WO1994015304A1 - Method and apparatus for transforming an arbitrary topology collection of nodes into an acyclic directed graph - Google Patents

Method and apparatus for transforming an arbitrary topology collection of nodes into an acyclic directed graph Download PDF

Info

Publication number
WO1994015304A1
WO1994015304A1 PCT/US1993/012317 US9312317W WO9415304A1 WO 1994015304 A1 WO1994015304 A1 WO 1994015304A1 US 9312317 W US9312317 W US 9312317W WO 9415304 A1 WO9415304 A1 WO 9415304A1
Authority
WO
WIPO (PCT)
Prior art keywords
node
nodes
cycle
adjacent
parent
Prior art date
Application number
PCT/US1993/012317
Other languages
French (fr)
Inventor
Florin Oprescu
Original Assignee
Apple Computer, Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Apple Computer, Inc. filed Critical Apple Computer, Inc.
Priority to DE69331705T priority Critical patent/DE69331705T2/en
Priority to JP6515318A priority patent/JPH08507623A/en
Priority to AU59541/94A priority patent/AU5954194A/en
Priority to EP94905424A priority patent/EP0674790B1/en
Publication of WO1994015304A1 publication Critical patent/WO1994015304A1/en

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00Data switching networks
    • H04L12/28Data switching networks characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks]
    • H04L12/40Bus networks
    • H04L12/40052High-speed IEEE 1394 serial bus
    • H04L12/40078Bus configuration
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F15/00Digital computers in general; Data processing equipment in general
    • G06F15/16Combinations of two or more digital computers each having at least an arithmetic unit, a program unit and a register, e.g. for a simultaneous processing of several programs
    • G06F15/163Interprocessor communication
    • G06F15/173Interprocessor communication using an interconnection network, e.g. matrix, shuffle, pyramid, star, snowflake
    • G06F15/17337Direct connection machines, e.g. completely connected computers, point to point communication networks
    • G06F15/17343Direct connection machines, e.g. completely connected computers, point to point communication networks wherein the interconnection is dynamically configurable, e.g. having loosely coupled nearest neighbor architecture
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F15/00Digital computers in general; Data processing equipment in general
    • G06F15/16Combinations of two or more digital computers each having at least an arithmetic unit, a program unit and a register, e.g. for a simultaneous processing of several programs
    • G06F15/163Interprocessor communication
    • G06F15/173Interprocessor communication using an interconnection network, e.g. matrix, shuffle, pyramid, star, snowflake
    • G06F15/17356Indirect interconnection networks
    • G06F15/17368Indirect interconnection networks non hierarchical topologies
    • G06F15/17381Two dimensional, e.g. mesh, torus

Definitions

  • the present invention relates to computer systems. More particularly, the present invention relates to a method and apparatus for establishing and utilizing a communications scheme between a plurality of arbitrarily assembled elements of a computer system.
  • a bus is a communications path, such as a wire or wires, running throughout the computer system.
  • Each component of the system need only plug into the bus to be theoretically connected to each of the other components in the system. Of course, each component cannot simultaneously communicate with other components because there may be only a single communications channel between the components. It is necessary when utilizing a communications bus to establish some form a sharing arrangement so that each component may use the bus to communicate with other components in an efficient manner that does not leave critical pieces of information from one component hanging, waiting for bus access.
  • the method by which components on the bus share the bus is generally referred to as a bus arbitration scheme.
  • the physical (and logical/electrical) configuration of the bus itself can and should be optimized to minimize system delays while remaining as flexible as possible.
  • each component In order to communicate with other components attached to a bus, each component must be equipped with hardware such as transmitting and receiving circuitry compatible with the communications protocol implemented for the bus .
  • One such communications standard is described in IEEE Standards Document P1394 entitled "High Performance Serial Bus ", said document attached as Appendix A.
  • the standard described in P1394 is intended to provide a low cost interconnect between cards on the same backplane, cards on other backplanes, and external peripherals.
  • Prior art buses or networks required knowing what was being plugged in where. For example, the back of many computers have specified ports for specific peripherals. Some computers implement several buses, such as the Macintosh which uses a bus referred to as ADB for components like a mouse and keyboard and SCSI bus for other peripherals. These types of buses provide for daisy chaining elements together but connections are of limited topology. Other known buses/networks require that the nodes of the network be arranged in a ring, a loop which must be closed in order to operate. Finally, star, or hub- and-spoke arrangements re-quired that each node be directly linked to a central master. Each of the prior art systems lacks a desirable measure of flexibility.
  • a system and method which take an arbitrarily assembled collection of nodes on a bus or network and imposes an optimized hierarchical tree structure where there is only one root node.
  • Nodes having both parent and child nodes are considered branch nodes while nodes having only parent nodes are leaf nodes .
  • Loops or cycles in the physical topology are resolved into a logical topology that is acyclic and directed.
  • a signaling scheme is developed in which nodes, via on board communications hardware, signal all connected nodes and respond accordingly until hierarchical relationships are established. Cycles are resolved by intelligently breaking links to yield an acyclic graph.
  • Direction is established by each node recognizing its parent/child status with respect to connected nodes until a single node is established as a root node.
  • Figure 1 illustrates a block diagram of the hardware layer implementation utilized in accordance with the present invention.
  • Figures 2 (a) -2 (b) illustrate arbitrarily assembled collection of nodes, one being acyclic and the other including multiple cycles.
  • Figure 3(a) is the arbitrarily assembled collection of nodes of Figure 2(a) undergoing the graph transformation process in accordance with the present invention.
  • Figures 3 (b)-3(d) illustrate alternative communications exchanges between nodes in implementing the present invention.
  • Figure 3(e) graphically illustrates the directed graph resulting from the arbitrarily assembled network of nodes of Figure 2 (a) .
  • Figure 4 illustrates a symmetrical graph arrangement which requires resolving a root contention.
  • Figures 5(a)-5(d) illustrate a plurality of nodes being processed through the various steps of the cycle resolution procedure.
  • FIG(a)-6(f) illustrate the process flow for carrying out the graph transformation procedure in accordance with the preferred embodiment of the present invention.
  • Figure 7 illustrates a directed acyclic graph with a possible unique address assignment order indicated.
  • the present invention for defining the bus topology may be applied to any arbitrarily assembled collection of nodes linked together as in a network of devices.
  • One point that must be noted is that it is necessary to distinguish a node from a physical computer component.
  • Each component to reside on the bus will have associated with it at least one node physical layer controller.
  • a given component may advantageously be associated with multiple nodes but in the usual case there will be a one- to-one correspondence between devices or components on the bus and nodes.
  • the node 10 includes arbitration state machine logic 11.
  • This arbitration state machine logic incorporates all the logic circuitry for carrying out the methodologies and algorithms to be described herein.
  • the circuitry may comprise a programmable logic array (PLA) or be uniquely designed to carry out the functions described herein.
  • PPA programmable logic array
  • the node by means of its logic, shall implement the minimum arbitration protocol including the bus initialization, tree identification, self identification, and the bus arbitration functions, all to be described in detail further herein.
  • the node 10 shown in Figure 1 also includes transmission multiplexers 12 and 13 and data transmitter, receiver and resynchronizer 14.
  • the node illustrated in Figure 1 is coupled to local host 15.
  • Local host 15 may be any device one wishes to attach to the bus such as a disk drive, CPU, keyboard or any other component which needs to communicate with other components in the system.
  • the node 10 communicates with other nodes through communications links.
  • a link is a connection between two ports and in immediate practical terms is a cable segment but in general it may be implemented as any physical communication channel.
  • a link shall be able, at minimum. to provide a half duplex communication channel between the two ports which it connects .
  • a port is the interface between a node and a link.
  • a port must have the ability to transmit and receive data and arbitration signaling.
  • a port needs to be able to determine whether or not it is connected to another port through a link.
  • One method of facilitating this is by having connected ports apply a biasing voltage through the link which is detectable by the port at the other end of the link.
  • the port will determine that it is not a connected port.
  • the illustrated node 10 has three external ports 21, 22 and 23 with connecting links 17, 18 and 19, respectively.
  • a node may have one or more ports.
  • a node shall be able to transmit and receive data on any one of its ports.
  • a node shall be able to receive data on one and only one of its enabled ports at a time and be able to retransmit this data on all remaining enabled ports.
  • a node shall be able to receive and transmit signaling messages through all of its ports simultaneously and independently. Separate signaling transceivers, encoders and decoders are required for each port of a node.
  • a minimum implementation node does not require a local host device. For example, such a node may function as a cable extension. From hereon devices and local hosts will be ignored and all references to bus topology will refer to nodes and node connections through various ports .
  • FIGS 2(a) and 2(b) illustrate arbitrarily assembled collections of nodes. From hereon, nodes will be illustrated merely as circles, but are deemed to each incorporate elements equivalent to those described with respect to Figure 1. Note, however, that each node may have more or less than the three external ports shown in that figure.
  • the lines illustrated connecting each of the nodes are the method by which links are shown. Ports are not illustrated, but are impliedly the interface where a link and a node connect .
  • the bus arbitration methodology to be described herein requires that the arbitrary topology be resolved into an acyclic directed graph.
  • an arbitrary topology graph a collection of nodes and links may form a cycle.
  • a cycle exists if starting from a specific node in the graph it is possible to return to the same node by traversing links and nodes without any link being traversed twice.
  • Figure 2(a) illustrates an acyclic graph because none of the nodes illustrated are connected within a loop.
  • Figure 2(b) is not an acyclic graph because the region in bounding box 25 contains a collection of nodes, 40-47 which form multiple cycles.
  • the bus arbitration methodology to be described requires that there be no cycles, so a method of cycle resolution will also be described further herein.
  • a directed graph is one in which a hierarchical arrangement has been established between adjacent nodes . Initially, there are no established parent-child relationships between nodes. That is, for example, node 31 may be the "parent node” for node 34, or be the "child node” for node 34. Thus, it is necessary to take a given arbitrary topology graph and transform it into an acyclic and- directed graph. The methods described herein will work to perform this transformation for any give arbitrary topology, regardless of the number of nodes or how they are physically linked and regardless of the signaling propagation time along the links.
  • Figure 3(a) shows the arbitrary graph of Figure 2 (a) wherein the nodes and links have status labels and communicated signals are indicated for the graph transformation process for directing a graph. It is instructive at this point to describe signal communications between nodes.
  • Figure 3(b) illustrates two nodes 50 and 51 (hereinafter node A and node B, respectively) coupled by link 52. As described, the link is the communications channel coupling transceiver ports of the respective nodes as described above with reference to Figure 1. During the graph transformation process, it becomes necessary for nodes to establish parent-child relationships with adjacent nodes .
  • Two nodes are said to be adjacent nodes if there is at least one link connected between a port of the first node and a port of the second node.
  • the relationship to be resolved is that node B is the parent of node A and that it is appropriate for the nodes to establish that relationship.
  • node A Prior to a direction being established, when it become appropriate for node A to establish node B as its parent, node A will transmit from its port to which link 52 is coupled the signal "You Are My Parent” (YAMP) .
  • This message content may take any form, so long as node A knows that it is signaling YAMP and node B is capable of understanding that the received message is YAMP.
  • YAMP signal 53 When YAMP signal 53 is received by node B, node B will respond to node A by sending "You Are My Child” (YAMC) through link 52 to node A.
  • the arbitration state machine logic 11 of node A will keep track of the time delay between sending YAMP signal 53 and receiving YAMC signal 54. The time measured signifies twice the propagation delay between nodes A and B.
  • node A Upon receiving the YAMC signal, node A will respond with a "You Are My Child Acknowledged" (YAMCA) signal 55.
  • YAMCA You Are My Child Acknowledged
  • This provides node B with the ability to also determine the propagation time delay between the nodes equal to the time delay between sending YAMC and receiving YAMCA.
  • the YAMCA message also has the effect of properly orienting the communication channel.
  • the three logical messages, YAMP, YAMC and YAMCA can alternatively be relayed by only two signal transmissions. In Figure 3(c) this situation is illustrated where node A asserts the YAMP signal 56 continuously until it receives the return YAMC signal 57.
  • the YAMCA signal is logically transmitted to node B when the YAMP signal is detected as no longer arriving.
  • this described triple asynchronous message exchange provides a mechanism by which both nodes involved in the message exchange can determine the propagation time delay through the link. This delay value is used in resolving contention events to be described further herein as well as during normal bus arbitration to optimize bus performance. The dynamic extraction of this parameter is not mandatory. As an alternative a maximum propagation time delay can be apriory defined at the expense of optimum bus performance.
  • node A and B have exchanged messages signifying that node B is the parent of node A, the link can be said to be directed.
  • a short hand graphical notation is illustrated in Figure 3(d) where the direction arrow 58 indicates that node B is established as the parent of node A and the link is directed.
  • a "leaf" node is defined as a node with only one connected port.
  • a node recognizes its status as a leaf node as soon as it is initialized after power-up or other bus initialization.
  • a "branch" node is a node which has a least two connected ports. Through all but one of the connected ports a branch node will have received the YAMP signal and have acknowledged it. Through its remaining port, a branch node has sent the YAMP signal thus establishing that it has a parent node.
  • a node does not achieve branch status until it has established that it has one parent (a node can have only one parent node) and all its other ports are connected to child nodes. Prior to achieving branch status, a node is considered a "cycle" node because until it is determined to be a branch the possibility exists that the node is part of a cycle which makes establishing direction impossible.
  • the graph transformation procedure begins at step 60 upon bus initialization (power-up or instigated) at which time the leaf nodes in the arbitrary topology recognize at step 61 and label themselves at step 68 as leaf nodes by determining that they have only one connected port at decision box 66.
  • nodes 33, 35, 36 and 37 are leaf nodes which, once initialized, at step 69 each transmits the YAMP signal through its only connected port to its adjacent node. The nodes receiving these signals will then propagate the YAMC signals back to the leaf nodes at step 70, thus establishing a direction for the given link between respective parent-child pairs when the YAMCA communication is completed.
  • each leaf node labels its one connected port as a parent port and each receiving port on the parent node is labeled a child port .
  • the nodes on the graph which are not initially leaf nodes are initially considered "cycle nodes" for the reason described above and proceed according to the Cycle Node Procedure 63. Any cycle node which has labeled all but one of its connected ports as child ports then propagates the YAMP signal from its remaining unlabeled port at step 85. When that direction is established for the link, the cycle node then becomes labeled a branch node.
  • node 34 has only one unlabeled port (having labeled the link connection to node 37 as being through a child port) so node 34 broadcasts the YAMP signal to node 31, resulting in node 34 becoming a branch node.
  • node 31 broadcasts the YAMP signal to node 30.
  • the YAMP signal at decision box 75, that node becomes the root node.
  • Figure 3(a) after node 30 has received the YAMP signals from nodes 31 and 32, its label changes from being a cycle node to being the root node.
  • FIG. 3(e) shows the resulting directed graph in response to the communicated signals shown in Figure 3(a) with each node labeled and the directions indicated by dark arrows.
  • a root contention situation may arise. This may happen for example in the case where the arbitrary topology has a symmetrical arrangement to it such as that shown in Figure .
  • nodes 160 and 161 have each established that it is a parent to the two leaf nodes to which it is coupled. Then, each has propagated the YAMP signal to the other at nearly the same time.
  • the root contention situation is recognized by both nodes involved at decision box 86. Each node is receiving a signal which designates it as a parent while it has sent the same signal out through the same port.
  • Each of the contending nodes responds to the other with the YAMC signal at step 91 which allows each to determine the "decision time period" which is equal to twice the propagation time between the nodes.
  • the root contention situation is resolved by utilizing a random decision mechanism incorporated on each arbitration state machine logic unit 11 of each node. For every "decision time period" that elapses, each node will randomly decide at step 92 (with a 50% probability) whether to again transmit the YAMP signal to the other. Almost certainly within a finite number of the cycles, one node will decide to designate the other its parent without that one reciprocating. The one that is designated the parent becomes the root at step 95.
  • predetermined selection criteria values may be assigned to nodes, the larger or smaller determining which dominates in a contention event .
  • the dynamic determination of the "decision time period”, while it offers optimum performance is not essential in implementing the present invention.
  • an apriory defined “decision time period” may be used as long as it is greater than the worst case link propagation that can be encountered in any bus using this algorithm. The same method used to resolve root contentions will also be used to resolve other contention events to be described further herein.
  • the result of the graph transformation process is the assignment of the root attribute to one and only one node in the graph.
  • the root node will have the ultimate decision in the bus arbitration scheme to be described and therefore can access the bus with maximum priority without the use of special priority time gaps. It is often desirable to be able to assign the root property to a predetermined node either when it is manufactured or dynamically (during run time) to optimize a given system.
  • a given bus may include a node which requires isochronous data transfer. Isochronous data is data that must be transmitted at a given time to be of any value. For example, music from a compact disk needs to be transferred and output in the order in which it is to be heard and with no significant delays, unlike data files which may be transferred piecemeal and not necessarily in order.
  • Nodes can be classified into three categories with respect to root designations. These designations may be applied during manufacturing by hard-wiring the designation into the device, programming the arbitration state machine logic or by higher level software making the decision then initiating a reboot while preserving that decision.
  • the three designations that a node may be assigned with respect to being designated a root are: nodes that do not want to be root, nodes that may (should) be root and nodes that shall be root. These designations are tested for at steps 81 and 83.
  • a node designated in the first category will begin the graph transformation procedure as soon as it is directed to do so. This will usually be immediately following the completion of the bus initialization procedure.
  • a node from the second category will delay the beginning of the graph transformation procedure for a predetermined amount of time after it is directed to begin the procedure at step 84. By this delay, the node increases its chance of becoming the root. (The YAMP signals are more likely to propagate to it due to the delay.) Despite the added delay, it is still possible that a "may be root" node will not wind up being designated the root . This will depend on the given topology and message propagation delays. The amount of delay can be defined during design to be greater than a reasonable worst case propagation delay through a fairly complex graph.
  • a node from the third category of root designation possibilities may only recognize the fact that it must be the root after the graph has already been transformed and all nodes have identified themselves.
  • the arbitration state machine logic may make this determination or software running on the host system. When this occurs, the node that has to be root agrees with all other nodes on the bus that it is going to be the one and only root and restarts the graph transformation process by signaling a preemptive bus initialization signal which is described further herein.
  • the node then waits at step 82 to become the root and does not participate in the graph transformation until it has received the YAMP signal on all of its ports, thus forcing it to be designated the root.
  • the graph can be said to be directed. There is a defined relationship existing between all adjacent nodes on the graph.
  • the "cycle detect" time-out event does not have to occur synchronous for all nodes of a graph as all message exchanges are asynchronous events . As such, it is possible for a node which has not yet reached its "cycle detect" time-out event to receive a message indicating that cycle resolution is ongoing. Such a node will terminate its cycle detect time-out interval and begin the appropriate cycle resolution process.
  • the method of cycle resolution may require different iteration processes depending upon the graph topology. Essentially there are two types of cycles, simple cycles and multiple cycles. Multiple cycles exist where there are nodes that meet the definition of multiple cycles nodes.
  • a multiple cycles node is a cycle node which has at least three connected ports through which it has not received the YAMP signal.
  • a multiple cycles node has at least three adjacent cycle nodes. It will belong to either at least two distinct graph cycles, or to at least one graph cycle and to a branch wherein there is a node having the "shall be root" designation (which therefore has not propagated a YAMP signal) . These ports are labeled by the node as cycle ports.
  • a node may only be considered a multiple cycles node during the cycle resolution process.
  • a single cycle node is one which has two and only two connected ports through which it has not yet received the YAMP signal.
  • FIG. 5(a) illustrates in greater detail the portion 25 of the arbitrary topology shown in Figure 2(b) .
  • nodes 40-47 form multiple cycles because there are two nodes, 43 and 47, which are linked to more than two other cycle nodes.
  • Figure 6(f) illustrates the process flow for multiple cycles resolution.
  • step 106 all multiple cycles nodes transmit through all their cycle ports a signal indicating "You Are My Cycle Parent" (YAMCP) . If the YAMCP signal generates a reply conveying the message "You Are My Cycle Child” (YAMCC) , the node responds with a "You Are My Cycle Child Acknowledged” (YAMCCA) at step 108 and labels the corresponding port as a cycle parent port.
  • YAMCP You Are My Cycle Parent
  • a cycle contention event has occurred such as will be the case between nodes 43 and 47.
  • the cycle contention event is resolved in the same manner as a root contention event wherein for each "decision time period" the contending nodes randomly (50% likelihood) each send or do not send the YAMCP signal until a decision time interval in which one of the two ports involved becomes a cycle parent port while the other one recognizes its status as a cycle child port.
  • a.single cycle node receives through one or more of its ports the YAMCP signal, it replies with the YAMCC signal and awaits the return of a YAMCCA. After these exchanges, the ports are labeled cycle child ports.
  • a single cycle node When a single cycle node has labeled one of its two cycle ports as a cycle child port, it conveys the YAMCP signal through its remaining cycle port at step 109. If the YAMCC signal is received as a reply without interference, then the node sends the YAMCCA and labels the port a cycle parent port. If, on the other hand, a contention situation occurs with the node receiving a YAMCP signal after transmitting the same, the cycle contention event is resolved at step 110 in exactly same manner as described above where other contentions between two nodes are resolved.
  • a single cycle node winds up with both of its ports labeled as cycle child ports, one of the two links leading to the node must be redirected at step 111.
  • the single cycle node selects one of its two connected cycle child ports, based upon an apriory defined criterion, and transmits the YAMCP signal. There will be no contention event, the receiving port with reply with the YAMCC to which the selected port replies YAMCCA and relabels the port that was a cycle child port as a cycle parent port.
  • every single cycle node will have one and only one port labeled as a cycle child port.
  • a single cycle node receives the YAMCP signal through a port that was previously labeled as a cycle parent port, it shall respond with the YAMCC and wait for the YAMCCA at which time the port is relabeled as a cycle child port. That single cycle node will then have both cycle ports labeled as cycle child ports. The node will then select the port that has the older cycle child port label and transmit the YAMCP signal, thus reversing the direction of the link to that port. This process will continue until all single cycle nodes have one cycle parent node and one cycle child node. It should be emphasized that the criteria used for port relabeling should follow one of two rules.
  • any apriory defined criterion is acceptable (always choose port X over port Y, etc.) . If the ports have been labeled at different times the relabeling process should always involve the port which has been labeled a cycle child port the earliest in time.
  • the method of a priory criterion selection may be a logical or hard-wired ordering of ports on a node. For example, each node may elect to make port 1 a higher priority than port 2 and port 2 a higher priority than port 3, etc. The node merely needs a rule for selecting one port in favor of another and to consistently apply that rule.
  • At least one multiple cycles node will have at least one of its ports labeled a cycle child port either as a result of a contention with another multiple cycles port or as a result of relabeling message exchange.
  • a multiple cycles node may end up with more than one cycle port labeled as a cycle child port such as node 43 in Figure 5(b) which shows the direction of all cycle links resulting from the above steps.
  • the multiple cycles node will select all but one of its cycle child ports and transmits through it the "link disabled" (LD) message.
  • LD link disabled
  • the ports Upon receiving the "link disabled completed” (LDC) reply signal the ports are labeled by the respective nodes as disabled child ports and removed from the logical bus.
  • a cycle node (single or multiple) upon receiving on one of its ports labeled a cycle parent port the LD signal shall reply with the LDC signal, label the port a disabled parent port and remove it from the logical bus.
  • the link between nodes 42 and 43 or between 43 and 47 will be disabled, breaking the multiple cycles into a simple cycle.
  • the method of selecting which link(s) to disable may be apriory defined as described above for selecting redirecting propagation. If the direction of the links in the 47-46-45-44-43 chain had been reversed, the breaking of two of the links to 43 would have eliminated any simple cycles as well.
  • a cycle node shall transmit on one of its two cycle ports, the YAMCP signal. If this results in any contention with an adjacent cycle node, the contention will be resolved in exactly the same manner as described above for resolving other contentions. If a node receives the YAMCP signal before initiating its own transmission, it will abort its own transmission and respond with the YAMCC signal. Upon receiving the reply YAMCCA signal, it shall label the port as a cycle child port and then transmit through the remaining port the YAMCP signal. In this way at no time will any cycle node have more than one cycle parent port.
  • FIG. 5(d) illustrates this case where a single cycle graph has a dangling "shall-be-root" node, node 5. Though this is a single cycle graph, node 3 believes it is a multiple cycles node and begins the multiple cycle resolution process . At the end of the multiple cycle resolution process the cycle will be eliminated. One node will be left with only one cycle port, hence the test of decision box 102 to end the cycle resolution process with that node becoming a branch node.
  • Figure 5(c) shows the nodes 40-47 with the link between nodes 42 and 43 depicted as logically not being present. Also indicated is the remaining directed loop encompassing nodes 43-47.
  • Each node will transmit its randomly selected message to its parent or parents and receive from its child the child's randomly selected message or the message forwarded by its child if the child has become a forwarding node. Every cycle node electing to transmit the CMB signal, but receiving a CMA signal will consider itself eliminated from the race to become the root cycle node and will from thereon merely forward any message received including the CMA message which has triggered the transformation.
  • the time delay between successive messages generated and put out by a node which is not yet a forwarding node is a system parameter defined at design time. Its value shall be greater than the maximum propagation delay through a practical implementable cycle.
  • a node which does not become a forwarding node will output a predefined number of messages which is also a system parameter defined at design time. This number shall be large enough to guarantee probabilistically, with the desired level of certainty that at the end of this process one and only one node in the cycle remains a non-forwarding node. This remaining node will then be considered the root cycle node .
  • the root cycle node will then send the LD signal through its child port at step 104 in the manner described above, thus logically breaking the link between the root cycle node and its child. This, therefore, breaks the cycle and when all cycles are broken the overall graph is acyclic.
  • root cycle node breaks the link with its child node is to accommodate the possibility that there is a a waiting "shall be root” node appended to the root cycle node. It should be noted that once a cycle is a directed cycle, cycle nodes will only have a single cycle child but may have more than one cycle parent (as in the case of the waiting "shall be root” node) . A shall be root dangling node or chain of nodes will have the attribute of being a cycle parent node to the root cycle node, a link to which should not be broken. Once each of the loops is broken, and no cycles remain, the procedure for transforming the graph as described in the earlier sections may proceed until the entire graph is both acyclic and directed.
  • the method for selecting which bus request to grant may be an apriory decision such as that described above where, for example, ports are selected from left to right or based on port numbering, etc.
  • the Bus Grant (BG) signal will be transmitted from the root to its requesting child. If that requesting child is itself a parent node which has propagated the bus request from one of its children, it will send the bus denial signal through all but one of its child ports in the same predetermined fashion as described above. Eventually one leaf node will receive the bus grant signal to which it will reply with a Bus Grant Acknowledged (BGA) signal which will be propagated back to the root node.
  • BGA Bus Grant Acknowledged
  • the node which is finally granted access to the bus will transmit an address assignment packet.
  • the node will transmit this packet on the bus and it will be received by all other nodes, each of which will count the number of address packets they receive.
  • the transmitted address packet may have any arbitrary information.
  • a node's unique physical address will be based on the number of address packets a node has counted before it transmits an address packet. Thus, no two nodes will acquire the same physical address despite not having address information assigned in advance.
  • the actual composition of the address packet is arbitrary and may be any bit stream efficiently utilizable by the system.
  • a node After transmitting the physical address assignment packet, a node will transmit a "Child ID Completed" signal (CIC) signal.
  • CIC Channel ID Completed
  • the parent node receiving this on its child port will then transmit the "Child Identification Completed Acknowledgment" (CICA) signal and label the port as an identified child port .
  • CICA Child Identification Completed Acknowledgment
  • the parent of the node which has just identified itself will then select its next child to transmit the physical address packet .
  • the parent node will request the bus and, when granted the bus, will propagate its physical address assignment packet. This procedure will continue following the predetermined selection criteria until all nodes determine a unique physical address assignment by counting.
  • Figure 7 illustrates the graph of Figure 3(e) in which a left-to-right predefined selection criteria is implemented.
  • the nodes are uniquely assigned addresses where node 33 receives the first and as described, the root node 30 receives the eighth and last address.
  • each node in the graph will have a unique physical address, which need not have been determined in advance and which may be utilized for system management or other purposes .
  • node self identification essentially follows the same routine as the physical address assignment procedure described above.
  • that packet may include further information such as identification of the physical device comprising the identification of the local host related to the node, how much power it requires, and, for example, whether it supports a "soft power-on" attribute, etc.
  • the node self-identification information may serve as the physical address assignment packet because the practice of sending any information at all provides the basis for counting to yield unique physical addresses.
  • topology mapping follows along the same lines as physical address assignment and node self- identification. This procedure thus has each node, when it is going through the address assignment or node self- identification process, further transmit information concerning all of its ports such as the number of child ports it has and whether or not it has any disabled ports. With respect to disabled ports, it may be desirable to implement a communication protocol between ports that are disabling so that they can identify from whom they are disabled. Thus, when a port identifies a disabled port it will give an identifier indicating its own ID as well as the port ID from which it has become disabled.
  • the bus server, host or any software level application may logically reconstruct the resolved bus topology. This is useful for many purposes including implementing redundancy where if a link unexpectedly goes down, previously disabled links may serve to prevent the loss of communication channels to any nodes. Fair Bus Access Arbitration
  • One arbitration scheme implemented in accordance with the present invention is that of fair bus access.
  • a node desires access to the bus, it sends through its parent port (unless it is the root) a bus request (BR) signal.
  • the parent upon receiving the BR signal from one child sends a bus denied signal (BD) through all its other child ports.
  • BD bus denied signal
  • the parent propagates the BR signal upward through its parent until the signal reaches the root.
  • the root issues a bus grant signal (BG) responsive to the first BR signal it receives and sends the BD signal through all of its other child ports which propagate downward thereby orienting the links.
  • BG bus grant signal
  • the BG signal propagates downward through the graph until it reaches the requesting node which then sends Bus Acknowledge (BA) signal followed by the packet of information that the node needed to send on the bus.
  • BA Bus Acknowledge
  • the predetermined selection criteria for the root node will be used for granting to one of the nodes bus access . This may be the same predetermined priority selecting criteria as described above.
  • a further aspect of fair bus access arbitration is that a parent node has priority over its children.
  • a parent node wants the bus, it sends the BD signal through all of its child ports, then propagates the BR signal up toward the root.
  • One potential problem with this mechanism is that if the parent has a large quantity of information to transmit on the bus a child node may have trouble getting adequate bus access. There is therefore introduced a gap system which is widely used and well-known in the art. After a node has utilized the bus, the node must wait for one gap period before it can again request the bus. This gives equal chance of being granted the bus to every node on the bus independent of its topological placement on the bus.
  • the length of the gap has to be greater than the worst case signal propagation delay through the bus.
  • the gap value can be predetermined and hard-wired into the node logic but such an approach will result in all but the most extreme case in a sub optimal utilization of the bus.
  • the topology mapping capability together with the measurement of the propagation delay between adjacent nodes performed during the graph transformation phase enables the calculation of an optimal fair gap that will maximize the bus performance for any specific implementation.
  • the root node may grant the bus to itself at any time. This is done by first sending the BD signal down through all of the nodes in the graph. Priority bus access for the root is very useful for the case where the root node is required to perform isochronous data transfer.
  • token passing bus access refers to the notion that a node may communicate on a bus when it is in possession of a token that is passed between nodes.
  • the token is passed from node to node in a cyclic fashion so that each node receives the bus in a predetermined point in the cycle.
  • Token passing is implemented in the present invention in following the same manner as the physical address assignment routine described above.
  • the predetermined selection mechanisms implemented are used to select the order in which the token will be passed from node to node. This order resembles the order as shown in Figure 7 which dictates the order of unique address assignment.
  • Each node, when it is assigned the token will propagate its information packet on the bus while the remaining nodes listen. The node will then pass the token to the next logical node based on the predetermined sequencing method as described above.
  • the state machine logic incorporated on each node is capable of triggering a bus initialization (BI) signal to be propagated from the node through all of its ports upon certain conditions.
  • BI bus initialization
  • a preemptive bus initialization There are a number of situations which may make it necessary or desirable to trigger a preemptive bus initialization. First, this may be a node response to an unforeseen error. Additionally, at the host level, it may be determined that a different node should acquire the root attribute, for example, an isochronous data transfer node. This assignment will be preserved throughout the bus initialization routine thereby causing the desired node to wait during the transform procedure until it receives the root designation. Another condition leading to a preemptive bus initialization may be the breakage of a link, in which case it may be necessary to calculate a new acyclic directed graph for the attached nodes .
  • a preemptive bus initialization should occur is when a device is added to the network, referred to as "hot addition" of peripherals.
  • the port to which a new device is connected will detect the presence of a new node and trigger a bus initialization which will be transparent to the user of the system but which allows the addition and subtraction of peripherals, for example, without having to shut down and repower.
  • a new acyclic directed graph is calculated which includes the presence of the added node. It is possible that upon removing certain nodes, it will not be necessary to trigger a bus initialization, for example, when a leaf node is removed, there is no harm to the network. However, if a branch node is disengaged from an operating bus, it is likely to be necessary to reconfigure the graph.

Abstract

A system and method are described which take an arbitrarily assembled collection of nodes on a bus or network and imposes an optimized hierarchical tree structure where there is only one root node. Nodes having both parent and child nodes are considered branch nodes while nodes having only parent nodes are leaf nodes. Loops or cycles in the physical topology are resolved into a logical topology that is acyclic and directed. A signaling scheme is developed in which nodes, via on board communications hardware, signal all connected nodes and respond accordingly until hierarchical relationships are established. Cycles are resolved by intelligently breaking links to yield an acyclic graph. Direction is established by each node recognizing its parent/child status with respect to connected nodes until a single node is established as a root node.

Description

METHOD AND APPARATUS FOR TRANSFORMING AN ARBITRARY TOPOLOGY COLLECTION OF NODES INTO AN ACYCLIC
DIRECTED GRAPH
BACKGROUND QT TSB I VENTION
Related Applications
This application is related to applications Serial No, 07/994,402, entitled "Method and Apparatus for Unique Address Assignment,. Node Self-Identification and Topology Mapping for a Directed Acyclic Graph ", Serial No. 07/994,983, entitled "Method and Apparatus for Arbitrating on an Acyclic Directed Graph " and Serial No. 07/994,117, entitled "Method and Apparatus for Transforming an Arbitrary Acyclic Topology Collection of Nodes into an Acyclic Directed Graph " each of which is assigned to the assignee of the present application and filed concurrently herewith.
Field of the Invention
The present invention relates to computer systems. More particularly, the present invention relates to a method and apparatus for establishing and utilizing a communications scheme between a plurality of arbitrarily assembled elements of a computer system.
Background
Components within a given computer system need the ability to convey signals amongst themselves. In very simple systems, it is possible to have each element of the system directly wired to all of the other parts of the system. However, in reality, in order to make computers expandable and to accommodate an unknown number of system parts, computer architects long ago developed the concept of a communications bus.
A bus is a communications path, such as a wire or wires, running throughout the computer system. Each component of the system need only plug into the bus to be theoretically connected to each of the other components in the system. Of course, each component cannot simultaneously communicate with other components because there may be only a single communications channel between the components. It is necessary when utilizing a communications bus to establish some form a sharing arrangement so that each component may use the bus to communicate with other components in an efficient manner that does not leave critical pieces of information from one component hanging, waiting for bus access. The method by which components on the bus share the bus is generally referred to as a bus arbitration scheme.
In addition to the critical need to optimize the bus arbitration scheme so as to maximize the flow of important information, the physical (and logical/electrical) configuration of the bus itself can and should be optimized to minimize system delays while remaining as flexible as possible.
In order to communicate with other components attached to a bus, each component must be equipped with hardware such as transmitting and receiving circuitry compatible with the communications protocol implemented for the bus . One such communications standard is described in IEEE Standards Document P1394 entitled "High Performance Serial Bus ", said document attached as Appendix A. The standard described in P1394 is intended to provide a low cost interconnect between cards on the same backplane, cards on other backplanes, and external peripherals.
Prior art buses or networks required knowing what was being plugged in where. For example, the back of many computers have specified ports for specific peripherals. Some computers implement several buses, such as the Macintosh which uses a bus referred to as ADB for components like a mouse and keyboard and SCSI bus for other peripherals. These types of buses provide for daisy chaining elements together but connections are of limited topology. Other known buses/networks require that the nodes of the network be arranged in a ring, a loop which must be closed in order to operate. Finally, star, or hub- and-spoke arrangements re-quired that each node be directly linked to a central master. Each of the prior art systems lacks a desirable measure of flexibility.
It would be desirable, and is therefore and object of the present invention, to be able to arbitrarily assemble elements of a computer system onto a bus where the arbitrary topology can be resolved by the system into a functioning system without requiring a predetermined arrangement of components. SUMMARY OF THE INVENTION
It is an object of the present invention to provide enhanced computer system performance by optimizing the way the various system components communicate regardless of the physical topology.
It is also an object of the present invention to provide a method for transforming an arbiting topology graph having simple or multiple cycles into an acyclic graph.
It is another object of the present invention to provide a method for transforming an acyclic graph into a directed acyclic graph.
It is still another object of the present invention to provide a method for transforming any arbitrary topology graph into an acyclic directed graph, said acyclic directed graph being useful for implementing computer system bus and network communications schemes.
These and other objects of the present invention are provided by a system and method which take an arbitrarily assembled collection of nodes on a bus or network and imposes an optimized hierarchical tree structure where there is only one root node. Nodes having both parent and child nodes are considered branch nodes while nodes having only parent nodes are leaf nodes . Loops or cycles in the physical topology are resolved into a logical topology that is acyclic and directed. A signaling scheme is developed in which nodes, via on board communications hardware, signal all connected nodes and respond accordingly until hierarchical relationships are established. Cycles are resolved by intelligently breaking links to yield an acyclic graph. Direction is established by each node recognizing its parent/child status with respect to connected nodes until a single node is established as a root node.
BRIEF DESCRIPTION OF TBE DRAWINGS
The objects, features and advantages of the present invention will be apparent from the following detailed description in which:
Figure 1 illustrates a block diagram of the hardware layer implementation utilized in accordance with the present invention.
Figures 2 (a) -2 (b) illustrate arbitrarily assembled collection of nodes, one being acyclic and the other including multiple cycles.
Figure 3(a) is the arbitrarily assembled collection of nodes of Figure 2(a) undergoing the graph transformation process in accordance with the present invention.
Figures 3 (b)-3(d) illustrate alternative communications exchanges between nodes in implementing the present invention.
Figure 3(e) graphically illustrates the directed graph resulting from the arbitrarily assembled network of nodes of Figure 2 (a) .
Figure 4 illustrates a symmetrical graph arrangement which requires resolving a root contention. Figures 5(a)-5(d) illustrate a plurality of nodes being processed through the various steps of the cycle resolution procedure.
Figure 6(a)-6(f) illustrate the process flow for carrying out the graph transformation procedure in accordance with the preferred embodiment of the present invention.
Figure 7 illustrates a directed acyclic graph with a possible unique address assignment order indicated.
DETAILED DESCRIPTION OF THE INVENTION
A method and apparatus for utilizing a bus having an arbitrary topology are described. In the following description, many specific details are set forth such as various computer components in order to provide a thorough understanding of the present invention. It will be obvious, however, to one skilled in the art that the present invention may be practiced without such specific details. In other instances, well-known control structures and coding techniques have not been described in detail in order not to obscure unnecessarily the present invention.
Throughout this detailed description, numerous descriptive terms are introduced to provide metaphorical clarity to the description. For example, frequent references will be made to parent-child relationships between nodes in a given topology. The purpose of this is to provide the concept of "direction" to the finally resolved graph. As will be described, once an arbitrary topology has been reduced to an acyclic directed graph, there will be one node identified as the "root" node. The root node will not have a parent node, all nodes logically immediately adjacent to the root node are the child nodes of the root. The "tree" metaphor is completed by the inclusion of nodes referred to as "branches" and "leaves".
The bus architecture described herein, though described with reference to components for a single computer, in general has a broader scope. The present invention for defining the bus topology may be applied to any arbitrarily assembled collection of nodes linked together as in a network of devices. One point that must be noted is that it is necessary to distinguish a node from a physical computer component. Each component to reside on the bus will have associated with it at least one node physical layer controller. In certain circumstance, a given component may advantageously be associated with multiple nodes but in the usual case there will be a one- to-one correspondence between devices or components on the bus and nodes.
Referring now to Figure 1 a block diagram of a node 10 is illustrated. The physical implementation of a node is somewhat arbitrary. In the preferred embodiment implementation of the present invention, the nodes are designed to comply with the IEEE P1394 High Performance Serial Bus communications protocol which is attached as Appendix A. The node 10 includes arbitration state machine logic 11. This arbitration state machine logic incorporates all the logic circuitry for carrying out the methodologies and algorithms to be described herein. The circuitry may comprise a programmable logic array (PLA) or be uniquely designed to carry out the functions described herein. Those skilled in the art, once described the functions to be carried out by the node logic will be able to implement the present invention without undue experimentation. The node, by means of its logic, shall implement the minimum arbitration protocol including the bus initialization, tree identification, self identification, and the bus arbitration functions, all to be described in detail further herein.
The node 10 shown in Figure 1 also includes transmission multiplexers 12 and 13 and data transmitter, receiver and resynchronizer 14. The node illustrated in Figure 1 is coupled to local host 15. Local host 15 may be any device one wishes to attach to the bus such as a disk drive, CPU, keyboard or any other component which needs to communicate with other components in the system. The node 10 communicates with other nodes through communications links. A link is a connection between two ports and in immediate practical terms is a cable segment but in general it may be implemented as any physical communication channel. A link shall be able, at minimum. to provide a half duplex communication channel between the two ports which it connects . A port is the interface between a node and a link. In accordance with the present invention, a port must have the ability to transmit and receive data and arbitration signaling. A port needs to be able to determine whether or not it is connected to another port through a link. One method of facilitating this is by having connected ports apply a biasing voltage through the link which is detectable by the port at the other end of the link. Thus, if a port has a link attached which is not connected to a port at the other end, a naked link, the port will determine that it is not a connected port. In Figure 1, the illustrated node 10 has three external ports 21, 22 and 23 with connecting links 17, 18 and 19, respectively.
Some of the rules of implementation for nodes in order to implement the present invention are that a node may have one or more ports. A node shall be able to transmit and receive data on any one of its ports. A node shall be able to receive data on one and only one of its enabled ports at a time and be able to retransmit this data on all remaining enabled ports. A node shall be able to receive and transmit signaling messages through all of its ports simultaneously and independently. Separate signaling transceivers, encoders and decoders are required for each port of a node. A minimum implementation node does not require a local host device. For example, such a node may function as a cable extension. From hereon devices and local hosts will be ignored and all references to bus topology will refer to nodes and node connections through various ports .
Graph Transformation
Figures 2(a) and 2(b) illustrate arbitrarily assembled collections of nodes. From hereon, nodes will be illustrated merely as circles, but are deemed to each incorporate elements equivalent to those described with respect to Figure 1. Note, however, that each node may have more or less than the three external ports shown in that figure. The lines illustrated connecting each of the nodes are the method by which links are shown. Ports are not illustrated, but are impliedly the interface where a link and a node connect .
The bus arbitration methodology to be described herein requires that the arbitrary topology be resolved into an acyclic directed graph. In an arbitrary topology graph a collection of nodes and links may form a cycle. A cycle exists if starting from a specific node in the graph it is possible to return to the same node by traversing links and nodes without any link being traversed twice. Figure 2(a) illustrates an acyclic graph because none of the nodes illustrated are connected within a loop. Figure 2(b), however, is not an acyclic graph because the region in bounding box 25 contains a collection of nodes, 40-47 which form multiple cycles. The bus arbitration methodology to be described requires that there be no cycles, so a method of cycle resolution will also be described further herein.
In addition to the requirement that a graph be acyclic, it must also be directed. A directed graph is one in which a hierarchical arrangement has been established between adjacent nodes . Initially, there are no established parent-child relationships between nodes. That is, for example, node 31 may be the "parent node" for node 34, or be the "child node" for node 34. Thus, it is necessary to take a given arbitrary topology graph and transform it into an acyclic and- directed graph. The methods described herein will work to perform this transformation for any give arbitrary topology, regardless of the number of nodes or how they are physically linked and regardless of the signaling propagation time along the links.
Nfi-ifi Com-nτιni--a ion
Initially, the process of transforming an acyclic arbitrary topology graph into a directed graph will be described. The case where cycle resolution is required will follow. Figure 3(a) shows the arbitrary graph of Figure 2 (a) wherein the nodes and links have status labels and communicated signals are indicated for the graph transformation process for directing a graph. It is instructive at this point to describe signal communications between nodes. Figure 3(b) illustrates two nodes 50 and 51 (hereinafter node A and node B, respectively) coupled by link 52. As described, the link is the communications channel coupling transceiver ports of the respective nodes as described above with reference to Figure 1. During the graph transformation process, it becomes necessary for nodes to establish parent-child relationships with adjacent nodes . Two nodes are said to be adjacent nodes if there is at least one link connected between a port of the first node and a port of the second node. In Figures 3(b)- 3(d) it will be assumed that the relationship to be resolved is that node B is the parent of node A and that it is appropriate for the nodes to establish that relationship.
Prior to a direction being established, when it become appropriate for node A to establish node B as its parent, node A will transmit from its port to which link 52 is coupled the signal "You Are My Parent" (YAMP) . This message content may take any form, so long as node A knows that it is signaling YAMP and node B is capable of understanding that the received message is YAMP. When YAMP signal 53 is received by node B, node B will respond to node A by sending "You Are My Child" (YAMC) through link 52 to node A. The arbitration state machine logic 11 of node A will keep track of the time delay between sending YAMP signal 53 and receiving YAMC signal 54. The time measured signifies twice the propagation delay between nodes A and B. Upon receiving the YAMC signal, node A will respond with a "You Are My Child Acknowledged" (YAMCA) signal 55. This provides node B with the ability to also determine the propagation time delay between the nodes equal to the time delay between sending YAMC and receiving YAMCA. For half duplex communication links, the YAMCA message also has the effect of properly orienting the communication channel. For full duplex communications channels the three logical messages, YAMP, YAMC and YAMCA can alternatively be relayed by only two signal transmissions. In Figure 3(c) this situation is illustrated where node A asserts the YAMP signal 56 continuously until it receives the return YAMC signal 57. The YAMCA signal is logically transmitted to node B when the YAMP signal is detected as no longer arriving.
The use of this described triple asynchronous message exchange provides a mechanism by which both nodes involved in the message exchange can determine the propagation time delay through the link. This delay value is used in resolving contention events to be described further herein as well as during normal bus arbitration to optimize bus performance. The dynamic extraction of this parameter is not mandatory. As an alternative a maximum propagation time delay can be apriory defined at the expense of optimum bus performance.
Once nodes A and B have exchanged messages signifying that node B is the parent of node A, the link can be said to be directed. Node A within its logic labels its port to which link 52 is coupled as a parent port (it talks to a parent node) and node B labels its port to which link 52 is coupled a child port (it talks to a child node) . It is important to maintain the labels that ports achieve because the methods to be described below will be in terms of the labels assigned to nodes and ports at a given time. A short hand graphical notation is illustrated in Figure 3(d) where the direction arrow 58 indicates that node B is established as the parent of node A and the link is directed.
Direction Determination
Referring back now to Figure 3(a) and to process Figures 6(a)-6(f), the process of directing the overall arbitrary topology will now be described. It is necessary to introduce a few more colorful definitions to aid in explaining the topology transformation process. First, a "leaf" node is defined as a node with only one connected port. A node recognizes its status as a leaf node as soon as it is initialized after power-up or other bus initialization. A "branch" node is a node which has a least two connected ports. Through all but one of the connected ports a branch node will have received the YAMP signal and have acknowledged it. Through its remaining port, a branch node has sent the YAMP signal thus establishing that it has a parent node. A node does not achieve branch status until it has established that it has one parent (a node can have only one parent node) and all its other ports are connected to child nodes. Prior to achieving branch status, a node is considered a "cycle" node because until it is determined to be a branch the possibility exists that the node is part of a cycle which makes establishing direction impossible.
The graph transformation procedure begins at step 60 upon bus initialization (power-up or instigated) at which time the leaf nodes in the arbitrary topology recognize at step 61 and label themselves at step 68 as leaf nodes by determining that they have only one connected port at decision box 66. In the graph depicted in Figure 3(a), nodes 33, 35, 36 and 37 are leaf nodes which, once initialized, at step 69 each transmits the YAMP signal through its only connected port to its adjacent node. The nodes receiving these signals will then propagate the YAMC signals back to the leaf nodes at step 70, thus establishing a direction for the given link between respective parent-child pairs when the YAMCA communication is completed. At step 71 each leaf node labels its one connected port as a parent port and each receiving port on the parent node is labeled a child port . The nodes on the graph which are not initially leaf nodes are initially considered "cycle nodes" for the reason described above and proceed according to the Cycle Node Procedure 63. Any cycle node which has labeled all but one of its connected ports as child ports then propagates the YAMP signal from its remaining unlabeled port at step 85. When that direction is established for the link, the cycle node then becomes labeled a branch node. Thus, after leaf node 37 establishes that node 34 is its parent, node 34 has only one unlabeled port (having labeled the link connection to node 37 as being through a child port) so node 34 broadcasts the YAMP signal to node 31, resulting in node 34 becoming a branch node. Likewise, once node 31 has identified that nodes 33 and 34 are its children, node 31 broadcasts the YAMP signal to node 30. When one node has received through all of its ports the YAMP signal at decision box 75, that node becomes the root node. In Figure 3(a) after node 30 has received the YAMP signals from nodes 31 and 32, its label changes from being a cycle node to being the root node. In the graph of Figure 3(a), it is not necessarily the case that node 30 would become the root. If some of the links in the tree provided long propagation delays, node 30 might have received a YAMP signal on one port and then transmitted a YAMP signal through its other port. Any of the nodes may become the root, even a leaf, the root property taking precedence. Figure 3(e) shows the resulting directed graph in response to the communicated signals shown in Figure 3(a) with each node labeled and the directions indicated by dark arrows.
Root Contention
In certain circumstances a root contention situation may arise. This may happen for example in the case where the arbitrary topology has a symmetrical arrangement to it such as that shown in Figure . In the arbitrary graph illustrated in Figure 4, nodes 160 and 161 have each established that it is a parent to the two leaf nodes to which it is coupled. Then, each has propagated the YAMP signal to the other at nearly the same time. The root contention situation is recognized by both nodes involved at decision box 86. Each node is receiving a signal which designates it as a parent while it has sent the same signal out through the same port. Each of the contending nodes responds to the other with the YAMC signal at step 91 which allows each to determine the "decision time period" which is equal to twice the propagation time between the nodes. The root contention situation is resolved by utilizing a random decision mechanism incorporated on each arbitration state machine logic unit 11 of each node. For every "decision time period" that elapses, each node will randomly decide at step 92 (with a 50% probability) whether to again transmit the YAMP signal to the other. Almost certainly within a finite number of the cycles, one node will decide to designate the other its parent without that one reciprocating. The one that is designated the parent becomes the root at step 95. Alternatively, predetermined selection criteria values may be assigned to nodes, the larger or smaller determining which dominates in a contention event . The dynamic determination of the "decision time period", while it offers optimum performance is not essential in implementing the present invention. As an alternative an apriory defined "decision time period" may be used as long as it is greater than the worst case link propagation that can be encountered in any bus using this algorithm. The same method used to resolve root contentions will also be used to resolve other contention events to be described further herein.
E-ϋ-i Assignment
As described above, the result of the graph transformation process is the assignment of the root attribute to one and only one node in the graph. The root node will have the ultimate decision in the bus arbitration scheme to be described and therefore can access the bus with maximum priority without the use of special priority time gaps. It is often desirable to be able to assign the root property to a predetermined node either when it is manufactured or dynamically (during run time) to optimize a given system. A given bus may include a node which requires isochronous data transfer. Isochronous data is data that must be transmitted at a given time to be of any value. For example, music from a compact disk needs to be transferred and output in the order in which it is to be heard and with no significant delays, unlike data files which may be transferred piecemeal and not necessarily in order.
Nodes can be classified into three categories with respect to root designations. These designations may be applied during manufacturing by hard-wiring the designation into the device, programming the arbitration state machine logic or by higher level software making the decision then initiating a reboot while preserving that decision. The three designations that a node may be assigned with respect to being designated a root are: nodes that do not want to be root, nodes that may (should) be root and nodes that shall be root. These designations are tested for at steps 81 and 83. A node designated in the first category will begin the graph transformation procedure as soon as it is directed to do so. This will usually be immediately following the completion of the bus initialization procedure. A node from the second category will delay the beginning of the graph transformation procedure for a predetermined amount of time after it is directed to begin the procedure at step 84. By this delay, the node increases its chance of becoming the root. (The YAMP signals are more likely to propagate to it due to the delay.) Despite the added delay, it is still possible that a "may be root" node will not wind up being designated the root . This will depend on the given topology and message propagation delays. The amount of delay can be defined during design to be greater than a reasonable worst case propagation delay through a fairly complex graph.
A node from the third category of root designation possibilities may only recognize the fact that it must be the root after the graph has already been transformed and all nodes have identified themselves. The arbitration state machine logic may make this determination or software running on the host system. When this occurs, the node that has to be root agrees with all other nodes on the bus that it is going to be the one and only root and restarts the graph transformation process by signaling a preemptive bus initialization signal which is described further herein. The node then waits at step 82 to become the root and does not participate in the graph transformation until it has received the YAMP signal on all of its ports, thus forcing it to be designated the root.
Once the root has been determined, the graph can be said to be directed. There is a defined relationship existing between all adjacent nodes on the graph.
£_ --J-£ Resolution
The procedures described above for directing a graph will only work for an acyclic graph. If there are cycles in the arbitrary topology, they must be broken by the procedure beginning at step 80. This is done by intelligently treating certain links as though they are not connected. The existence of a cycle is detected at step 79 when, after a predetermined time-out period has elapsed, a node is still labeled a cycle node rather than a leaf, branch or root. The "cycle detect" timing starts immediately after the end of the bus initialization function. The time-out period need be no longer than the worst case duration of the graph transformation process (adding in delay time for a "may be root" node and a possible root contention event) .
The "cycle detect" time-out event does not have to occur synchronous for all nodes of a graph as all message exchanges are asynchronous events . As such, it is possible for a node which has not yet reached its "cycle detect" time-out event to receive a message indicating that cycle resolution is ongoing. Such a node will terminate its cycle detect time-out interval and begin the appropriate cycle resolution process.
The method of cycle resolution may require different iteration processes depending upon the graph topology. Essentially there are two types of cycles, simple cycles and multiple cycles. Multiple cycles exist where there are nodes that meet the definition of multiple cycles nodes. A multiple cycles node is a cycle node which has at least three connected ports through which it has not received the YAMP signal. A multiple cycles node has at least three adjacent cycle nodes. It will belong to either at least two distinct graph cycles, or to at least one graph cycle and to a branch wherein there is a node having the "shall be root" designation (which therefore has not propagated a YAMP signal) . These ports are labeled by the node as cycle ports. A node may only be considered a multiple cycles node during the cycle resolution process. In contrast, a single cycle node is one which has two and only two connected ports through which it has not yet received the YAMP signal.
Multiple Cycles Resolution
During cycle resolution, if there exist any multiple cycles nodes at step 100 then the Multiple Cycles Resolution Process 100 must be performed. Figure 5(a) illustrates in greater detail the portion 25 of the arbitrary topology shown in Figure 2(b) . In this figure, nodes 40-47 form multiple cycles because there are two nodes, 43 and 47, which are linked to more than two other cycle nodes. Figure 6(f) illustrates the process flow for multiple cycles resolution.
When the cycle detect time-out period terminates, at step 106 all multiple cycles nodes transmit through all their cycle ports a signal indicating "You Are My Cycle Parent" (YAMCP) . If the YAMCP signal generates a reply conveying the message "You Are My Cycle Child" (YAMCC) , the node responds with a "You Are My Cycle Child Acknowledged" (YAMCCA) at step 108 and labels the corresponding port as a cycle parent port.
If a node that has transmitted a YAMCP signal receives on the transmitting port a YAMCP signal before a YAMCC signal, a cycle contention event has occurred such as will be the case between nodes 43 and 47. The cycle contention event is resolved in the same manner as a root contention event wherein for each "decision time period" the contending nodes randomly (50% likelihood) each send or do not send the YAMCP signal until a decision time interval in which one of the two ports involved becomes a cycle parent port while the other one recognizes its status as a cycle child port. When a.single cycle node receives through one or more of its ports the YAMCP signal, it replies with the YAMCC signal and awaits the return of a YAMCCA. After these exchanges, the ports are labeled cycle child ports.
When a single cycle node has labeled one of its two cycle ports as a cycle child port, it conveys the YAMCP signal through its remaining cycle port at step 109. If the YAMCC signal is received as a reply without interference, then the node sends the YAMCCA and labels the port a cycle parent port. If, on the other hand, a contention situation occurs with the node receiving a YAMCP signal after transmitting the same, the cycle contention event is resolved at step 110 in exactly same manner as described above where other contentions between two nodes are resolved.
If a single cycle node winds up with both of its ports labeled as cycle child ports, one of the two links leading to the node must be redirected at step 111. This is a common situation and occurs where a single cycle node is situated logically between two multiple cycles nodes such as may be the case with node 41. The single cycle node selects one of its two connected cycle child ports, based upon an apriory defined criterion, and transmits the YAMCP signal. There will be no contention event, the receiving port with reply with the YAMCC to which the selected port replies YAMCCA and relabels the port that was a cycle child port as a cycle parent port. Ultimately in the cycle resolution process, every single cycle node will have one and only one port labeled as a cycle child port.
If a single cycle node receives the YAMCP signal through a port that was previously labeled as a cycle parent port, it shall respond with the YAMCC and wait for the YAMCCA at which time the port is relabeled as a cycle child port. That single cycle node will then have both cycle ports labeled as cycle child ports. The node will then select the port that has the older cycle child port label and transmit the YAMCP signal, thus reversing the direction of the link to that port. This process will continue until all single cycle nodes have one cycle parent node and one cycle child node. It should be emphasized that the criteria used for port relabeling should follow one of two rules. If the ports of a single cycle node are labeled as cycle child ports within the same time unit any apriory defined criterion is acceptable (always choose port X over port Y, etc.) . If the ports have been labeled at different times the relabeling process should always involve the port which has been labeled a cycle child port the earliest in time. The method of a priory criterion selection may be a logical or hard-wired ordering of ports on a node. For example, each node may elect to make port 1 a higher priority than port 2 and port 2 a higher priority than port 3, etc. The node merely needs a rule for selecting one port in favor of another and to consistently apply that rule.
Once all single cycle nodes each have one cycle parent and one cycle child, at least one multiple cycles node will have at least one of its ports labeled a cycle child port either as a result of a contention with another multiple cycles port or as a result of relabeling message exchange. Depending upon the number of cycles in a graph a multiple cycles node may end up with more than one cycle port labeled as a cycle child port such as node 43 in Figure 5(b) which shows the direction of all cycle links resulting from the above steps. When this occurs, the multiple cycles node will select all but one of its cycle child ports and transmits through it the "link disabled" (LD) message. Upon receiving the "link disabled completed" (LDC) reply signal the ports are labeled by the respective nodes as disabled child ports and removed from the logical bus. A cycle node (single or multiple) upon receiving on one of its ports labeled a cycle parent port the LD signal shall reply with the LDC signal, label the port a disabled parent port and remove it from the logical bus. Thus, even where a physical cable connects two nodes, it will be ignored as a link for communications purposes between the two nodes. In Figure 5(c), the link between nodes 42 and 43 or between 43 and 47 will be disabled, breaking the multiple cycles into a simple cycle. The method of selecting which link(s) to disable may be apriory defined as described above for selecting redirecting propagation. If the direction of the links in the 47-46-45-44-43 chain had been reversed, the breaking of two of the links to 43 would have eliminated any simple cycles as well.
In the case where there is a dangling "shall be root" node which still has its cycle node attribute, the node will respond as described above with respect to cycle parent messages. At this point there will be no multiple cycles remaining with the broken links taken into account . The above breaking of the multiple cycles for the example shown in Figure 5(b) will either yield no remaining cycles or one remaining simple cycle, a cycle in which all cycle nodes are single cycle nodes.
Simple Cvcle Resolution
After the resolution of any existing multiple cycles, there may be remaining simple cycles. These will either be ones that were always simple cycles, or ones that result from the breaking of multiple cycles. The nodes comprising any simple cycles resulting from the breaking of multiple cycles will already be directed with respect to adjacent cycle nodes. Nodes in simple cycles that have always been part of simple cycles will have to be directed with respect to adjacent cycle nodes before the cycle can be broken. For those single cycle nodes which do not yet have their cycle ports labeled at decision box 101, the following procedure should be carried out at step 103 in order to establish the direction with respect to the adjacent cycle nodes. At the beginning of the single cycle resolution process, a cycle node shall wait for a period of time equivalent to the "decision time period" defined previously. Then, based on any apriory arbitrarily defined criteria, a cycle node shall transmit on one of its two cycle ports, the YAMCP signal. If this results in any contention with an adjacent cycle node, the contention will be resolved in exactly the same manner as described above for resolving other contentions. If a node receives the YAMCP signal before initiating its own transmission, it will abort its own transmission and respond with the YAMCC signal. Upon receiving the reply YAMCCA signal, it shall label the port as a cycle child port and then transmit through the remaining port the YAMCP signal. In this way at no time will any cycle node have more than one cycle parent port. If a cycle node winds up with two cycle child ports, it shall select the oldest labeled port, unlabel it and transmit therethrough the YAMCP signal. At the conclusion of this process, all cycle ports are labeled either cycle parent or cycle child thus yielding direction to the links in the loop. It should be noted that there is a special case where multiple cycle elimination will resolve all cycles. Figure 5(d) illustrates this case where a single cycle graph has a dangling "shall-be-root" node, node 5. Though this is a single cycle graph, node 3 believes it is a multiple cycles node and begins the multiple cycle resolution process . At the end of the multiple cycle resolution process the cycle will be eliminated. One node will be left with only one cycle port, hence the test of decision box 102 to end the cycle resolution process with that node becoming a branch node.
Cvcle Elimination
Figure 5(c) shows the nodes 40-47 with the link between nodes 42 and 43 depicted as logically not being present. Also indicated is the remaining directed loop encompassing nodes 43-47.
Once direction has been established for a cycle either as described in the preceding section or resulting from multiple cycles resolution, it is necessary to identify a "root cycle node" in order to break the cycle. Each node in the loop will randomly decide to send either one of two messages through its cycle parent port. The number of possible messages has to be at least two but may be greater. The only condition is that an order relation be defined between all possible messages and this relation be known to all the nodes. The content of these messages is insignificant other than that they be distinguishable and one be designated to have priority over the other. For illustrative purposes, we will describe this as "Cycle Message A" (CMA) and "Cycle Message B" (CMB) . Each node will transmit its randomly selected message to its parent or parents and receive from its child the child's randomly selected message or the message forwarded by its child if the child has become a forwarding node. Every cycle node electing to transmit the CMB signal, but receiving a CMA signal will consider itself eliminated from the race to become the root cycle node and will from thereon merely forward any message received including the CMA message which has triggered the transformation. The time delay between successive messages generated and put out by a node which is not yet a forwarding node is a system parameter defined at design time. Its value shall be greater than the maximum propagation delay through a practical implementable cycle.
A node which does not become a forwarding node will output a predefined number of messages which is also a system parameter defined at design time. This number shall be large enough to guarantee probabilistically, with the desired level of certainty that at the end of this process one and only one node in the cycle remains a non-forwarding node. This remaining node will then be considered the root cycle node . The root cycle node will then send the LD signal through its child port at step 104 in the manner described above, thus logically breaking the link between the root cycle node and its child. This, therefore, breaks the cycle and when all cycles are broken the overall graph is acyclic.
The reason the root cycle node breaks the link with its child node is to accommodate the possibility that there is a a waiting "shall be root" node appended to the root cycle node. It should be noted that once a cycle is a directed cycle, cycle nodes will only have a single cycle child but may have more than one cycle parent (as in the case of the waiting "shall be root" node) . A shall be root dangling node or chain of nodes will have the attribute of being a cycle parent node to the root cycle node, a link to which should not be broken. Once each of the loops is broken, and no cycles remain, the procedure for transforming the graph as described in the earlier sections may proceed until the entire graph is both acyclic and directed.
Unique Physical Addreaa Assignment
Once a directed acyclic graph has been established from the original arbitrary topology, it is then possible to assign unique physical addresses to each node on the graph. This process begins with all leaf nodes requesting the bus by transmitting through their single connected ports the Bus Request (BR) signal. The parent node receiving the signal will wait until it has received the BR signal from all of its child ports and then will propagate the BR signal to its parent. The BR signals will propagate through the graph until the root has received the BR signal from all of its children. Once the root has received a bus request through all of its child ports, it will make a decision for granting the bus through one port and propagating a Bus Denial (BD) signal through its remaining child ports. The method for selecting which bus request to grant may be an apriory decision such as that described above where, for example, ports are selected from left to right or based on port numbering, etc. The Bus Grant (BG) signal will be transmitted from the root to its requesting child. If that requesting child is itself a parent node which has propagated the bus request from one of its children, it will send the bus denial signal through all but one of its child ports in the same predetermined fashion as described above. Eventually one leaf node will receive the bus grant signal to which it will reply with a Bus Grant Acknowledged (BGA) signal which will be propagated back to the root node. The propagation of the BD and BGA signals serve to orient the communication links which may be necessary for the case of half duplex communications channels. All of the denied nodes will then wait for activity by the node which finally receives the BG signal.
The node which is finally granted access to the bus will transmit an address assignment packet. The node will transmit this packet on the bus and it will be received by all other nodes, each of which will count the number of address packets they receive. The transmitted address packet may have any arbitrary information. A node's unique physical address will be based on the number of address packets a node has counted before it transmits an address packet. Thus, no two nodes will acquire the same physical address despite not having address information assigned in advance. The actual composition of the address packet is arbitrary and may be any bit stream efficiently utilizable by the system. After transmitting the physical address assignment packet, a node will transmit a "Child ID Completed" signal (CIC) signal. The parent node receiving this on its child port will then transmit the "Child Identification Completed Acknowledgment" (CICA) signal and label the port as an identified child port . In response to the next BR signal propagation, the parent of the node which has just identified itself will then select its next child to transmit the physical address packet . Once all the child nodes of a parent node have identified themselves, the parent node will request the bus and, when granted the bus, will propagate its physical address assignment packet. This procedure will continue following the predetermined selection criteria until all nodes determine a unique physical address assignment by counting. Figure 7 illustrates the graph of Figure 3(e) in which a left-to-right predefined selection criteria is implemented. The nodes are uniquely assigned addresses where node 33 receives the first and as described, the root node 30 receives the eighth and last address.
When this procedure is completed, each node in the graph will have a unique physical address, which need not have been determined in advance and which may be utilized for system management or other purposes .
HΔdfi Sel -Identifica ion
The process of node self identification essentially follows the same routine as the physical address assignment procedure described above. As each node transmits its physical address assignment packet, that packet may include further information such as identification of the physical device comprising the identification of the local host related to the node, how much power it requires, and, for example, whether it supports a "soft power-on" attribute, etc. In fact, the node self-identification information may serve as the physical address assignment packet because the practice of sending any information at all provides the basis for counting to yield unique physical addresses.
With respect to the node self-identification packet, the particular information concerning the node need only be "listened" to by those nodes affected by the nature of the announcing node. This procedure, as with the above, proceeds until all nodes have transmitted their node self identification information. Topoloqy Mapping
The method of topology mapping follows along the same lines as physical address assignment and node self- identification. This procedure thus has each node, when it is going through the address assignment or node self- identification process, further transmit information concerning all of its ports such as the number of child ports it has and whether or not it has any disabled ports. With respect to disabled ports, it may be desirable to implement a communication protocol between ports that are disabling so that they can identify from whom they are disabled. Thus, when a port identifies a disabled port it will give an identifier indicating its own ID as well as the port ID from which it has become disabled.
By assembling all the topology information about all the ports received during the topology mapping procedure, the bus server, host or any software level application may logically reconstruct the resolved bus topology. This is useful for many purposes including implementing redundancy where if a link unexpectedly goes down, previously disabled links may serve to prevent the loss of communication channels to any nodes. Fair Bus Access Arbitration
Once the topology mapping, node self identification or physical address assignment routines have completed, the bus can be considered up and running. One arbitration scheme implemented in accordance with the present invention is that of fair bus access. When a node desires access to the bus, it sends through its parent port (unless it is the root) a bus request (BR) signal. The parent, upon receiving the BR signal from one child sends a bus denied signal (BD) through all its other child ports. The parent then propagates the BR signal upward through its parent until the signal reaches the root. The root issues a bus grant signal (BG) responsive to the first BR signal it receives and sends the BD signal through all of its other child ports which propagate downward thereby orienting the links. The BG signal propagates downward through the graph until it reaches the requesting node which then sends Bus Acknowledge (BA) signal followed by the packet of information that the node needed to send on the bus. When the packet is completed, all nodes return or enter into an idle state. In the case where the root receives nearly simultaneous requests for the bus, the predetermined selection criteria for the root node will be used for granting to one of the nodes bus access . This may be the same predetermined priority selecting criteria as described above.
A further aspect of fair bus access arbitration is that a parent node has priority over its children. Thus, when a parent node wants the bus, it sends the BD signal through all of its child ports, then propagates the BR signal up toward the root. One potential problem with this mechanism is that if the parent has a large quantity of information to transmit on the bus a child node may have trouble getting adequate bus access. There is therefore introduced a gap system which is widely used and well-known in the art. After a node has utilized the bus, the node must wait for one gap period before it can again request the bus. This gives equal chance of being granted the bus to every node on the bus independent of its topological placement on the bus. In order to guarantee a fair arbitration protocol the length of the gap has to be greater than the worst case signal propagation delay through the bus. The gap value can be predetermined and hard-wired into the node logic but such an approach will result in all but the most extreme case in a sub optimal utilization of the bus. The topology mapping capability together with the measurement of the propagation delay between adjacent nodes performed during the graph transformation phase enables the calculation of an optimal fair gap that will maximize the bus performance for any specific implementation.
Priority Bus Arbitration
In the bus arbitration scheme implemented in accordance with the above fair bus access arbitration, it may be desirable that the root always have bus priority. When this is implemented, the root node may grant the bus to itself at any time. This is done by first sending the BD signal down through all of the nodes in the graph. Priority bus access for the root is very useful for the case where the root node is required to perform isochronous data transfer.
Token Passing Bus Arbitration
As an alternative to the fair and priority bus access arbitrations schemes described above, the present invention may be utilized in implementing a token passing bus arbitration scheme. Metaphorically speaking, token passing bus access refers to the notion that a node may communicate on a bus when it is in possession of a token that is passed between nodes. The token is passed from node to node in a cyclic fashion so that each node receives the bus in a predetermined point in the cycle. Token passing is implemented in the present invention in following the same manner as the physical address assignment routine described above. The predetermined selection mechanisms implemented are used to select the order in which the token will be passed from node to node. This order resembles the order as shown in Figure 7 which dictates the order of unique address assignment. Each node, when it is assigned the token will propagate its information packet on the bus while the remaining nodes listen. The node will then pass the token to the next logical node based on the predetermined sequencing method as described above.
Preem ive Bus Initialisation
An important feature that may be implemented in accordance with the present invention is the notion of preemptive bus initialization. The state machine logic incorporated on each node is capable of triggering a bus initialization (BI) signal to be propagated from the node through all of its ports upon certain conditions. When a node has determined it is necessary to signal a bus initialization condition, it will propagate the BI signal out through all of its ports for a length of time sufficient to guarantee that all adjacent nodes have received it and then released. A node will then go into the initiating procedures which then lead to the graph transformation process in the above described procedures .
There are a number of situations which may make it necessary or desirable to trigger a preemptive bus initialization. First, this may be a node response to an unforeseen error. Additionally, at the host level, it may be determined that a different node should acquire the root attribute, for example, an isochronous data transfer node. This assignment will be preserved throughout the bus initialization routine thereby causing the desired node to wait during the transform procedure until it receives the root designation. Another condition leading to a preemptive bus initialization may be the breakage of a link, in which case it may be necessary to calculate a new acyclic directed graph for the attached nodes . Finally, an important situation in which a preemptive bus initialization should occur is when a device is added to the network, referred to as "hot addition" of peripherals. The port to which a new device is connected will detect the presence of a new node and trigger a bus initialization which will be transparent to the user of the system but which allows the addition and subtraction of peripherals, for example, without having to shut down and repower. A new acyclic directed graph is calculated which includes the presence of the added node. It is possible that upon removing certain nodes, it will not be necessary to trigger a bus initialization, for example, when a leaf node is removed, there is no harm to the network. However, if a branch node is disengaged from an operating bus, it is likely to be necessary to reconfigure the graph.
Although the present invention has been described in terms of preferred embodiments, it will be appreciated that various modifications and alterations might be made by those skilled in the art without departing from the spirit and scope of the invention. The invention should, therefore, be measured in terms of the claims which follow.

Claims

C AIMS
1. In an electronic system comprising a plurality of discrete components, each component having at least a first communications node having at least one external port, said plurality of components being interconnected in an arbitrary topology connected by a plurality of communications links, each of said communications links coupling a pair of nodes through one of the external ports on each of said pair of interconnected nodes, each of said communications nodes being capable of detecting a connection to another node and being capable of propagating and acknowledging signals with adjacent nodes through a communications link, said arbitrarily interconnected plurality of discrete components constituting an arbitrary topology collection of nodes, a method for transforming said arbitrary topology collection of nodes into a logical bus characterized by a directed acyclic graph comprising the steps of:
resolving said arbitrary topology collection of nodes into an acyclic graph by eliminating cycles from the graph; and directing hierarchically the relationships between the nodes on the acyclic graph by propagating directing signals between adjacent nodes on the graph wherein adjacent nodes are nodes directly coupled together by a communications link.
2. The method of claim 1 wherein said resolving step comprises the steps of:
detecting whether any cycles exist, wherein a cycle exists if it is possible to start from a given node in the graph and return to that node after traversing links and nodes without any link being traversed twice;
eliminating any multiple cycles, wherein a multiple cycle exists if there are any nodes which have more than two connected links which are part of a cycle; and
eliminating any single cycles whereby the resulting graph will have an acyclic topology.
3. The method of claim 2 wherein said detecting step comprises determining that some of said plurality of communications nodes have not received directing signals indicating a hierarchical relationship with an adjacent node after a predetermined period of time has elapsed.
4. The method of claim 2 wherein in a multiple cycle at least two nodes are characterized as multiple cycles nodes, being those nodes which have more than two connected links which are part of a cycle, the remaining nodes of the cycles being single cycle nodes, the step of eliminating multiple cycles comprising for each multiple cycle the steps of:
directing hierarchically the nodes of the multiple cycle wherein all multiple cycles nodes in the cycle propagate a "You Are My Cycle Parent" (YAMCP) signal to which responding nodes signal back "You Are My Cycle Child" (YAMCC) , said single cycle nodes upon having identified one adjacent node as a cycle parent node, identifying its other adjacent node as a cycle child node, said directing step continuing until all single cycle nodes each have identified adjacent nodes as one cycle parent node and one cycle child node at which time at least one of said multiple cycles nodes will have an adjacent node identified a cycle child node; and logically disabling a selected link so as to eliminate the multiple cycle from the graph.
5. The method of claim 4 wherein said logically disabling step comprises the step of propagating a "link disabled" (LD) signal from a multiple cycles node which has an adjacent node identified as a cycle child node to said adjacent cycle child node, said adjacent cycle child node responding with a "link disabled completed" (LDC) signal at which time the link between the two involved nodes is treated as not existing in the logical bus to be resolved.
6. The method of claim 4 wherein said directing step further comprises a step of resolving conflicts between adjacent nodes that attempt to identify each other as parent nodes by having the conflicting nodes periodically randomly identify or not identify the other as a parent node until one achieves parent node status with respect to the other.
7. The method of claim 2 wherein said step of eliminating single cycles comprises for each single cycle the steps of:
directing the links of the cycle if they have not been previously directed as a result of resolving a multiple cycle so that each node in the single cycle has identified one of its two adjacent nodes as a cycle parent node and the other adjacent node as a cycle child node;
selecting one communications link in the cycle to be logically disabled; and
logically disabling the selected communications link.
8. The method of claim 7 wherein said selecting one link to be logically disabled step comprises the steps of:
determining one node of the cycle to be the root cycle node;
propagating from the root cycle node to one of its adjacent nodes the LD signal; and
logically removing the link between the root cycle node and said one of its adjacent nodes from the logical bus to be resolved.
9. The method of claim 8 wherein said step of determining one node to be the root cycle node comprises the steps of:
iterating the following steps until only one node remains in contention to become the root cycle node:
each node propagating to its adjacent node identified as its cycle parent node either a first priority signal or a second priority signal, being determined randomly; and
each of said nodes which elected to propagate said second priority signal eliminating itself from contention to become the root cycle node and thereafter forwarding any of said first or second priority signals received through from its cycle child node to its parent node.
10. The method of claim 1 wherein said step of directing hierarchically the relationships between nodes on the acyclic graph comprise the steps of: establishing parent-child relationships between all adjacent nodes; and
designating one node on the graph to be the root node where all other nodes are either leaf nodes or branch nodes, said leaf nodes being characterized as nodes having only one adjacent node on the graph.
11. The method of claim 10 wherein said step of establishing parent-child relationships between adjacent nodes comprises the steps of:
propagating from each leaf node a "You Are My Parent" (YAMP) signal to its only adjacent node, wherein in response to said YAMP signal, said adjacent node propagates to said adjacent leaf node "You Are My Child" (YAMC) signal thereby establishing a parent-child relationship between each leaf node and its adjacent node; and
each branch node, upon receiving and acknowledging the YAMP signal from all but one of its adjacent nodes, propagating the YAMP signal to its remaining adjacent node, said step continuing until all branch nodes have propagated said YAMP signal to an adjacent node with one branch node. the last to receive the YAMP signal from all adjacent nodes achieving the status of being the root node for the graph.
12. The method of claim 11 further comprising the step of resolving root contentions between adjacent nodes attempting to identify each other as parent nodes by having the conflicting nodes periodically randomly identify or not identify the other as a parent node until one achieves parent node status with respect to the other, that node becoming the root node.
13. The method of claim 11 further comprising the step of designating one node as a "shall-be-root" node, said node waiting for a predetermined period of time before propagating said YAMP signal to an adjacent node, thereby increasing the chances that it results in being the root node.
14. The method of claim 11 further comprising the step of designating one node as a "must-be-root" node, said node not propagating said YAMP signal to any adjacent node, thereby ensuring said node becomes the root node.
15. In a computer system comprising a plurality of components interconnected by a plurality of communications links, each of said plurality of components having a communications node with at least a first port to which said communications links may be coupled, said interconnected plurality of components forming an arbitrary topology of nodes, the method of transforming said arbitrary topology into an acyclic graph comprising the step of:
detecting whether any cycles exist in the arbitrary topology, wherein a cycle exists if it is possible to start from a given node in the graph and return to that node after traversing links and nodes without any link being traversed twice;
eliminating any multiple cycles, wherein a multiple cycle exists if there are any nodes which have more than two connected links which are part of a cycle; and
eliminating any single cycles whereby the resulting graph will have an acyclic topology.
16. The method of claim 15 wherein said detecting step comprises determining that some of said plurality of communications nodes have not received directing signals indicating a hierarchical relationship with an adjacent node after a predetermined period of time has elapsed.
17. The method of claim 15 wherein in a multiple cycle at least two nodes are characterized as multiple cycles nodes, being those nodes which have more than two connected links which are part of a cycle, the remaining nodes of the cycles being single cycle nodes, the step of eliminating multiple cycles comprising for each multiple cycle the steps of:
directing hierarchically the nodes of the multiple cycle wherein all multiple cycles nodes in the cycle propagate a "You Are My Cycle Parent" (YAMCP) signal to which responding nodes signal back "You Are My Cycle Child" (YAMCC) , said single cycle nodes upon having identified one adjacent node as a cycle parent node, identifying its other adjacent node as a cycle child node, said directing step continuing until all single cycle nodes each have identified adjacent nodes as one cycle parent node and one cycle child node at which time at least one of said multiple cycles nodes will have an adjacent node identified a cycle child node; and
logically disabling a selected link so as to eliminate the multiple cycle from the graph.
18. The method of claim 17 wherein said logically disabling step comprises the step of propagating a "link disabled" (LD) signal from a multiple cycles node which has an adjacent node identified as a cycle child node to said adjacent cycle child node, said adjacent cycle child node responding with a "link disabled completed" (LDC) signal at which time the link between the two involved nodes is treated as not existing in the logical bus to be resolved.
19. The method of claim 17 wherein said directing step further comprises a step of resolving conflicts between adjacent nodes that attempt to identify each other as parent nodes by having the conflicting nodes periodically randomly identify or not identify the other as a parent node until one achieves parent node status with respect to the other.
20. The method of claim 15 wherein said step of eliminating single cycles comprises for each single cycle the steps of:
directing the links of the cycle if they have not been previously directed as a result of resolving a multiple cycle so that each node in the single cycle has identified one of its two adjacent nodes as a cycle parent node and the other adjacent node as a cycle child node;
selecting one communications link in the cycle to be logically disabled; and
logically disabling the selected communications link.
21. The method of claim 20 wherein said selecting one link to be logically disabled step comprises the steps of:
determining one node of the cycle to be the root cycle node;
propagating from the root cycle node to one of its adjacent nodes the LD signal; and logically removing the link between the root cycle node and said one of its adjacent nodes from the logical bus to be resolved.
22. The method of claim 21 wherein said step of determining one node to be the root cycle node comprises the steps of:
iterating the following steps until only one node remains in contention to become the root cycle node:
each node propagating to its adjacent node identified as its cycle parent node either a first priority signal or a second priority signal, being determined randomly; and
each of said nodes which elected to propagate said second priority signal eliminating itself from contention to become the root cycle node and thereafter forwarding any of said first or second priority signals received through from its cycle child node to its parent node.
PCT/US1993/012317 1992-12-21 1993-12-16 Method and apparatus for transforming an arbitrary topology collection of nodes into an acyclic directed graph WO1994015304A1 (en)

Priority Applications (4)

Application Number Priority Date Filing Date Title
DE69331705T DE69331705T2 (en) 1992-12-21 1993-12-16 METHOD AND DEVICE FOR CONVERTING ANY TOPOLOGY FROM A NODE COLLECTION IN AN ACYCLIC DIRECTED GRAPH
JP6515318A JPH08507623A (en) 1992-12-21 1993-12-16 Method and apparatus for converting an arbitrary topology collection of nodes into an acyclic directed graph
AU59541/94A AU5954194A (en) 1992-12-21 1993-12-16 Method and apparatus for transforming an arbitrary topology collection of nodes into an acyclic directed graph
EP94905424A EP0674790B1 (en) 1992-12-21 1993-12-16 Method and apparatus for transforming an arbitrary topology collection of nodes into an acyclic directed graph

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US99412892A 1992-12-21 1992-12-21
US07/994,128 1992-12-21

Publications (1)

Publication Number Publication Date
WO1994015304A1 true WO1994015304A1 (en) 1994-07-07

Family

ID=25540312

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US1993/012317 WO1994015304A1 (en) 1992-12-21 1993-12-16 Method and apparatus for transforming an arbitrary topology collection of nodes into an acyclic directed graph

Country Status (7)

Country Link
US (1) US5784557A (en)
EP (2) EP1146431A3 (en)
JP (1) JPH08507623A (en)
AU (1) AU5954194A (en)
DE (1) DE69331705T2 (en)
HK (1) HK1041330A1 (en)
WO (1) WO1994015304A1 (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1079580A2 (en) * 1999-08-23 2001-02-28 Sanyo Electric Co., Ltd. Method and apparatus for avoiding loop connection
US7418005B2 (en) 2001-07-24 2008-08-26 Matsushita Electric Industrial Co., Ltd. Bus optimizing method and communication node

Families Citing this family (97)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5634004A (en) * 1994-05-16 1997-05-27 Network Programs, Inc. Directly programmable distribution element
US7388092B2 (en) * 1996-05-03 2008-06-17 Applera Corporation Oligonucleotides and analogs labeled with energy transfer dyes
EP0825506B1 (en) 1996-08-20 2013-03-06 Invensys Systems, Inc. Methods and apparatus for remote process control
US5978795A (en) * 1997-01-14 1999-11-02 Microsoft Corporation Temporally ordered binary search method and system
US6160795A (en) * 1997-03-21 2000-12-12 Siemens Aktiengesellschaft Network communication
US6633577B1 (en) 1997-03-26 2003-10-14 Nec Corporation Handshaking circuit for resolving contention on a transmission medium regardless of its length
US6038600A (en) * 1997-07-16 2000-03-14 Hewlett-Packard Company Method and system for automatic detection of bridged and repeated network device connections
US6145018A (en) * 1997-11-24 2000-11-07 Intel Corporation Method for hindering some types of nodes from becoming a bus arbitration controller
US6041348A (en) * 1997-12-01 2000-03-21 Lsi Logic Corporation N-port algorithm for disabling a node within a network during reset
JP3171241B2 (en) * 1998-03-06 2001-05-28 日本電気株式会社 Communication method
US6324181B1 (en) * 1998-04-16 2001-11-27 3Com Corporation Fibre channel switched arbitrated loop
FR2779301B1 (en) * 1998-05-26 2000-07-21 Thomson Multimedia Sa METHOD FOR IDENTIFYING DEVICES IN A COMMUNICATION NETWORK AND APPARATUS FOR IMPLEMENTING
US6108698A (en) * 1998-07-29 2000-08-22 Xerox Corporation Node-link data defining a graph and a tree within the graph
US6597658B1 (en) 1998-12-28 2003-07-22 At&T Corp. Hierarchical telecommunications network with fault recovery
US7062456B1 (en) 1999-02-09 2006-06-13 The Chase Manhattan Bank System and method for back office processing of banking transactions using electronic files
US6334205B1 (en) 1999-02-22 2001-12-25 International Business Machines Corporation Wavefront technology mapping
US6385668B1 (en) * 1999-04-08 2002-05-07 Lucent Technologies Inc. Method and apparatus for compound hardware configuration control
WO2000070531A2 (en) 1999-05-17 2000-11-23 The Foxboro Company Methods and apparatus for control configuration
US7089530B1 (en) * 1999-05-17 2006-08-08 Invensys Systems, Inc. Process control configuration system with connection validation and configuration
US6600727B1 (en) 1999-05-27 2003-07-29 Cisco Technology, Inc. Distributed network repeater system
US6788980B1 (en) 1999-06-11 2004-09-07 Invensys Systems, Inc. Methods and apparatus for control using control devices that provide a virtual machine environment and that communicate via an IP network
US6496485B1 (en) * 1999-06-14 2002-12-17 Hewlett-Packard Company Method of loop breaking tree indentification in a system bus hierarcy
JP4444397B2 (en) 1999-06-25 2010-03-31 富士通マイクロエレクトロニクス株式会社 How to build a topology
US6628607B1 (en) * 1999-07-09 2003-09-30 Apple Computer, Inc. Method and apparatus for loop breaking on a serial bus
US6574633B1 (en) * 1999-11-01 2003-06-03 Honeywell International Inc. Method for dynamically grouping limited range physical entities in a topological space
JP2002077182A (en) * 2000-08-29 2002-03-15 Sony Corp Network error display device and error detection display method
US7328211B2 (en) 2000-09-21 2008-02-05 Jpmorgan Chase Bank, N.A. System and methods for improved linguistic pattern matching
US7539154B1 (en) * 2000-10-17 2009-05-26 Cisco Technology, Inc. Method and apparatus to detect and break loop configuration
US7328232B1 (en) 2000-10-18 2008-02-05 Beptech Inc. Distributed multiprocessing system
JP2002152265A (en) * 2000-11-15 2002-05-24 Tadashi Yamamoto Connection path selection method for communication channel network, and device and recording medium for recording connection path selection program
ATE347765T1 (en) * 2001-03-09 2006-12-15 Koninkl Philips Electronics Nv SYSTEM OF DEVICES THAT COMMUNICATE USING A BUS STRUCTURE
JP2002314543A (en) * 2001-04-13 2002-10-25 Matsushita Electric Ind Co Ltd Topology correction method and communication node
US20030035379A1 (en) * 2001-07-14 2003-02-20 Zimmel Sheri L. Apparatus and method for optimizing telecommunication network design using weighted span classification
US20030051007A1 (en) * 2001-07-14 2003-03-13 Zimmel Sheri L. Apparatus and method for optimizing telecommunication network design using weighted span classification for low degree of separation demands
US20030023706A1 (en) * 2001-07-14 2003-01-30 Zimmel Sheri L. Apparatus and method for optimizing telecommunications network design using weighted span classification and rerouting rings that fail to pass a cost therehold
US20030046378A1 (en) * 2001-07-14 2003-03-06 Zimmel Sheri L. Apparatus and method for existing network configuration
US20030055918A1 (en) * 2001-07-14 2003-03-20 Zimmel Sheri L. Apparatus and method for optimizing telecommunication network design using weighted span classification for high degree of separation demands
US7239614B2 (en) * 2001-07-16 2007-07-03 International Business Machines Corporation Methods and apparatus for the propagation of multicast transmissions in a communications network
US7103054B2 (en) * 2001-07-16 2006-09-05 International Business Machines Corporation Methods and arrangements for building a subsource address multicast distribution tree using point to point routing records
WO2003034221A1 (en) * 2001-10-17 2003-04-24 Beptech Inc. Method of communicating across an operating system
US6950870B2 (en) 2002-02-06 2005-09-27 Harris Corporation Method and apparatus for loop detection and dissolution in a communication network
US7987246B2 (en) * 2002-05-23 2011-07-26 Jpmorgan Chase Bank Method and system for client browser update
US8463947B2 (en) 2002-08-28 2013-06-11 Tellabs Operations, Inc. Method of finding rings for optimal routing of digital information
US7346709B2 (en) * 2002-08-28 2008-03-18 Tellabs Operations, Inc. Methods for assigning rings in a network
US7340650B2 (en) 2002-10-30 2008-03-04 Jp Morgan Chase & Co. Method to measure stored procedure execution statistics
US20040103199A1 (en) * 2002-11-22 2004-05-27 Anthony Chao Method and system for client browser update from a lite cache
US7149752B2 (en) * 2002-12-03 2006-12-12 Jp Morgan Chase Bank Method for simplifying databinding in application programs
US7085759B2 (en) 2002-12-06 2006-08-01 Jpmorgan Chase Bank System and method for communicating data to a process
US7292585B1 (en) * 2002-12-20 2007-11-06 Symantec Operating Corporation System and method for storing and utilizing routing information in a computer network
US8032439B2 (en) 2003-01-07 2011-10-04 Jpmorgan Chase Bank, N.A. System and method for process scheduling
US7401156B2 (en) 2003-02-03 2008-07-15 Jp Morgan Chase Bank Method using control interface to suspend software network environment running on network devices for loading and executing another software network environment
US20040257993A1 (en) * 2003-03-26 2004-12-23 Olav Lysne Method and device for network reconfiguration
US7379998B2 (en) * 2003-03-31 2008-05-27 Jp Morgan Chase Bank System and method for multi-platform queue queries
US20040230602A1 (en) * 2003-05-14 2004-11-18 Andrew Doddington System and method for decoupling data presentation layer and data gathering and storage layer in a distributed data processing system
US7366722B2 (en) * 2003-05-15 2008-04-29 Jp Morgan Chase Bank System and method for specifying application services and distributing them across multiple processors using XML
US8095659B2 (en) 2003-05-16 2012-01-10 Jp Morgan Chase Bank Service interface
US7069278B2 (en) 2003-08-08 2006-06-27 Jpmorgan Chase Bank, N.A. System for archive integrity management and related methods
US20050065964A1 (en) * 2003-09-19 2005-03-24 Ziemann David M. Update of a tree-based database
US7421696B2 (en) * 2003-12-22 2008-09-02 Jp Morgan Chase Bank Methods and systems for managing successful completion of a network of processes
US20050144174A1 (en) * 2003-12-31 2005-06-30 Leonid Pesenson Framework for providing remote processing of a graphical user interface
US8037102B2 (en) 2004-02-09 2011-10-11 Robert T. and Virginia T. Jenkins Manipulating sets of hierarchical data
US20050222990A1 (en) * 2004-04-06 2005-10-06 Milne Kenneth T Methods and systems for using script files to obtain, format and disseminate database information
WO2005102016A2 (en) * 2004-04-26 2005-11-03 Jp Morgan Chase Bank System and method for routing messages
US7614037B2 (en) * 2004-05-21 2009-11-03 Microsoft Corporation Method and system for graph analysis and synchronization
US20060031444A1 (en) * 2004-05-28 2006-02-09 Drew Julie W Method for assigning network resources to applications for optimizing performance goals
US7392471B1 (en) 2004-07-28 2008-06-24 Jp Morgan Chase Bank System and method for comparing extensible markup language (XML) documents
US7764629B2 (en) * 2004-08-11 2010-07-27 Cray Inc. Identifying connected components of a graph in parallel
US7366974B2 (en) * 2004-09-03 2008-04-29 Jp Morgan Chase Bank System and method for managing template attributes
US20060059210A1 (en) * 2004-09-16 2006-03-16 Macdonald Glynne Generic database structure and related systems and methods for storing data independent of data type
US20090132466A1 (en) * 2004-10-13 2009-05-21 Jp Morgan Chase Bank System and method for archiving data
US7801923B2 (en) 2004-10-29 2010-09-21 Robert T. and Virginia T. Jenkins as Trustees of the Jenkins Family Trust Method and/or system for tagging trees
US7627591B2 (en) 2004-10-29 2009-12-01 Skyler Technology, Inc. Method and/or system for manipulating tree expressions
US7630995B2 (en) 2004-11-30 2009-12-08 Skyler Technology, Inc. Method and/or system for transmitting and/or receiving data
US7636727B2 (en) 2004-12-06 2009-12-22 Skyler Technology, Inc. Enumeration of trees from finite number of nodes
US8316059B1 (en) 2004-12-30 2012-11-20 Robert T. and Virginia T. Jenkins Enumeration of rooted partial subtrees
US8615530B1 (en) 2005-01-31 2013-12-24 Robert T. and Virginia T. Jenkins as Trustees for the Jenkins Family Trust Method and/or system for tree transformation
US7681177B2 (en) 2005-02-28 2010-03-16 Skyler Technology, Inc. Method and/or system for transforming between trees and strings
US7899821B1 (en) 2005-04-29 2011-03-01 Karl Schiffmann Manipulation and/or analysis of hierarchical data
US8065606B1 (en) 2005-09-16 2011-11-22 Jpmorgan Chase Bank, N.A. System and method for automating document generation
US7499933B1 (en) 2005-11-12 2009-03-03 Jpmorgan Chase Bank, N.A. System and method for managing enterprise application configuration
US7610172B2 (en) * 2006-06-16 2009-10-27 Jpmorgan Chase Bank, N.A. Method and system for monitoring non-occurring events
US8104076B1 (en) 2006-11-13 2012-01-24 Jpmorgan Chase Bank, N.A. Application access control system
CN104407518B (en) 2008-06-20 2017-05-31 因文西斯系统公司 The system and method interacted to the reality and Simulation Facility for process control
US8127060B2 (en) 2009-05-29 2012-02-28 Invensys Systems, Inc Methods and apparatus for control configuration with control objects that are fieldbus protocol-aware
US8463964B2 (en) 2009-05-29 2013-06-11 Invensys Systems, Inc. Methods and apparatus for control configuration with enhanced change-tracking
US8407340B2 (en) * 2010-04-09 2013-03-26 Microsoft Corporation Page load performance analysis
US8893210B2 (en) * 2010-08-20 2014-11-18 Sony Corporation Server load balancing for interactive television
US8918801B2 (en) 2010-08-30 2014-12-23 Sony Corporation Transmission apparatus, transmission method, reception apparatus, reception method, program, and broadcasting system
WO2012047201A1 (en) * 2010-10-05 2012-04-12 Hewlett-Packard Development Company, L.P. Acyclic graph navigator
US9038177B1 (en) 2010-11-30 2015-05-19 Jpmorgan Chase Bank, N.A. Method and system for implementing multi-level data fusion
US9277422B2 (en) * 2011-02-15 2016-03-01 Tata Consultancy Services Limited Dynamic self configuration engine for cognitive networks and networked devices
US9292588B1 (en) 2011-07-20 2016-03-22 Jpmorgan Chase Bank, N.A. Safe storing data for disaster recovery
US9118539B2 (en) * 2012-07-30 2015-08-25 Cisco Technology, Inc. Managing grey zones of unreachable nodes in computer networks
US9300569B2 (en) * 2012-07-31 2016-03-29 Cisco Technology, Inc. Compressing data packet routing information using bloom filters
US10540373B1 (en) 2013-03-04 2020-01-21 Jpmorgan Chase Bank, N.A. Clause library manager
KR101974069B1 (en) * 2018-04-30 2019-04-30 한국과학기술원 System and method for visualizing binary structure
CN111125849A (en) * 2019-11-30 2020-05-08 浙江华云信息科技有限公司 Ring-breaking processing method for tree-like structure graph model

Family Cites Families (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4344134A (en) * 1980-06-30 1982-08-10 Burroughs Corporation Partitionable parallel processor
US5355371A (en) * 1982-06-18 1994-10-11 International Business Machines Corp. Multicast communication tree creation and control method and apparatus
US4698752A (en) * 1982-11-15 1987-10-06 American Telephone And Telegraph Company At&T Bell Laboratories Data base locking
IT1159351B (en) * 1983-02-03 1987-02-25 Cselt Centro Studi Lab Telecom DISTRIBUTED STRUCTURE REFEREE CIRCUIT FOR BUS REQUESTS FOR A MULTIPROCESSOR SYSTEM
US4706080A (en) * 1985-08-26 1987-11-10 Bell Communications Research, Inc. Interconnection of broadcast networks
US4782444A (en) * 1985-12-17 1988-11-01 International Business Machine Corporation Compilation using two-colored pebbling register allocation method such that spill code amount is invariant with basic block's textual ordering
US4845744A (en) * 1986-10-16 1989-07-04 American Telephone And Telegraph Company, At&T Bell Laboratories Method of overlaying virtual tree networks onto a message passing parallel processing network
US4740954A (en) * 1986-12-31 1988-04-26 Bell Communications Research, Inc. Multicast routing algorithm
US4881166A (en) * 1987-07-24 1989-11-14 Amoco Corporation Method for consistent multidatabase transaction processing
DE3838945A1 (en) * 1987-11-18 1989-06-08 Hitachi Ltd NETWORK SYSTEM WITH LOCAL NETWORKS AND WITH A HIERARCHICAL CHOICE OF PATH
US4811337A (en) * 1988-01-15 1989-03-07 Vitalink Communications Corporation Distributed load sharing
US5088032A (en) * 1988-01-29 1992-02-11 Cisco Systems, Inc. Method and apparatus for routing communications among computer networks
US5079767A (en) * 1988-09-27 1992-01-07 Digital Equipment Corporation Method of multicast message distribution
US5138615A (en) * 1989-06-22 1992-08-11 Digital Equipment Corporation Reconfiguration system and method for high-speed mesh connected local area network
US5150360A (en) * 1990-03-07 1992-09-22 Digital Equipment Corporation Utilization of redundant links in bridged networks
US5280619A (en) * 1990-05-17 1994-01-18 Texas Instruments Incorporated System for accessing shared data using a serialization graph constructed from a history file showing completed locking dependencies between transactions
US5309437A (en) * 1990-06-29 1994-05-03 Digital Equipment Corporation Bridge-like internet protocol router
CA2093355A1 (en) * 1990-10-03 1992-04-04 David C. Douglas Parallel computer system
GB9112898D0 (en) * 1991-06-14 1991-07-31 Digital Equipment Int Communication networks

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
BOKHARI AND RAZA: "Augmenting computer networks", PROCEEDINGS OF THE 1984 INTERNATIONAL CONFERENCE ON PARALLEL PROCESSING, 24 August 1984 (1984-08-24), OHIO, US, pages 338 - 345, XP001613382 *
CIDON AND GOPAL: "Dynamic tree detection in computer networks", PROCEEDINGS OF IEEE INFOCOM, 2 April 1987 (1987-04-02), SAN FRANCISCO, US, pages 181 - 187 *
COMPUTER SCIENCE CONFERENCE, 20 February 1975 (1975-02-20), WASHINGTON DC, US, pages 24 *
DATABASE INSPEC INSTITUTE OF ELECTRICAL ENGINEERS, STEVENAGE, GB; BOWIE: "Distributed operating systems" *
WU ET AL: "Prototype of Star architecture - a status report", AFIPS CONFERENCE PROCEEDINGS, 18 July 1985 (1985-07-18), CHICAGO, US, pages 191 - 201, XP001613371 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1079580A2 (en) * 1999-08-23 2001-02-28 Sanyo Electric Co., Ltd. Method and apparatus for avoiding loop connection
EP1079580A3 (en) * 1999-08-23 2003-08-06 Sanyo Electric Co., Ltd. Method and apparatus for avoiding loop connection
US7418005B2 (en) 2001-07-24 2008-08-26 Matsushita Electric Industrial Co., Ltd. Bus optimizing method and communication node

Also Published As

Publication number Publication date
EP0674790A1 (en) 1995-10-04
EP1146431A3 (en) 2001-12-19
DE69331705T2 (en) 2002-12-19
HK1041330A1 (en) 2002-07-05
EP1146431A2 (en) 2001-10-17
JPH08507623A (en) 1996-08-13
US5784557A (en) 1998-07-21
EP0674790B1 (en) 2002-03-13
AU5954194A (en) 1994-07-19
DE69331705D1 (en) 2002-04-18

Similar Documents

Publication Publication Date Title
EP0674790B1 (en) Method and apparatus for transforming an arbitrary topology collection of nodes into an acyclic directed graph
EP0674788B1 (en) Method and apparatus for arbitrating on an acyclic directed graph
US5394556A (en) Method and apparatus for unique address assignment, node self-identification and topology mapping for a directed acyclic graph
WO1994015305A1 (en) Method and apparatus for transforming an arbitrary acyclic topology collection of nodes into an acyclic directed graph

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AT AU BB BG BR BY CA CH CZ DE DK ES FI GB HU JP KP KR KZ LK LU LV MG MN MW NL NO NZ PL PT RO RU SD SE SK UA UZ VN

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): AT BE CH DE DK ES FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN ML MR NE SN TD TG

DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
121 Ep: the epo has been informed by wipo that ep was designated in this application
WWE Wipo information: entry into national phase

Ref document number: 1994905424

Country of ref document: EP

WWP Wipo information: published in national office

Ref document number: 1994905424

Country of ref document: EP

REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

NENP Non-entry into the national phase

Ref country code: CA

WWG Wipo information: grant in national office

Ref document number: 1994905424

Country of ref document: EP