CN102014111B - Data transmission method, message engine, communication node and network system - Google Patents
Data transmission method, message engine, communication node and network system Download PDFInfo
- Publication number
- CN102014111B CN102014111B CN 200910195309 CN200910195309A CN102014111B CN 102014111 B CN102014111 B CN 102014111B CN 200910195309 CN200910195309 CN 200910195309 CN 200910195309 A CN200910195309 A CN 200910195309A CN 102014111 B CN102014111 B CN 102014111B
- Authority
- CN
- China
- Prior art keywords
- request
- write request
- read
- message
- response
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
Images
Abstract
The invention discloses a data transmission method, a message engine, a communication node and a network system. The data transmission method comprises the following steps of: splitting a message corresponding to one data transmission item into a plurality of reading requests and writing requests, wherein every i (i is more than 1) reading requests correspond to one writing request; transmitting the reading request according to the request sequence of the message; transmitting the writing request after returning the i reading requests corresponding to the writing request; and transmitting an answer word of the message after returning the responses of all writing requests in the message. What needs to be stressed is that the returns of the reading response and the writing end are in unordered mode. The data transmission method, the message engine, the communication node and the network system can improve the throughput rate of message processing and the utilization rate of a chain path.
Description
Technical field
The present invention relates to the interconnected communication field of high-performance calculation, particularly data transmission method, message engine, communication node and network system.
Background technology
At present, the normal mode of transmitting based on message that adopts of the transfer of data between the communication node of computer network, the carrying out that comes the driving data transmission by the mode of message transmission.
Increasing along with the data traffic of the required processing of computer network, transfer of data frequency and capacity between communication node are also increasing.How to realize that the parallel of message is one of key issue of transmission of messages always; existing transmission of messages utilization variety of way realize the parallel of message; but existing various mechanism often following situation can occur: initiation order message formerly receives the reason such as insufficient space and can't finish owing to, destination node in poor shape such as transmission network; will occupied bandwidth; cause subsequent message to get clogged, this situation will affect the efficient of Message Processing.And for the applied environment of high bandwidth, the link utilization of this message process mode is also lower.
Summary of the invention
The problem that the present invention solves be prior art based on the transfer of data of message transfer mode, the problem that the efficient of its Message Processing and the utilance of link are still waiting to improve.
For addressing the above problem, the invention provides a kind of data transmission method, be applied to high bandwidth, it comprises:
The message of a corresponding transfer of data is split into a plurality of read requests and a plurality of write request, the corresponding write request of the individual read request of every i (i＞1);
Request by described message sequentially sends read request;
After the response of i read request corresponding to write request is all returned, send this write request;
After the response of all write requests in described message is returned, send the answer word of described message.
Correspondingly, the present invention also provides a kind of message engine, and for the treatment of message transmission between each communication node in the network system, described message engine comprises: split cells, read request control unit, write request control unit, answer the word transmitting element, wherein,
Described split cells splits into a plurality of read requests and a plurality of write request with the message of a corresponding transfer of data, the corresponding write request of the individual read request of every i (i＞1), and information that will split write request is sent to answer word transmitting element;
Described read request control unit sends read request by the request of described message order to the source node of this transfer of data, obtain source node corresponding read response after, will describedly read to respond to be sent to the write request control unit;
Described write request control unit after the response that obtains i read request corresponding to write request, sends this write request to the destination node of this transfer of data, after the corresponding write response that obtains destination node, described write response is sent to answer word transmitting element;
Answer the word transmitting element, according to the information of write request, after the response of all write requests that obtain destination node, answer word to the source node transmission of this transfer of data.
Correspondingly, the present invention also provides a kind of communication node, comprising:
Split cells splits into a plurality of read requests and a plurality of write request, the corresponding write request of the individual read request of every i (i＞1) with the message of a corresponding transfer of data;
The read request control unit is sequentially carried out read request by the request of described message;
Network communication unit is sent to network with the information of write request after splitting and the response of reading of described read request, and from network, receive to should item number defeated answer word reportedly.
Correspondingly, the present invention also provides a kind of communication node, comprising:
The network receiving element receives the information of the write request of a corresponding transfer of data, and to should item number reportedly defeated read request read response, the corresponding write request of the individual read request of every i (i＞1);
The write request control unit after the response of i read request corresponding to write request of acquisition, is carried out this write request;
Answer the word transmitting element, according to the information of write request, produce and answer word should item number reportedly after the response of all defeated write requests obtaining;
The network transmitting element will be answered word and will be sent to network.
Correspondingly, the present invention also provides a kind of network system that comprises above-mentioned message engine.
Correspondingly, the present invention also provides a kind of network system that comprises above-mentioned communication node.
Compared with prior art, data transmission method of the present invention, message engine, communication node and network system have the following advantages: after message is sequentially initiated, response according to data comes driving request, write request need to receive read accordingly the response after send, the answer word that message is finished need to send after all write responses of receiving described message.The processing procedure of message has formed the mode of data flow driven, thereby can realize the parallel processing of a plurality of message, has eliminated head and has blocked.For the situation of bandwidth applications, described message based data transmission method can significantly improve the utilance of throughput and the link of Message Processing.
Description of drawings
Fig. 1 is a kind of execution mode flow chart of data transmission method of the present invention;
Fig. 2 is a kind of execution mode structural representation of message engine of the present invention;
Fig. 3 is the example schematic that data transmission method of the present invention and message engine are applied to single message;
Fig. 4 is that data transmission method of the present invention and message engine are applied to the concurrent example schematic of multiple messages;
Fig. 5 is a kind of example structure schematic diagram of communication node of the present invention;
Fig. 6 is the another kind of example structure schematic diagram of communication node of the present invention.
Embodiment
Data transfer mode by analyzing prior art as can be known, two internodal data transmission procedures can roughly be divided into two stages in the computer network: the process of reading out data from the main memory of source node, and in the main memory of destination node the process of data writing.Correspondingly, message corresponding to this transfer of data also comprises the read request of corresponding reading out data process, and the write request of corresponding data writing process.Guarantee carrying out smoothly of data writing, its prerequisite is that data can successfully read, so finishing also of write request must depend on finishing of corresponding read request.Correspondingly, finishing of message corresponding to transfer of data also depends on finishing of write request.Therefore be to point out finishing of request by the mode of response based on prior art, if drive its corresponding write request with the response of read request, drive finishing of message with the response of write request, should be able to guarantee the correct processing of message.
Based on this, the invention provides a kind of data transmission method that is applied to high bandwidth, adopted the message processing facility of above-mentioned data flow driven mode.
With reference to shown in Figure 1, a kind of execution mode that the present invention is applied to the data transmission method of high bandwidth comprises:
Step s1 splits into a plurality of read requests and a plurality of write request, the corresponding write request of the individual read request of every i (i＞1) with the message of a corresponding transfer of data;
Step s2 sequentially sends read request by the request of described message;
Step s3, i read request corresponding to write request read the response all return after, send this write request;
Step s4 after the write response of all write requests in described message is returned, sends the answer word of described message.
In the above-mentioned execution mode, after message is initiated, at first message is split by two kinds of request types of read and write.
Split write request boundary's situation is carried out according to message-length and message addresses.Splitting read request then can carry out according to the memory access granularity of main memory support.For example, write request is split as the unit take 512B, the memory access granularity of supposing the main memory support is 128B, and then read request corresponding to this write request can be 4.The benefit of processing like this is not need to arrange buffering in read data, has read the again retaking of a year or grade response of 512B data by the time, can whenever read the 128B data and read response with regard to directly returning.And the memory access granularity of supposition main memory support is 64B, and then read request corresponding to this write request can be 8.
With the corresponding write request of every i read request, be equivalent to set up the follow-up driving mapping table that drives write request by the response of read request, follow-up write request drives and can carry out according to this table.
After message is split, just can the actual processing of carrying out message.Described request by message sequentially sends read request and refers to send read request by the address sequencing of the read request after splitting.
After sending read request, sending of the request of subsequent message is exactly to drive according to above-mentioned response, and any one write request only just can be sent after the response that has collected its corresponding read request.And the answer word that the expression entire message is finished only just can send after the response that has collected its all write request.
Summary is got off, it is sequentially to carry out that above-mentioned execution mode only has sending of read request, sending then of follow-up write request decided by the degree of returning of reading to respond fully, thereby be actually separate execution between each write request, whether whether each write request only need be paid close attention to relative read request and finish, namely have response to get final product.Based on described explanation, in the situation that multiple messages is concurrent, the processing procedure of each bar message also is separately independently, there is no each other contact.The processing of every message all can be by special-purpose separately link.Therefore, such mode is fit to the applied environment of high bandwidth very much, and higher to the utilance of link.
It seems from the complete process of above execution mode, but because each bar message parallel processing, and finishing dealing with of every message is the Determines that is moved by the actual computer network system, it is out of order reading to respond with returning of write response, sequence independence with the message initiation, therefore solved in the prior art owing to front a piece of news is not finished the problem of blocking subsequent message, significantly improved the throughput of Message Processing.
With reference to shown in Figure 2, a kind of execution mode of message engine of the present invention, it comprises for the treatment of message transmission between each communication node in the network system: split cells 10, read request control unit 20, write request control unit 30, answer word transmitting element 40, wherein,
Described split cells 10 splits into a plurality of read requests and a plurality of write request with the message of a corresponding transfer of data, the corresponding write request of the individual read request of every i (i＞1), and information that will split write request is sent to answer word transmitting element 40;
Described read request control unit 20 sends read request by the request of described message order to the source node of this transfer of data, obtain source node corresponding read response after, will describedly read to respond to be sent to write request control unit 30;
Described write request control unit 30, after the response of i read request corresponding to write request of acquisition, destination node to this transfer of data sends this write request, after the corresponding write response that obtains destination node, described write response is sent to answer word transmitting element 40;
Answer word transmitting element 40, according to the information of write request, after the response of all write requests that obtain destination node, answer word to the source node transmission of this transfer of data.
In one embodiment, described read request control unit 20 can comprise the control subelement that equates with write request number after the fractionation, i read request corresponding to the corresponding described write request of control of each control subelement.
In one embodiment, described write request control unit 30 can comprise with split after the control subelement that equates of write request number, each control subelement obtain that corresponding read request control subelement sends read response after, the transmission write request.
Below in conjunction with above-mentioned data transmission method and message engine, give an example to further specify by the processing procedure of single message and multiple messages respectively.
Figure 3 shows that a kind of processing procedure of single message for example.In conjunction with Fig. 2 and shown in Figure 3, after the message of a corresponding transfer of data was initiated, the split cells 10 in the message engine at first split into this message 3 groups of read request R1, R2, R3 and 3 write request W1, W2, W3.Wherein each group read request respectively comprises 3 read requests, the corresponding write request W1 of the read request among the read request group R1, the corresponding write request W2 of the read request among the read request group R2, the corresponding write request W3 of the read request among the read request group R3.After fractionation was finished, the read request control unit 20 in the message engine sent read request with the order of R1, R2, R3.
Continue with reference to shown in Figure 3, read after response returns in all of read request group R3, illustrate that the read request operation among read request group R3 this moment finishes, related data is read from the main memory of source node, and the write request control unit 30 in the message engine sends write request W3 immediately.And the at this moment response of reading of read request group R1, R2 is not all returned, illustrate that relevant read request operation all do not finish, related data is not all read from the main memory of source node, thereby the write request control unit 30 in the message engine does not send write request W1, W2.
Continue with reference to shown in Figure 3, read request group R2 read the response all return after, illustrate that the read request operation among read request group R2 this moment finishes, related data is read from the main memory of source node, and the write request control unit 30 in the message engine sends write request W2 immediately.And the response of reading of read request group R1 is just all returned afterwards, illustrates that the read request operation among read request group R1 this moment has been finished, and related data has been read from the main memory of source node, and the write request control unit 30 in the message engine sends write request W1 immediately.Can see that herein the sequence independence that the sending order of write request W1, W2, W3 and read request send determines and only return degree by the response of each self-corresponding read request.And because write request W3 sends at first, also when reception reads to respond, it has begun to receive write response at read request group R1, R2 herein.Therefore, also be separate between each write request W1, W2, the W3.
Continue with reference to shown in Figure 3, after the write response of write request W1, W2, W3 is all returned, illustrate that this moment, all write request operations were finished, related data has write in the main memory of destination node, answer word transmitting element 40 in the message engine sends the answer word that expression message is finished immediately, represents that namely data transfer is finished between source node and destination node main memory.
Figure 4 shows that the concurrent processing procedure of multiple messages for example.In conjunction with Fig. 2 and shown in Figure 4, current total n bar message is initiated simultaneously, and each bar message is processed in the direction of arrows from left to right, and grid row in the left side represent the read request data, and grid row in the right side represent write response.Split cells 10 in the message engine at first splits every message, and its fractionation mode can be with reference to the related description of above-mentioned single message, the read request order is sent after then respectively each bar message being split by read request control unit 20.Herein, after a plurality of read requests of a corresponding write request all obtain response, returning read request data, thereby activate write request, also is that the write request control unit 30 in the message engine sends corresponding write request.
Continuing has 2 read request data to return with reference to shown in Figure 4 in the message 0,2 write requests of its correspondence also send (2 activating write request) by the write request control unit 30 in the message engine immediately.And respectively having 3 read request data to return in the message 1,2, n, 3 write requests of its correspondence also send (3 activating write request) by the write request control unit 30 in the message engine immediately.
Continue with reference to shown in Figure 4, after write request sent, write response was also returned successively.And simultaneously, the read request data of returning successively also activate corresponding write request successively.
Continue with reference to shown in Figure 4, the response of the write request of message 1 and message n is all returned at first, message engine is after the write response of message 1 and message n is all returned, and by the answer word of answering word transmitting element 40 transmission message 1 and message n, message 1 and message n finish at first.Subsequently, the response of the write request of message 0 is also all returned, and message engine is after the write response of message 0 is all returned, and by the answer word of answering word transmitting element 40 transmission message 0, message 0 is finished.The response of the write request of message 2 is just all returned at last, and the answer word transmitting element 40 of message engine also just sends at last the answer word of message 2, and message 2 is finished at last.
Can see from the processing procedure that above-mentioned multiple messages is concurrent, although each bar message is initiated simultaneously, the deadline of each bar message is also different.Each bar message is in processing procedure, and with other message and onrelevant, its deadline is by the write response decision of reading to respond and drive the answer word that drives write request in the processing procedure.
Comprehensive above explanation, described message engine also can all or part ofly be integrated in the communication node according to functional requirement.
With reference to shown in Figure 5, a kind of embodiment of communication node of the present invention comprises:
Read request control unit 102 is sequentially carried out read request by the request of described message;
With reference to shown in Figure 6, the another kind of embodiment of communication node of the present invention comprises:
Network receiving element 201 receives the information of the write request of a corresponding transfer of data, and to should item number reportedly defeated read request read response, the corresponding write request of the individual read request of every i (i＞1);
Write request control unit 202 after the response of i read request corresponding to write request of acquisition, is carried out this write request;
Answer word transmitting element 203, according to the information of write request, produce and answer word should item number reportedly after the response of all defeated write requests obtaining;
Network transmitting element 204 will be answered word and will be sent to network.
Correspondingly, the present invention also provides a kind of network system that comprises above-mentioned message engine.
Correspondingly, the present invention also provides a kind of network system that comprises above-mentioned communication node.
In sum, data transmission method of the present invention, message engine, communication node and network system, after the message order is initiated, response according to data comes driving request, write request need to receive read accordingly the response after send, the answer word that message is finished need to send after all write responses of receiving described message.The processing procedure of message has formed the mode of data flow driven, thereby can realize the parallel processing of a plurality of message, has eliminated head and has blocked.For the situation of bandwidth applications, described message based data transmission method can significantly improve the utilance of throughput and the link of Message Processing especially.
Although the present invention discloses as above with preferred embodiment, the present invention is defined in this.Any those skilled in the art without departing from the spirit and scope of the present invention, all can make various changes or modifications, so protection scope of the present invention should be as the criterion with the claim limited range.
Claims (12)
1. a data transmission method is applied to high bandwidth, it is characterized in that, comprising:
The message of a corresponding transfer of data is split into a plurality of read requests and a plurality of write request, and every i corresponding write request of read request is to set up the follow-up driving mapping table that drives write request by the response of read request, i〉1;
Address sequencing by the read request after splitting sends read request;
After the response that has collected i corresponding to a described write request read request, send described write request, the sending order of described write request and the sending order of read request are irrelevant, determine and only return degree by the response of each self-corresponding read request;
After the response that has collected all write requests of described message, send the answer word that the expression entire message is finished.
2. data transmission method as claimed in claim 1 is characterized in that, splits read request and carries out according to the memory access granularity of main memory support.
3. data transmission method as claimed in claim 1 is characterized in that, splits write request boundary's situation is carried out according to message-length and message addresses.
4. a message engine for the treatment of message transmission between each communication node in the network system, is characterized in that, described message engine comprises: split cells, read request control unit, write request control unit, answer the word transmitting element, wherein,
Described split cells, the message of a corresponding transfer of data is split into a plurality of read requests and a plurality of write request, every i corresponding write request of read request, to set up the follow-up driving mapping table that drives write request by the response of read request, and the information of split write request is sent to answer word transmitting element, i 1;
Described read request control unit sends read request by the address sequencing of the read request after splitting to the source node of this transfer of data, obtain source node corresponding read response after, will describedly read to respond to be sent to the write request control unit;
Described write request control unit, after the response that collects i read request corresponding to write request, destination node to this transfer of data sends this write request, the sending order of described write request and the sending order of read request are irrelevant, determine and only return degree by the response of each self-corresponding read request, after the corresponding write response that obtains destination node, described write response is sent to answer word transmitting element;
Answer the word transmitting element, according to the information of write request, after the response of all write requests that collect destination node, send the answer word that the expression entire message is finished to the source node of this transfer of data.
5. message engine as claimed in claim 4 is characterized in that, described read request control unit comprises the control subelement that equates with write request number after the fractionation, i read request corresponding to the corresponding described write request of control of each control subelement.
6. message engine as claimed in claim 4, it is characterized in that, described write request control unit comprise with split after the control subelement that equates of write request number, each control subelement obtain that corresponding read request control subelement sends read response after, the transmission write request.
7. a communication node is characterized in that, comprising:
Split cells splits into a plurality of read requests and a plurality of write request with the message of a corresponding transfer of data, and every i corresponding write request of read request is to set up the driving mapping table of follow-up response driving write request by read request, i〉1;
The read request control unit is carried out read request by the address sequencing of the read request after splitting;
Network communication unit, the information of write request after splitting and the response of reading of described read request are sent to network, the sending order of described write request and the sending order of read request are irrelevant, determine and only return degree by the response of each self-corresponding read request, and from network, receive to should item number defeated answer word reportedly.
8. communication node as claimed in claim 7 is characterized in that, described read request control unit comprises the control subelement that equates with write request number after the fractionation, i read request corresponding to the corresponding described write request of control of each control subelement.
9. a communication node is characterized in that, comprising:
The network receiving element receives the information of the write request of a corresponding transfer of data, and to should item number reportedly defeated read request read response, write request of every i read request correspondence, follow-up response by i read request drives described write request, i〉1;
The write request control unit after the response that collects i read request corresponding to write request, is carried out this write request, and the sending order of described write request and the sending order of read request are irrelevant, determines and only return degree by the response of each self-corresponding read request;
Answer the word transmitting element, according to the information of write request, obtaining to produce the answer word that the expression transfer of data is finished to should item number reportedly after the response of all defeated write requests;
The network transmitting element will be answered word and will be sent to network.
10. communication node as claimed in claim 9 is characterized in that, described write request control unit comprises the control subelement that equates with the write request number, each control subelement obtain corresponding read response after, carry out write request.
11. network system that comprises each described message engine of claim 4 to 6.
12. network system that comprises each described communication node of claim 7 to 10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910195309 CN102014111B (en) | 2009-09-04 | 2009-09-04 | Data transmission method, message engine, communication node and network system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910195309 CN102014111B (en) | 2009-09-04 | 2009-09-04 | Data transmission method, message engine, communication node and network system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102014111A CN102014111A (en) | 2011-04-13 |
| CN102014111B true CN102014111B (en) | 2013-09-18 |
Family
ID=43844127
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200910195309 Active CN102014111B (en) | 2009-09-04 | 2009-09-04 | Data transmission method, message engine, communication node and network system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102014111B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108337286A (en) * | 2017-01-20 | 2018-07-27 | 深圳市中兴微电子技术有限公司 | One kind cutting packet method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5235595A (en) * | 1987-05-06 | 1993-08-10 | Fischer & Porter Company | Packet switching |
| CN101014050A (en) * | 2007-02-02 | 2007-08-08 | 浙江大学 | Connection-based communication protocol of CAN bus applied for a great deal of data |
| CN101382955A (en) * | 2008-09-28 | 2009-03-11 | 中国科学院计算技术研究所 | File reading method in cluster file system and system |
-
2009
- 2009-09-04 CN CN 200910195309 patent/CN102014111B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5235595A (en) * | 1987-05-06 | 1993-08-10 | Fischer & Porter Company | Packet switching |
| CN101014050A (en) * | 2007-02-02 | 2007-08-08 | 浙江大学 | Connection-based communication protocol of CAN bus applied for a great deal of data |
| CN101382955A (en) * | 2008-09-28 | 2009-03-11 | 中国科学院计算技术研究所 | File reading method in cluster file system and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102014111A (en) | 2011-04-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1205562C (en) | Split transaction protocol for a bus system | |
| CN1049059C (en) | A real time data imaging network system and a method of operating same | |
| CN100368998C (en) | Multiple data base data syne method | |
| CN104615684B (en) | A kind of mass data communication concurrent processing method and system | |
| CN102387179A (en) | Distributed file system and nodes, saving method and saving control method thereof | |
| CN1376280A (en) | Transaction scheduling for a bus system | |
| CN1866910A (en) | Data message transmission method and Ethernet bridge apparatus based on VLAN | |
| CN85107738A (en) | On-off controller | |
| CN102378971B (en) | Method for reading data and memory controller | |
| CN1329809C (en) | Controller of magnetic disk array and its working method | |
| WO2004028123A8 (en) | System and method of accessing contact information on a communication device | |
| CN1859282A (en) | Storage controller and control method | |
| CN1574785A (en) | Method and system for maintenance of packet order using caching | |
| CN106503020A (en) | Daily record data processing method and processing device | |
| CN104866528A (en) | Multi-platform data acquisition method and system | |
| CN112948293A (en) | DDR arbiter and DDR controller chip of multi-user interface | |
| CN103326925A (en) | Message push method and device | |
| CN102014111B (en) | Data transmission method, message engine, communication node and network system | |
| CN101526925B (en) | Processing method of caching data and data storage system | |
| CN103123575A (en) | Data write-in method | |
| CN1313903C (en) | Method and apparatus for obtaining physical layer chip condition information | |
| CN105740166A (en) | Cache reading and reading processing method and device | |
| CN106897021A (en) | A kind of method and apparatus for reading and writing data | |
| CN105721338A (en) | Method and device for processing received data | |
| CN105162722B (en) | A kind of message classification dispatching method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |