Crash recovery without complete remirror

(54) CRASH RECOVERY WITHOUT COMPLETE REMIRROR

(75) Inventors: Richard Ohran, Provo; Dan Price, Orem; Randall Johnson, Pleasant Grove, all of UT (US)

(73) Assignee: Legato Systems, Inc., Mountain View, CA (US)

( * ) Notice: Subject to any disclaimer, the term of this patent is extended or adjusted under 35 U.S.C. 154(b) by 0 days.

(21) Appl. No.: 09/455,072

(22) Filed: Dec. 6, 1999

Int. C I. G06F 12/16

U.S. CI 711/162; 707/204; 709/213;

714/6

(58) Field of Search 711/161, 162;

714/6; 707/204; 709/213

(56) References Cited

U.S. PATENT DOCUMENTS

4,342,079 A 7/1982 Stewart et al 364/200

5,155,845 A 10/1992 Beat et al 395/575

5,212,784 A 5/1993 Sparks 395/575

5,649,152 A 7/1997 Ohran et al 395/441

5,764,903 A 6/1998 Yu 395/200.38

5,812,748 A 9/1998 Ohran et al 395/182.02

5,835,953 A 11/1998 Ohran 711/162

5,907,673 A 5/1999 Hirayama 395/182.14

5,978,565 A 11/1999 Ohran et al 395/182.11

(51) (52)

A primary computer system and a backup computer system each have an associated memory. For each write request, a copy of the request is forwarded to a delay buffer and memory queue associated with the primary computer system, and a copy is forwarded to a memory queue of the backup computer system. The backup computer system transmits an acknowledgement signal to the primary computer system when the backup computer system receives its copy of the request. The write request in the delay buffer of the primary computer system is executed in the primary memory only upon receipt of this acknowledgement signal. Thus, the backup computer system knows of every request executed in the primary memory. The write request is executed in the backup memory at any time after the backup computer system receives the write request. The write requests are deleted from the memory queues (primary and backup) when the associated computer system confirms that the write request was executed in the memory of the opposite computer system. Should the primary (or backup) computer system shut down, the requests are accumulated in the opposite backup (or primary) memory queue. When the primary (or backup) computer system becomes operational again, the requests in the opposite backup (or primary) memory queue are executed in the primary (or backup) memory. Thus, no memory is lost when the primary (or backup) computer system shuts down and complete remirroring of data is not required.

31 Claims, 4 Drawing Sheets