System and method for maintaining replicated data coherency in a data processing system

• US 5,434,994 A
• Filed: 05/23/1994
• Issued: 07/18/1995
• Est. Priority Date: 05/23/1994
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Yellepeddy, Krishna K., Shaheen, Amal A.
• Primary Examiner(s): Lall, Parshuiam S.
• Assistant Examiner(s): LINTSAI, PAULIN
• Application Number:
  US08/247,422
• Time in Patent Office: 421 Days
• Field of Search: 395/200, 395/500, 395/600
• US Class Current: 709/223
• CPC Class Codes: G06F 12/0815 Cache consistency protocols
• CPC Class Codes: G06F 16/1767 Concurrency control, e.g. optimistic or pessimistic approaches
• CPC Class Codes: Y10S 707/99953 Recoverability
• CPC Class Codes: Y10S 707/99954 Version management

Abstract

A system and method for maintaining data coherency in a system in which data is replicated on two or more servers. Each server is able to update the data replica present on the server. Updates are logged for each server. Reconciliation of server data replicas is aggressively initiated upon the occurrence of predefined events. These events include arrival at a scheduled time, a request for data by a client system, server and network failure recovery. Reconciliation is managed by a coordinator server selected to ensure that at most one coordinator server per network partition is selected. Logged updates are merged and transmitted to each server containing a data replica. The logged updates are applied unless a conflict is detected. Conflicts are collected and distributed for resolution. Reconciliation is managed between servers without regard to operating system or physical file system type.

Claims

1. A system for maintaining coherency of replicated data in a distributed data processing system having a plurality of processors interconnected by a communications medium and a plurality of storage devices, each of said storage devices controlled by at least one of said plurality of processors, said replicated data being replicated on two or more of said plurality of storage devices, said system comprising:

means for detecting an event occurring in said data processing system;

means for initiating an update of said replicated data in response to said detected event;

means for selecting one of said plurality of processors to coordinate said update of said replicated data;

said selected one of said plurality of processors to coordinate said update of said replicated data including means for updating said replicated data, said means for updating causing each copy of said replicated data to become a duplicate of all other copies of said replicated data, said updating means including;

means for requesting modification logs from said plurality of processors;

means for receiving all requested modification logs;

means for merging all received modification logs; and

means for sending said merged modification logs to said plurality of processors.

2. The system of claim 1, wherein certain of said plurality of processors are designated server processors and wherein said means for selecting selects a coordinating processor from said server processors.

3. The system of claim 1, wherein said means for selecting includes:

means for establishing a priority for each of said plurality of processors;

means for querying each higher priority processor to determine which of said each higher priority processor will coordinate said update of said replicated data;

means for detecting responses to said query; and

means for signalling assumption of coordination role to processors of lesser priority upon selection of one of said plurality of processors to coordinate said update of said replicated data.

4. A method of aggressively maintaining data coherency between two or more replicas of data distributed between two or more servers, said method comprising the steps of:

logging changes to one of said two or more data replicas;

detecting an event occurring in one or more of said two or more servers;

initiating reconciliation of said two or more replicas;

selecting one of said two or more servers to coordinate said reconciliation;

determining changes needed to each of said two or more replicas based on said logged changes to said one or said two or more data replicas;

making said determined changes to said each of said two or more replicas and detecting any data conflicts; and

recording data conflicts associated with said determined changes made to said each of said two or more replicas for later resolution.

5. The method of claim 4, wherein said event includes a scheduled update, return to service of a server, and network link reconnection.

6. The method of claim 4, wherein said selected one of said two or more servers is controlled by an operating system that differs from an operating system of at least one other of said two or more servers.

7. The method of claim 4, wherein each of said two or more data replicas is stored on a physical file system and said selected one of said two or more servers maintains a physical file system that differs from a physical file system of at least one other of said two or more servers.

8. The method of claim 4, wherein said step of selecting one of said two or more servers to coordinate said reconciliation comprises the steps of:

assigning a priority to each of said two or more servers;

sending by each of said two or more servers a message to each server of a higher priority querying whether said higher priority server will become said selected one of said two or more servers to coordinate said reconciliation;

detecting higher priority server responses to said querying messages sent to each of said two or more of said servers; and

if no response from a higher priority server is detected by a particular one of said two or more of said servers, sending a message by said particular one of said two or more of said servers to each server of a lower priority signalling assumption of coordinator server role by said particular one of said two or more of said servers.

9. The method of claim 8, wherein one and only one coordinator server is selected for each system partition.

10. The method of claim 9, wherein at least two of said two or more of said servers are controlled by different operating systems.

11. The method of claim 10, further comprising:

means for detecting coordinator server failure and causing said step of selecting one of said two or more servers to be repeated.