LOAD BALANCING USING REPLICATION DELAY

• US 20100036952A1
• Filed: 08/11/2008
• Published: 02/11/2010
• Est. Priority Date: 08/11/2008
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Golwalkar, Yogesh Vilas, Rajamani, Magesh, Hazlewood, Kristin Marie
• Granted Patent: US 7,890,632 B2
• Time in Patent Office: Days
• US Class Current: 709/226
• CPC Class Codes: G06F 15/173 using an interconnection network, e.g. matrix, shuffle, pyramid, star, snowflake
• CPC Class Codes: H04L 61/4523 using lightweight directory access protocol [LDAP]

Abstract

A method, system, and computer usable program product for load balancing using replication delay are provided in the illustrative embodiments. In response to a request to update, a system updates data associated with a write server, forming updated data of a data partition. The system receives a read request for the data partition. The system calculates a time difference between an arrival time of the request to update and an arrival time of the read request. The system receives a set of average replication delays for a set of replica servers serving the data partition. The system directs the read request to a replica server in the set of replica servers whose average replication delay is less than or equal to the time difference.

Claims

1. A computer implemented method for load balancing using replication delay , the computer implemented method comprising:

updating, responsive to a request to update, a data associated with a write server, forming updated data of a data partition;

receiving a read request for the data partition;

calculating a time difference between an arrival time of the request to update and an arrival time of the read request;

receiving a set of average replication delays for a set of replica servers serving the data partition; and

directing the read request to a replica server in the set of replica servers whose average replication delay is one of (i) less than and (ii) equal to the time difference.

2. The computer implemented method of claim 1, further comprising:

replicating the updated data at the replica server from the data associated with the write server, wherein the replicating occurs after a replication delay.

3. The computer implemented method of claim 1, further comprising:

directing the read request to the write server if no replica server in the set of replica servers has an average replication delay of one of (i) less than and (ii) equal to the time difference.

5. The computer implemented method of claim 1, wherein the data partition is a directory partition, wherein the write server and the set of replica servers are servers serving the directory partition.

4. The computer implemented method of claim 4, further comprising:

receiving a response to the read request from the replica server; and

sending the response, responsive to the read request.

6. A computer implemented method for load balancing using replication delay, the computer implemented method comprising:

receiving a replication delay information for a replica server for each request to update over one of (i) a preset number of requests to update and (ii) all requests to update received during a present amount of time;

determining a replication delay for each request to update;

computing an average replication delay from the replication delays for each request to update; and

sending the average replication delay for the replica server to a load balancing system.

7. The computer implemented method of claim 6, wherein the replication delay information comprises:

a first time when an update corresponding to a first request to update became available in a data source; and

a second time when the update became available for serving from the replica server.

8. The computer implemented method of claim 7, further comprising:

resetting a starting time for a replica server at the first time.

9. The computer implemented method of claim 6, further comprising:

adding additional time to the average replication delay, wherein the additional time is one of (i) a percentage of the average replication delay and (ii) a preset amount of time.

10. The computer implemented method of claim 6, further comprising:

storing the replication delay information; and

storing the average replication delay.

11. A data processing system for load balancing using replication delay, the data processing system comprising:

a storage device, wherein the storage device stores computer usable program code; and

a processor, wherein the processor executes the computer usable program code, and wherein the computer usable program code comprises;

computer usable code for updating, responsive to a request to update, a data associated with a write server, forming updated data of a data partition;

computer usable code for receiving a read request for the data partition;

computer usable code for calculating a time difference between an arrival time of the request to update and an arrival time of the read request;

computer usable code for receiving a set of average replication delays for a set of replica servers serving the data partition; and

computer usable code for directing the read request to a replica server in the set of replica servers whose average replication delay is one of (i) less than and (ii) equal to the time difference.

12. The computer implemented method of claim 11, further comprising:

computer usable code for replicating the updated data at the replica server from the data associated with the write server, wherein the replicating occurs after a replication delay.

13. The computer implemented method of claim 11, further comprising:

computer usable code for directing the read request to the write server if no replica server in the set of replica servers has an average replication delay of one of (i) less than and (ii) equal to the time difference.

15. The computer implemented method of claim 11, wherein the data partition is a directory partition, wherein the write server and the set of replica servers are servers serving the directory partition.

14. The computer implemented method of claim 14, further comprising:

computer usable code for receiving a response to the read request from the replica server; and

computer usable code for sending the response, responsive to the read request.

16. A computer usable program product comprising a computer usable medium including computer usable code for load balancing using replication delay, the computer usable code comprising:

computer usable code for receiving a replication delay information for a replica server for each request to update over one of (i) a preset number of requests to update and (ii) all requests to update received during a present amount of time;

computer usable code for determining a replication delay for each request to update;

computer usable code for computing an average replication delay from the replication delays for each request to update; and

computer usable code for sending the average replication delay for the replica server to a load balancing system.

17. The computer usable program product of claim 16, wherein the replication delay information comprises:

a first time when an update corresponding to a first request to update became available in a data source; and

a second time when the update became available for serving from the replica server.

18. The computer usable program product of claim 17, further comprising:

computer usable code for resetting a starting time for a replica server at the first time.

19. The computer usable program product of claim 16, further comprising:

computer usable code for adding additional time to the average replication delay, wherein the additional time is one of (i) a percentage of the average replication delay and (ii) a preset amount of time.

20. The computer usable program product of claim 16, further comprising:

computer usable code for storing the replication delay information; and

computer usable code for storing the average replication delay.