Method and system for deciding when to checkpoint an application based on risk analysis

• US 7,392,433 B2
• Filed: 01/25/2005
• Issued: 06/24/2008
• Est. Priority Date: 01/25/2005
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Oliner, Adam J., Sahoo, Ramendra K.
• Primary Examiner(s): McCarthy; Christopher S
• Application Number:
  US11/042,611
• Publication Number:
  US 20060168473A1
• Time in Patent Office: 1,246 Days
• Field of Search: None
• US Class Current: 714/38.13
• CPC Class Codes: G06F 11/1471 involving logging of persistent data for recovery

Abstract

Briefly, according to the invention in an information processing system including a plurality of information processing nodes, a request for checkpointing by an application includes node health criteria (or parameters). The system has the authority to grant or deny the checkpointing request depending on the system health or availability. This scheme significantly improves not only the system performance, but also the application running time as the system. By skipping a checkpoint the application can use the same time to run the application instead of spending extra time for checkpointing.

Claims

1. In a multi-node information processing system, a method for checkpointing an application comprising steps of:

receiving from an application a request for checkpointing, wherein the request includes node health-related parameters;

determining whether to grant the checkpointing request, based on node health parameters;

determining a last time when the application was checkpointed; and

determining how long the application takes to be checkpointed.

2. The method of claim 1 wherein the health-related parameters comprise node failure rates.

3. The method of claim 1 wherein the health-related parameters comprise error rates.

4. The method of claim 1 wherein determining whether to grant the checkpointing request comprises determining a probability of occurrence of a failure within a specified time window.

5. The method of claim 1 wherein the request for checkpointing comprises the node health-related parameters.

6. The method of claim 1, wherein determining whether to grant or deny the checkpoint request comprises calculating an expected loss of work and a cost of checkpointing interval.

7. The method of claim 1 further comprising determining a set of node health-related parameters affecting a job or an application running in the system before receiving the request for check pointing.

8. In a multi-node information processing system, a method for checkpointing an application comprising steps of:

receiving from an application a request for checkpointing, wherein the request includes node health-related parameters;

determining a risk associated with shipping a checkpoint associated with the application;

comparing a cost of checkpointing with the risk of shipping the checkpoint;

granting the request for checkpointing if the risk exceeds the cost.