Method for improving performance in a computer storage system by regulating resource requests from clients

• US 7,228,354 B2
• Filed: 06/28/2002
• Issued: 06/05/2007
• Est. Priority Date: 06/28/2002
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Chambliss, David Darden, Jadav, Divyesh
• Primary Examiner(s): Flynn; Nathan J.
• Assistant Examiner(s): Nguyen; Dustin
• Application Number:
  US10/187,227
• Publication Number:
  US 20040003087A1
• Time in Patent Office: 1,803 Days
• Field of Search: 709/226, 709/227, 709/245, 709/225, 711/121, 711/211, 710/15, 710/5, 710/6, 718/104
• US Class Current: 709/229
• CPC Class Codes: G06F 11/1076 Parity data used in redundant arrays of independent storages, e.g. in RAID systems
• CPC Class Codes: H04L 41/083 for increasing network speed
• CPC Class Codes: H04L 41/0896 Bandwidth or capacity management, i.e. automatically increasing or decreasing capacities flow or congestion control using dynamic resource allocation, e.g. in-call renegotiation, H04L47/76
• CPC Class Codes: H04L 41/5022 by giving priorities, e.g. assigning classes of service

Abstract

A method and system for optimizing the performance of a storage system by classifying each client request for resources based on operational limits of the resources and controlling when to submit the request for processing based on service class. The operational limits are determined from performance characteristics of the system resources and from the level of performance guaranteed to each client. By regulating the clients'"'"' usage of resources using the resource operational limits, total system performance requirements and guarantees can be achieved.

Claims

1. In a storage system having one or more resources, one or more clients, a request classifier and a flow controller, a method for optimizing performance comprising steps of:

receiving service requests from clients;

transmitting the service requests to a gateway for recording timestamps on the service requests;

sending the service requests from the gateway to the request classifier;

associating, by the request classifier, each service request with a service class, said service class corresponding to a set of operational limits for limiting each service request'"'"'s use of the storage system'"'"'s resources, wherein the set of operational limits are determined from performance characters of the system resources and from level of performance guaranteed to each client;

determining, by the flow controller, with use of the set of operational limits corresponding to the service class, whether and by how much time the submission of a service request for processing should be delayed; and

submitting the service request to the one or more resources for processing based on the determining step of the flow controller.

2. The method of claim 1 wherein the set of operational limits corresponds to desired maximum rates at which the resources process the service requests in said service class so as not to affect the service requests in other service classes that need the same resources.

3. The method of claim 1 wherein the set of operational limits includes a rate of transmitting data between the clients and the resources.

4. The method of claim 1 wherein the set of operational limits includes a rate of performing disk I/O operations.

5. The method of claim 1 wherein the set of operational limits includes a rate of performing cache-miss disk I/O operations.

6. The method of claim 1 wherein the set of operational limits includes a rate of performing non-sequential operations by the resources.

7. The method of claim 1 wherein the storage system includes multiple channels for accessing data at the resources and the set of operational limits includes a rate of data transmission over the channels.

8. The method of claim 1 wherein the set of operational limits includes a rate at which cache pages are displaced.

9. The method of claim 2 wherein a single set of operational limits applies to a total usage rate for all resources belonging to a same workload group.

10. The method of claim 2 wherein separate sets of operational limits respectively apply to separate resources of the same workload group.

11. The method of claim 2 wherein the flow controller delays the submission of a request to ensure that the service requests submitted for a given service class do not exceed the operational limits corresponding to said service class.

12. The method of claim 2 wherein the flow controller rejects some of the service requests to ensure that those submitted for a particular service class do not exceed the operational limits for said particular service class.

13. The method of claim 2 wherein:

each service class corresponds to a set of service level objectives, said service level objectives including a desired response time for processing the service requests in said service class during a time interval and a usage constraint associated with each resource, and the desired response time should be achieved whenever the usage constraints is satisfied.

14. The method of claim 13 wherein the desired response time is such that an aggregate response time of all service requests be less than a target response time for said service class.

15. The method of claim 13 wherein the desired response time is such that a percentage of the service requests whose response time exceeds a target response time be less than a tolerance limit for said service class.

16. The method of claim 13 wherein the resource usage constraint is such that actual usage of a resource be less than a limit corresponding to said resource and said service class.

17. The method of claim 13 wherein the resource usage constraint is such that actual usage of a resource be less than a first limit corresponding to said resource and said service class, and that a peak requested usage rate be less than a second limit corresponding to said resource and said service class.

18. The method of claim 14 wherein the service level objectives are evaluated to determine the resource operational limits.

19. The method of claim 18 wherein the evaluation includes considering expected and worst-case utilization of the resources in view of the service level objectives.

20. The method of claim 19 further comprising the step of determining whether the storage system could meet the target aggregate response time based on said evaluation.

21. The method of claim 20 further comprising the steps of:

accepting the service level objectives if they can be achieved based on said evaluation;

rejecting the service level objectives if they cannot be achieved based on said evaluation; and

applying the accepted service level objectives as the resource operational limits.

22. The method of claim 20 wherein the evaluation includes comparing actual performance of the storage system with past operational limits and resource usage.

23. A storage system with improved performance comprising:

one or more resources;

one or more clients having requests for resources;

a gateway for recording request timestamps;

a request classifier for associating each request with a service class, said service class corresponding to a set of operational limits for limiting each request'"'"'s of the storage system'"'"'s resources, wherein the set of operational limits are determined from performance characteristics of the system resources and from level of performance guaranteed to each client; and

a flow controller for determining, with use of the set of operational limits corresponding to the service class, whether and by how much time the submission of a request for processing should be delayed submitting the request to the one or more resources for processing based on the determining of the flow controller.

24. A computer-program product stored on a storage system to improve performance, the system having one or more resources, one or more clients having requests for resources, a request classifier and a flow controller, the computer-program product comprising computer program code for:

directing a gateway to record a timestamp for each request;

directing the request classifier to associate each request with a service class, said service class corresponding to a set of operational limits for limiting each request'"'"'s of the storage system'"'"'s resources, wherein the set of operational limits are determined from performance characteristics of the system resources and from level of performance guaranteed to each client;

directing the flow controller to determined, with use of the set of operational limits corresponding to the service class, whether and by how much time the submission of a request for processing should be delayed; and

directing the flow controller to submit the request to one or more resources for processing based on the determining of the flow controller.