Capacity planning for computing systems hosting multi-tier application based on think time value and resource cost of composite transaction using statistical regression analysis

• US 9,135,075 B2
• Filed: 03/09/2007
• Issued: 09/15/2015
• Est. Priority Date: 03/09/2007
• Status: Active Grant

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Cherkasova, Ludmila, Zhang, Qi, Mathews, Guy, Greene, Wayne
• Primary Examiner(s): Kawsar, Abdullah Al
• Application Number:
  US11/684,569
• Publication Number:
  US 20080221941A1
• Time in Patent Office: 3,112 Days
• Field of Search: 718/104
• US Class Current: 1/1
• CPC Class Codes: G06F 11/3409 for performance assessment
• CPC Class Codes: G06F 11/3414 Workload generation, e.g. scripts, playback
• CPC Class Codes: G06F 11/3423 where the assessed time is active or idle time
• CPC Class Codes: G06F 11/3442 for planning or managing the needed capacity
• CPC Class Codes: G06F 11/3452 Performance evaluation by statistical analysis
• CPC Class Codes: G06F 2201/87 Monitoring of transactions
• CPC Class Codes: G06F 2209/508 Monitor
• CPC Class Codes: G06F 9/505 considering the load

Abstract

A method comprises receiving, by a workload profiler, a representative workload of a computing system under analysis. The workload profiler determines a workload profile of the computing system that reflects a transaction mix that varies over time. A capacity analyzer receives the workload profile, and determines a maximum capacity of the computing system under analysis for serving the workload profile while satisfying a defined quality of service (QoS) target.

Claims

1. A method comprising:

receiving, by a workload profiler, a representative workload of a computing system under live analysis;

determining, by the workload profiler, a workload profile of the computing system that reflects a transaction mix that varies over time,

wherein the workload profile includes, for each of a plurality of monitoring windows, a plurality of composite transactions where a mix of the plurality of composite transactions varies across the plurality of monitoring windows;

wherein workload profiler further determines a think time value for at least one session in the representative workload, said at least one session comprising a plurality of related composite transactions for providing a service to a client, and said think time value reflecting idle time of the computing system between the plurality of related composite transactions in the at least one session;

determining, by a resource cost calculator, resource cost for at least one composite transaction in the representative workload, said resource cost reflecting an amount of utilization of at least one resource of the computing system in serving the at least one composite transaction,

wherein the amount of utilization is determined by determining for one of the plurality of composite transactions an approximate amount of utilization using a resource cost calculator and a statistical regression-based analysis;

receiving, by a capacity analyzer, the determined think time value for the at least one session and the determined resource cost for the at least one composite transaction; and

determining, by the capacity analyzer, a maximum number of clients for each of a plurality of different tiers of at least one multi-tier application supportable by the computing system under analysis for serving the workload profile while satisfying a defined quality of service (QoS) target based on the think time value and the determined resource cost;

wherein said capacity analyzer employs an analytical model based on a network of queues representing different tiers of the at least one multi-tier application for determining the maximum number of clients for each of a plurality of different tiers of the at least one multi-tier application; and

performing, by the capacity analyzer, capacity planning based on the determined maximum number of clients.

2. The method of claim 1 wherein the workload profile reflects each of a plurality of composite transaction types present in the representative workload as a fraction of overall transaction types present in the representative workload.

3. The method of claim 1, further comprising:

wherein said statistical regression-based analysis comprises;

determining an approximate utilization (U′

_CPU,n) as;

4. The method of claim 1 further comprising:

determining, by the workload profiler, a subset of composite transactions included in the received representative workload to include in the workload profile.

5. The method of claim 4 wherein said determining the subset comprises:

determining K most popular composite transaction types in the representative workload, wherein K is a number that is at least 1 and that is less than a total number of composite transaction types included in the representative workload.

6. The method of claim 5 further comprising:

determining, by a resource cost calculator, resource cost for composite transactions included in the representative workload that are of the determined K most popular composite transaction types, said resource cost reflecting an amount of utilization of at least one resource of the computing system in serving the at least one composite transaction.

7. The method of claim 1 wherein said workload profiler and said capacity analyzer comprise computer-executable software code stored to a non-transitory computer-readable medium.

8. The method of claim 1 further comprises:

determining a maximum number of clients that can be served by the computing system under analysis while satisfying said defined QoS target.

9. The method of claim 8 wherein said maximum number of clients that the computing system can support comprises a maximum number of clients that the computing system can serve within a defined time interval in compliance with said defined QoS target.

10. The method of claim 8 wherein said maximum number of clients comprises a maximum number of clients that the computing system can serve concurrently in compliance with said defined QoS target.

11. The method of claim 1 wherein said defined QoS target comprises a defined maximum response time for said computing system to respond to a client request.

12. A system comprising:

at least one processor;

at least one multi-tier application hosted by a computing system;

a workload profiler that receives a representative workload of the computing system under live analysis, said representative workload comprising a plurality of composite transactions, wherein a mix of the plurality of composite transactions vary over a plurality of monitoring windows;

said workload profiler further determines a think time value for at least one session in the representative workload, said at least one session comprising a plurality of related composite transactions for providing a service to a client, and said think time value reflecting idle time of the computing system between the plurality of related composite transactions in the at least one session;

a resource cost calculator that determines resource cost for at least one composite transaction in the representative workload, said resource cost reflecting an amount of utilization of at least one resource of the computing system in serving the at least one composite transaction,

wherein the amount of utilization is determined by determining for one of the plurality of composite transactions an approximate amount of utilization using the resource cost calculator and a statistical regression-based analysis;

a capacity analyzer that receives the determined think time value for the at least one session and the determined resource cost for the at least one composite transaction, and determines a maximum number of clients for each of a plurality of different tiers of the at least one multi-tier application supportable by the computing system under analysis while satisfying a defined quality of service (QoS) target based on the determined think time value and the determined resource cost; and

wherein said capacity analyzer employs an analytical model based on a network of queues representing different tiers of the at least one multi-tier application for determining the maximum number of clients for each of a plurality of different tiers of the at least one multi-tier application and performs capacity planning based on the determined maximum number of clients.

13. The system of claim 12 further comprising:

said workload profiler further determines a subset of the plurality of composite transactions in the representative workload, said subset comprising K most popular composite transaction types in the representative workload, wherein K is a number that is at least 1 and that is less than a total number of composite transaction types included in the representative workload; and

said at least one composite transaction for which said resource cost calculator determines said resource cost comprises composite transactions included in said determined subset.

14. The system of claim 12 wherein said resource cost calculator comprises a regression-based solver that performs the statistical regression-based analysis for determining said resource cost.

15. A method comprising:

receiving, by a workload profiler, a representative workload of a computing system under live analysis, said representative workload comprising a plurality of composite transactions;

determining, by the workload profiler, a subset of the plurality of composite transactions for which a corresponding resource cost is to be determined, wherein the subset of the plurality of composite transactions comprises, for each of a plurality of monitoring windows, a plurality of composite transactions, where a mix of the plurality of composite transactions varies across the plurality of monitoring windows;

determining, by the workload profiler, a think time value for at least one session in the representative workload, said at least one session comprising a plurality of related composite transactions for providing a service to a client, and said think time value reflecting idle time of the computing system between the plurality of related composite transactions in the at least one session;

determining, by a regression-based solver, a resource cost for the at least one composite transaction in the determined subset, said resource cost reflecting an amount of utilization of at least one resource of the computing system in serving the at least one composite transaction in the determined subset;

wherein determining the resource cost is based on an approximated CPU cost of average transaction service times of the computing system; and

determining, by a capacity analyzer, based at least in part on the determined resource cost for the at least one composite transaction and the determined think time value for the at least one session, a maximum number of clients for each of a plurality of different tiers of at least one multi-tier application supportable by the computing system under analysis while satisfying a defined quality of service (QoS) target;

wherein said capacity analyzer employs an analytical model based on a network of queues representing different tiers of the at least one multi-tier application for determining the maximum number of clients for each of a plurality of different tiers of the at least one multi-tier application; and

performing, by the capacity analyzer, capacity planning based on the determined maximum number of clients.

16. The method of claim 15 further comprises:

determining a maximum capacity of the computing system for serving a workload while satisfying a defined quality of service (QoS) target.

17. The method of claim 16 further comprises:

determining a maximum number of clients that can be served by the computing system while satisfying said defined QoS target.