Method, apparatus, and program for measuring server performance using multiple clients

• US 6,961,769 B2
• Filed: 09/20/2001
• Issued: 11/01/2005
• Est. Priority Date: 09/20/2001
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Arora, Rajiv, Gordon, Jesse Mannes
• Primary Examiner(s): Walsh, John B.
• Application Number:
  US09/956,767
• Publication Number:
  US 20030055956A1
• Time in Patent Office: 1,503 Days
• Field of Search: 709/224
• US Class Current: 709/224
• CPC Class Codes: G06F 11/3428 Benchmarking

Abstract

Server performance is bench marked using multiple master processes. A local master process is created for each client computer system applying load on the server. Then a global master process is created on a physical computer system on which neither client nor server processes are running. The global master is responsible for the proper execution of the bench mark. The global master uses an inter-master communications protocol that enforces coordination among the disparate client systems and processes.

Claims

1. A method, in a global master process, for measuring server performance, comprising:

invoking a plurality of local master processes, wherein each of the plurality of local master processes executes on a separate client in a network;

instructing the plurality of local master processes to begin a bench mark phase;

receiving results from the plurality of local master processes;

(a) instructing the plurality of local master processes to run an iteration of a test;

(b) receiving a plurality of metrics from each of the plurality of local master processes;

(c) computing a confidence interval for the plurality of metrics for each of the plurality of local master processes;

(d) determining whether each confidence interval is less than a respective threshold; and

repeating steps (a) through (d) until all confidence intervals are less than their respective thresholds.

2. The method of claim 1, wherein the global master process executes on a first client.

3. The method of claim 2, wherein at least one of the local master processes executes on a second client.

4. The method of claim 1, wherein each of the local master processes applies load on a server.

5. An apparatus for measuring server performance in a global master process, comprising:

invocation means for invoking a plurality of local master processes, wherein each of the plurality of local master processes executes on a separate client in a network;

signal means for instructing the plurality of local master processes to begin a bench mark phase;

receipt means for receiving results from the plurality of local master processes,

(a) instructing means for instructing the plurality of local master processes to run an iteration of a test;

(b) receiving means for receiving a plurality of metrics from each of the plurality of local master processes;

(c) computing means for computing a confidence interval for the plurality of metrics for each of the plurality of local master processes;

(d) determining means for determining whether each confidence interval is less than a respective threshold; and

repeating means for repeating the means recited in (a) through (d) until all confidence intervals are less than their respective thresholds.

6. The apparatus of claim 5, wherein the global master process executes on a first client.

7. The apparatus of claim 6, wherein at least one of the local master processes executes on a second client.

8. The apparatus of claim 5, wherein each of the local master processes applies load on a server.

9. An apparatus for measuring server performance, comprising:

a global master client, wherein the global master client executes a global master process;

a plurality of bench mark clients, wherein the plurality of bench mark clients execute a plurality of local master processes; and

a server, wherein the local master processes create worker processes that apply load on the server under control of the global master process, wherein the global master process executes a set of instructions to (a) instruct the plurality of local master processes to run an iteration of a test;

(b) receive a plurality of metrics from each of the plurality of local master processes;

(c) compute a confidence interval for the plurality of metrics for each of the plurality of local master processes;

(d) determine whether each confidence interval is less than a respective threshold; and

repeat instructions (a) through (d) until all confidence intervals are less than their respective thresholds.

10. A computer program product for measuring server performance in a global master process, comprising:

instructions for invoking a plurality of local master processes, wherein each of the plurality of local master processes executes on a separate client in a network;

instructions for instructing the plurality of local master processes to begin a bench mark phase;

instructions for receiving results from the plurality of local master processes;

(a) instructions for instructing the plurality of local master processes to run an iteration of a test;

(b) instructions for receiving a plurality of metrics from each of the plurality of local master processes;

(c) instructions for computing a confidence interval for the plurality of metrics for each of the plurality of local master processes;

(d) instructions for determining whether each confidence interval is less than a respective threshold; and

instructions for repeating instructions (a) through (d) until all confidence intervals are less than their respective thresholds.

11. The computer program product of claim 10, wherein the global master process executes on a first client.

12. The computer program product of claim 11, wherein at least one of the local master processes executes on a second client.

13. The computer program product of claim 10, wherein each of the local master processes applies load on a server.