| US 7,539,608 B1 | ||
| Techniques for determining effects on system performance of a memory management parameter | ||
| Benoit Dageville, Foster City, Calif. (US); Sambavi Muthukrishnan, Redmond, Wash. (US); and Mohamed Zait, Sunnyvale, Calif. (US) | ||
| Assigned to Oracle International Corporation, Redwood Shores, Calif. (US) | ||
| Filed on Oct. 09, 2002, as Appl. No. 10/268,475. | ||
| Claims priority of provisional application 60/379827, filed on May 10, 2002. | ||
| Int. Cl. G06F 13/10 (2006.01); G06F 17/50 (2006.01); G06F 11/30 (2006.01); G06F 9/44 (2006.01) | ||
| U.S. Cl. 703—21 [703/1; 703/13; 702/182; 702/186; 717/135] | 49 Claims |
| 1. A method for estimating performance of a system, based on a value for a parameter that is used to manage memory for the
system, comprising the steps of:
while processing a workload within the system using a first value for the parameter, generating data that indicates the workload,
wherein the data that indicates the workload includes data that identifies one or more performance characteristics for a plurality
of operations that executed, at least in part, during the workload; and
based on the data that indicates the workload, generating an estimated performance for the workload that the system would
have experienced if a second value for the parameter, different from the first value for the parameter, had been used while
processing said workload;
wherein the step of generating an estimated performance for the workload includes:
based at least on the performance characteristics identified in the data indicating the workload, generating individual estimated
performances for each operation in a plurality of concurrent operations that executed concurrently during the workload;
wherein each individual estimated performance estimates a performance that the system would have experienced if the second
value for the parameter had been used while processing said workload;
determining the estimated performance for the workload based on the individual estimated performances.
|