Method R Discussion GroupMethod R helps developers, DBAs, and decision-makers optimize Oracle-based software in every phase of the software life cycle. |
||
Discuss Method R products, techniques, and events. Ask new questions or post your own responses. |
We are using some modules of Oracle Applications. After future rollouts, we know that 50% more users will use a particular Module (say Inventory) and 200% more users will use another module (say Account Payables).
Now based on these two application users growth, can we predict total workload in terms of technical metrics like user calls/sec for example? It is required to forecast utilization (CPU or I/O subsystem). Regards, Neeraj
I've been presenting on exactly this topic for the past year and a half or so. Look for the presentation called "Preventing Performance Problems."
The main message is that while it's good to understand what queueing theory tells you, on real systems with heterogenous workload, it is sufficient to manage workload to a utilization statistic. You simply cannot add any workload that will send utilization to the right of its "knee value" without causing response time problems for everyone on the system, where "knee value" is defined as the utilization value at which response time divided by utilization is at its minimum. The key is that you have to know where that knee is for each resource on your system. This is one of the things that we discuss in the "Preventing Performance Problems" seminar. How to measure all the things I've mentioned here (like how much of a given resource does a given task execution consume) is the subject of virtually all our company's present software development effort. |
Powered by FogBugz