Release 10.1A: OpenEdge Getting Started:
Database Essentials

Establishing a performance baseline

The primary reason to establish the baseline is to enable you to quantify changes in performance from changes in your load or application. The most difficult part of establishing a baseline is determining which operations are critical to the effective use of the system. You want the list of operations to be complete, but you also want to avoid having too many items on the list. Too many items will increase the amount of work needed to establish reliable baselines.

The basic rules outlined below should allow you to narrow down the number of tasks that need baselines to a manageable number. You might want to have two sets of baselines—one for daytime processing and one for end-of-day processing.

Include the following:

Tasks that are done many times throughout the day (such as creating orders, process control, and data entry tasks).
Tasks that are important to the user (such as Web site queries, customer support screens, and order entry tasks).

Do not include the following:

Periodic tasks (such as monthly and weekly reports).
Little-used portions of your application.
Reporting, as it generally falls into the above two categories. Also, you can schedule most reporting outside of your primary operating hours.

Collecting your baseline statistics

Once you have determined what items you want to benchmark, you can plan your strategy.

You can modify the application code to collect this data, which is the most accurate method, but it is also time consuming and costly. An easier way to perform data collection is to time the operations on a stopwatch. This is fairly accurate and easy to implement. To determine the best timing baseline for each task, perform timing in isolation while nothing is running on the system. When timing baselines have been established, repeat the task during hours of operation to establish your under-load baselines.

Understanding your results

Once your task times have been established, you must analyze the results.

As mentioned, it is best to establish the baselines while there are no reports of any problems on the system. This will establish what is normal on your system. If users are reporting problems, you can compare the current timings against your baselines to see if the problem is real or imagined. If there is a material difference in the current timing, you must start analyzing performance on the system with monitoring tools such as PROMON, VSTs, Fathom Management, and operating system utilities.