There is a significant difference or shift in focus when working on batch processes. Our concern shifts from response time per unit of work to total response time. In other words, we are more concerned about how long it takes to process 500MB of IO or 5,000 seconds of CPU compared to the response time of a single physical IO. Another way of looking at this is our unit of work becomes the entire batch process or a step in the batch process.

While the response-time curve can be used when working with batch processes, because of the longer and singular process time focus, it is not nearly as useful. The response-time curve shines when it relates time to small units of work, like a logical IO or the execution of a SQL statement. Because our focus has shifted from small units of work to an entire process or process segment, our method of reflecting the situation must also change. Instead of using a response-time curve, the situation can be conveyed numerically in a table format (see Table 9-5) or by using a simple timeline.

Table 9-5 shows how to set up a batch process analysis. The entire batch process has been segmented into three steps, or segments. Step determination is based on your objectives, available statistics, and your audience. The time data comes from the same sources as with online transactions, but, as you'll see, with a slight yet significant twist. When the process steps have been defined and the respective data collected, a table similar to Table 9-5 can be constructed.