Planning area datasource and parallel processing

[Former Member]

I'm using a parallel processing profile on the datasource of a planning area. However, I don't see any performance gains because of it. In fact, it runs a little longer than without parallel processing.

Also, I noticed, there is only one background job that does the reading from the planning area with the parallel profile. How is the parallel profile supposed to help?

[srinivas_krishnamoorthy · Accepted solution]

the performance of the parallel profile depends on two other important things like block size and memory resources available. Since parallel processing is internal to the process, I dont think you will see multiple jobs active in SM37. the way parallel profile is supposed to help is that one process can do the read, while the other can do the write as required based on block sizes.

Think this figuratively like two/multiple workers laying bricks on a foundation. Assume that only one worker can lay brick at a time. So while one worker lays the brick on the foundation, the other can carry the bricks from the adjacent pile and be ready to lay as soon as the other one is done. The block size is the number of bricks that can be carried/laid by the worker. There is an optimum combination that will bring the desired results. the speed of building is dependent on the speed of carrying the bricks from pile (reading) and speed of laying the the bricks (writing). If the relative times are not comparable(writing takes>>reading time), then parallel profile does not help too much and can deteriorate the performance if block size is low. I hope you get the drift.