Hello,
When inserting/updating large amounts of data to TM1 cubes in parallel threads, the processing of data can be done in parallel threads (processes), but when it comes to committing data to cube, it does not go in parallel and it works in series, one by one, for multiple threads. This limits the total execution time that can be shortened via multithreaded processes. Whatever the parallelization we implement for processing the large data in a large server having many CPU cores, you can not parallelize the commit operation. This means we can never achieve 100% CPU utilization with parallel execution of TM1 process for the duration of the whole main process, assuming we have many CPU cores.
Maybe IBMers can consider this as an improvement point for the tool. As we deal with larger amounts of data with TM1 cubes, this becomes more and more important limitation for the performance of the tool.
Kind regards,
Mucahit
------------------------------
Mucahit Erdal
------------------------------
#PlanningAnalyticswithWatson