We should to a better job aggregating data from the kickstart records. A common question is regarding average cpu utilization (utime+stime / duration from ks record?) and average and max memory usage. I think we should do this across the workflow, but also per transform.