As mentioned already, "its all about the TLB". With memory intensive tasks using large pages, it lessens the usage of TLB from what would otherwise be the case, and that potentially benefits (by DAT avoidance) not only the task using large pages, but also every other workload utilising that TLB (which i presume implies it impacts all LPARs sharing a TLB?). Clearly HIS data is ideal for gaining access to the necessary stats, although I have not had the luxury of seeing those stats in our environment. We are only moving to large pages very gradually, and I suspect the best opportunity to see a measurable step-change will tend to be with a big-bang switch to large page use. By implication, the higher the TLB missses are to begin with, the greater the chance of seeing some benefit should be (assuming that the tasks converted to use large pages are actually using memory intensively, so to be notably contributing to the original TLB miss situation).
On 2 May 2013 16:59, Mike Schwab <[email protected]> wrote: > I think one person found a lot of OpenMVS programs being repeatedly > loaded. So they wrote a little program that LOADed the object modules > into the LF area and ended. They now remain in core until shutdown. > > > -- > Mike A Schwab, Springfield IL USA > Where do Forest Rangers go to get away from it all? > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to [email protected] with the message: INFO IBM-MAIN > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
