Dynamic Platform Optimizer is a hypervisor function which rearranges logical partition processors and memory on the system to improve the affinity between processors, and memory of logical partitions. In this post I will describe when to use this function and how to prevent IBM i Lpar from crash. I run an IBMi LPAR where memory and processor resources constantly grow. For obvious reasons, it is impossible to power off the entire box and power on the Lpars in order which guarantees the best processor and memory affinity. On POWER7 with multiple processor units (CECs) and partitions with big number of processors or memory you may experience performance degradation caused by a fact , that a processor has to read a data stored in a ‘remote cash’ (cash in different processor drawer). To improve the affinity IBM brought Dynamic Platform Optimizer (DPO) in firmware 760 for MMD systems. If you would like to read more about it, check another site.
Unfortunately, as many new functions, this one also hasn’t been perfect from the very first day. One of my lpars crashed during optimization process. It was one of the IBM i system with an independent ASP online.
Therefore, whenever you submit DPO on IBMi system with an iASP, make sure that you have PTFs MF60599 for V7R1 or MF60602 for V7R2 applied. More information you can read in this APAR.
DPO is handled by following commands from the HMC:
Show the score for the entire server
lsmemopt -m SERVERNAME-o calcscore -r sys
Show the score for lpars
lsmemopt -m SERVERNAME -o calcscore -r lpar
optmem -m SERVERNAME -o start -t affinity
Check status of the process
lsmemopt -m SERVERNAME