Originally posted by: bob
I have two AMS2300 dual controller boxes with 16gb cache, 8 FC ports and both of them suffer from high cpu load
and high latency on all ports for luns in all pools.
CPU util is at 60% average with long peaks at 70-95%. Ctrl 0 and 1 are about the same cpu util.
We run 3 HDP pools:
000 3 x 8+1 15krpm RGs, PVOL pool.
010 3 x 8+1 15krpm RGs, SVOL pool mixed with some testing VMs.
020 SATA 2TB 8+2 x 3
021 SATA 1TB 12+2 x 2.
We got some consultants in to diag the systems but their results were inconclusive. They presented us with some graphs
showing most metrics where ok. The load for all SAS pools where low with some peaks for the SATA drives. They claim
the occasional SATA peaks of 1500-2000 IOPS stuffs the cache and slows the entire system. Write pending remains under 20%
when it's at it's highest. Each box push about 150-200MBps in average over 24 hours.
We also have about 150 shadowimage pairs per box covering about 2TB, could they cause high cpu utilization? No truecopy, no cache partitions
or such. SI pairs are used for backups of low load oracle databases. They are synced during office hours and
split at night for backup to tape.
Sadly we don't have the budget for this client to buy tuning manager. Are we on the right track beleving the high
cpu util is causing the system wide latency problem?