Unix managers, heres a tough performance issue we are having.
We upgraded to 8.1.6.3.1 from 81620 and all of a sudden, our on-line hot
backups take twice as long. It is the compress the datafile to backup
mount point that now takes twice as long as before. With Oracle shut down,
the speed returns to normal.
Before the upgrade to 81631, the compress would grab 90% of the cpu, after
the upgrade 50% of the CPU remains unused, and the compress is only able to
grab 48% of the CPU.
We have extensive charts and graphs to document this which also show us that
the issue is not disk saturation on the source or target, nor is it network
saturation.
Anyone?
Quick technical details:
-- * Oracle down it takes 5 Minutes
/oracle/Q01/sapdata6/pooli_1: date;time compress <pooli.data1 >
/mnt/sapbackup/pooli.data1.Z;date
Mon Aug 6 21:10:30 EDT 2001
real 5m2.50s
user 4m39.01s
sys 0m17.13s
Mon Aug 6 21:15:33 EDT 2001
-- * Oracle running it takes 11 Minutes
/oracle/Q01/sapdata6/pooli_1: date;time compress <pooli.data1 >
/mnt/sapbackup/pooli.data1_4.Z;date
Mon Aug 6 21:53:10 EDT 2001
real 11m28.41s
user 4m44.20s
sys 0m26.53s
Mon Aug 6 22:04:38 EDT 2001
Received on Wed Aug 08 2001 - 22:30:19 NZST