About a month ago the performance of several batch jobs turned south in a 
big way.  Nothing had changed on the system.  No new applications were 
installed.  No hardware changes. This is a partition on an 870.  4 
processors, 10 Gb memory.   Faulting is at an acceptable level for all 
pools.  Most of these jobs were I/O intensive jobs so I started looking at 
journaling and disk performance.  I looked at the receiver size threshold 
(7 Gb)  to make sure it was large enough to spread it across enough arms 
(100 arms).  I looked that percentage of fast writes to writes per the 
Journaling performance redbook to see if the system was over running the 
write cache and that isn't a problem (99.96 of the writes were fast 
writes).  However, one thing I noticed was that for a 24 hr period 18 of 
the disks had 50 to 93 million writes while the rest of the disk had only 
1- 2 million writes.   How can I find out what is generating all these 
writes to these particular disks?  I am not a performance expert by any 
means and could use any suggestions.  Thanks.

Mark Garton

This thread ...

Return to Archive home page | Return to MIDRANGE.COM home page