RE: Journal performance problems

[ouuch@xxxxxxxxxxx]

On 1 May 2003 at 8:42, Andy Nolen-Parkhouse wrote:

>> As far as verifying whether your single-drive ASP is the problem, it
> sounds like you are doing the right thing by turning on monitoring.  A
> drive tends to max out its throughput when it reaches about 40% busy,
> after that you need more arms.  If this drive is at or above 40% during
> these batch jobs, then you can be reasonably sure that it is performing
> at its maximum and that you would benefit from adding another drive.  You
> can get a good idea of this from the WRKDSKSTS display if you can take a
> five-minute sample while your batch job(s) are running.  

Well, during some heavy-duty batchjobs, ASP2 disk goes up to 70-80% for some
time. Could it be better to RAID the drives in ASP2? I currently have 2 drives 
to be
built in on saturday and am trying to decide if they go to ASP2 or ASP1.
Would 4 drives in a RAID-config be better than 2 drives mirrored?

I have put back the two busiest journal receivers into ASP1. So I hope I see 
some 
performance differences during night processing.
> 
> I truly don't think that a larger write-cache will be much benefit for
> batch jobs.  If these jobs are pumping out write transactions for an hour
> and a half, you will just fill up the cache quickly and then be right
> back where you started.  The write cache will do an excellent job of
> evening out transient spikes, but won't do much for continuous overload.
> 
At busy times, one of the journals uses 3-4 journal receivers inside 60 
minutes, each
receiver at 1GB size. 

IBM has a licensed programm - Journal cache. Could this be helpful for us?

Regards,

Oliver