Re: Database Page Faults/sec
- From: "John Fullbright" <fjohn@donotspamenetappdotcom>
- Date: Wed, 25 Apr 2007 12:26:37 -0700
A database page fault occurs whenever a page is read that is not already in
database cache. It's really a reflection of read activity.
A database page fault stall occurs when, after a page fault, a page cannot
be read into cache because all of the pages in cache are locked. This
number should always be 0. A number larger than 0 would indicate a serious
bottleneck.
You'd be much better off looking at physical disk - average sec/write when
trying to determine if you have a disk bottleneck. If you average more than
20 ms or have spikes lasting more than a few seconds over 50 ms, the you
have a poorly performing disk subsystem.
BES essentially opens an uncached MAPI session for each BES user an polls.
To se high read activity for BES would not be out of the ordinary. I
believe it's the RIM whitepaper (referenced in optimizing storage) that
states a Blackberry user is 3.64 times the IO of a non blackberry user.
Are the storage subsystems on each of your servers appropriately sized to
handle the extra load?
"Mark Dolan" <MarkDolan@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:1C3031EA-733B-4CDC-A945-D92A4629B1E0@xxxxxxxxxxxxxxxx
Can someone tell me what an acceptable level for Databse Page Faults/sec
should be. I've got three servers and that perfmon counter is pegged out
on
two of them (averaging 1,914). The other server has an average of 215.
We're having some bad blackberry issues (delivery delays to the handheld)
on
the two with the high Database Page Faults. All other counters seem to be
OK, this is the only one we see that is high.
.
- Prev by Date: Re: SMTP Logging? -- Spam Problem
- Next by Date: Re: SystemMailbox{GUID Number}
- Previous by thread: Re: Database Page Faults/sec
- Next by thread: Re: sent mail from our email account to same account
- Index(es):
Relevant Pages
|
