ADSM-L

Re: DB Cache Hit Rate Question

2003-02-20 11:56:17
Subject: Re: DB Cache Hit Rate Question
From: Farren Minns <fminns AT WILEY.CO DOT UK>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Wed, 19 Feb 2003 16:24:32 +0000
OK, all the info I have received has been of great help and indeed the
Technical Reference guide is very informative.

But what I still don't understand is this. I'm running expiration on the db
at the moment and the cache hit it average at 98.5%. Not 99%, but not too
bad. But the rest of the time, i.e. when running backups etc, it drops to
as low as 95%.

Should the cache hit rate consistently be above 98%? This is the answer I
cannot find.

Many thanks again to all who have provided help with this.

Farren Minns - John Wiley & Sons



Please respond to "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>

Sent by:        "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>

To:        ADSM-L AT VM.MARIST DOT EDU
cc:
Subject:        Re: DB Cache Hit Rate Question


Farren,

I'm not sure exactly how the TSM database internals work, but typically a
database cache is populated with "cache prefetch."  This isn't _exactly_
correct, but it'll do conceptually.  The db goes to read one row (check a
file for expiration).  That one db row request loads one or more
blocks/pages of db into the cache. Those blocks contain many other rows, so
most of the next few thousand file checks get cache hits in the prefetched
cached blocks.

Hope this helps.

Alex Paschal
Freightliner, LLC
(503) 745-6850 phone/vmail

-----Original Message-----
From: Farren Minns [mailto:fminns AT WILEY.CO DOT UK]
Sent: Tuesday, February 18, 2003 8:25 AM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: DB Cache Hit Rate Question


Good afternoon all TSMers.

I'm a little confused about the cache hit ratio information. If I run
expiration on a database, surely the TSM server has to trawl through the
entire thing on disk as it does not already reside in the buffer. So how
would I expect to see hit ratios of 99% or higher. I won't go into details
here as I'm more interested in the process of how the database and cache
work than out particular performance problems. But, as we do very little in
the way of restores, surely the amount of time that a requested database
file is present in the cache must be very low, so a cache hit ratio of
96-98% would make sense.

Please forgive my ignorance here as I'm not a expert in this field at all.

Many thanks in advance

All the best

Farren Minns

John Wiley & Sons Ltd




<Prev in Thread] Current Thread [Next in Thread>