ADSM-L

Re: JFS vs logical volumes

1995-06-28 19:20:34
Subject: Re: JFS vs logical volumes
From: Paul Zarnowski <VKM AT CORNELLC.CIT.CORNELL DOT EDU>
Date: Wed, 28 Jun 1995 19:20:34 EDT
Thanks Greg.

I don't think we were recommended to use JFS because of bugs with the LVM
code, but perhaps that was in the back of the person's mind.  When we ran
with the LVM code, we did not experience any bugs that we could attribute
to LVM problems.  When setting up ADSM, I think working with LVMs is a bit
easier than with JFS, at least for large systems.  There are fewer layers
of disk concepts to deal with, and with JFS you waste a bit more space
(but not very much).

I am a bit surprised to hear that using JFS performs just as well as LVM.
Normally when double cacheing is done (on any type of system, not ADSM in
particular), the performance suffers somewhat, or at least the CPU overhead
increases.  In our case, we do seem to utilize the CPU on our RS/6000 100%
during nightly backups, and for other DB-intensive operations, so if double
cacheing does increase CPU usage then I would expect throughput to suffer.

On the other hand, if the studies that were done were on unsaturated ADSM
servers, then I can believe that no performance difference was detected
(if the CPU is not pegged, then it doesn't matter as much how much CPU
overhead there would be with double-cacheing).

Anyway, if anyone there does any further studies or hears of any, I (and
probably others) would appreciate hearing about them.

Thanks again.
..Paul

Paul Zarnowski                     Phone:   607/255-4757
Cornell Information Technologies   Fax:     607/255-6523
Cornell University                 US Mail: 315 CCC, Ithaca, NY 14853-2601
<Prev in Thread] Current Thread [Next in Thread>