Re: Disk I/O changes
From: Jonathan Lewis <Jonathan_at_jlcomp.demon.co.uk>
Date: 1995/11/22
Message-ID: <817076951snz_at_jlcomp.demon.co.uk>#1/1
Date: 1995/11/22
Message-ID: <817076951snz_at_jlcomp.demon.co.uk>#1/1
In article <dsmith.68.000AC4D1_at_gulfaero.com>
dsmith_at_gulfaero.com "Doug Smith" writes:
: Be sure you also have SOME disk cache (ie. 32-64
: meg.).
:
Large disk caches have been quoted before in this newsgroup as being a necessity with RAID5 configurations. But I have a question about this...
A write request from Oracle has to be a guaranteed write to disc, and on typical Unix systems is done with some sort of 'write through cache' or 'wait for sync' call.
How does this work if you are trying to use a cache on a RAID5 device to improve write times ? Aren't you leaving a window of failure open if your RAID device reports a write when the data is still in cache ?
-- Jonathan LewisReceived on Wed Nov 22 1995 - 00:00:00 CET