The use of scsi raid over a new ide drive may not be as much as you think in
some situations. In situations where the writes will always be at the tail
of the file like WAL ide may perform really well if this is the only writes
that are occuring on the disk. The heads will alway be in the right place
for the write so latency is just going to be spin time not positioning. This
may be compared to all the writes occuring a big fast raid system where all
the system writes are occuring in different positions on the same disk set.
Seek time increases due to the multiplicity of writes occuring in different
area. Even scsi smarts with out of order writes can only make this process
optimal. Dedicating physical disk sets can eliminate some of the problems.
--
Ian Willis
-----Original Message-----
From: Alex Howansky [mailto:alex@wankwood.com]
Sent: Saturday, 17 March 2001 8:53 AM
To: pgsql-general@postgresql.org
Subject: [GENERAL] Re: Re: Fast Inserts and Hardware Questions
> [Interesting stats]
>
> > Wow, does WAL make so much of a difference as to make my
> > two-generations-old, built-from-spare-parts scrap box faster than my
> > latest and greatest production server? Yikes!
>
> What RAID level are you running? RAID-5 isn't always very
> fast on writes. That and WAL might be what's up.
Yes, it is RAID-5 on the big box. Unfortunately, I don't have any spare RAID
equipped boxes sitting around, so I can't experiment with the different RAID
levels. Still, you'd think that even a "slow" RAID-5 configuration would be
faster than a $98 IDE drive...
> Have you tried 7.1b5 on the big box?
Not yet, I'm waiting for a quiet weekend. I'll post my stats when (if...) I
get
around to doing that.
--
Alex Howansky
Wankwood Associates
http://www.wankwood.com/
---------------------------(end of broadcast)---------------------------
TIP 6: Have you searched our list archives?
http://www.postgresql.org/search.mpl
