Skip site navigation (1) Skip section navigation (2)

Re: Fwd: large cache

From: Scott Marlowe <scott(dot)marlowe(at)gmail(dot)com>
To: Brian Modra <epailty(at)googlemail(dot)com>
Cc: pgsql-admin(at)postgresql(dot)org
Subject: Re: Fwd: large cache
Date: 2009-06-30 05:50:58
Message-ID: dcc563d10906292250v6787acc1s49f16963cfe8c4b8@mail.gmail.com (view raw or flat)
Thread:
Lists: pgsql-admin
On Mon, Jun 29, 2009 at 11:19 PM, Brian Modra<epailty(at)googlemail(dot)com> wrote:
> Hi,
> my database requires 1TB of disk space, but I want to avoid RAID (e.g.
> 8x 300GB SAS drives in RAID 10), and use rather a 1TB SATA disk.
> To avoid the "slowness" of SATA showing in queries, I want to cache
> lots of the database, but I don't want to buy more than 16GB RAM.
> (Budget limits).
> So I'm considering configuring a large cache, requiring larger than
> 16GB, and allowing it to "spill over" into swap space on a single
> 300GB SAS drive.

No, spilling to swap will make it slower than just hitting the drives.

If you're on a budget, look into using 4 1TB to 1.5TB SATA drives
(they're pretty cheap) with linux software RAID.  You can get pretty
good read performance from a RAID-10 of 4 drives, better parallel
performance.  Also, I was pricing RAM today for a server, and the
difference between 16 and 32Gig ($350 or so) was so little I didn't
even consider going with 16G.  Still, 16 or 32, it's only a tiny
portion of the 1TB you're talking about storing and accessing.  If
it's gonna be read-mostly, then leaving out the RAID controller can
save you a lot of money.

You can buy a 4 x 1TB drive machine with quad core opteron and 16Gig
for about $2500, $2850 if you want 32Gig.  That's a good deal for a
machine with redundancy on the drives and good performance.

In response to

pgsql-admin by date

Next:From: Rajdeep DasDate: 2009-06-30 07:13:08
Subject: Timestamp Datatype Changing automatically to Timestamp without time zone
Previous:From: Brian ModraDate: 2009-06-30 05:19:13
Subject: Fwd: large cache

Privacy Policy | About PostgreSQL
Copyright © 1996-2014 The PostgreSQL Global Development Group