From: | Bill Moran <wmoran(at)collaborativefusion(dot)com> |
---|---|
To: | Ow Mun Heng <Ow(dot)Mun(dot)Heng(at)wdc(dot)com> |
Cc: | justin <justin(at)emproshunts(dot)com>, Scott Marlowe <scott(dot)marlowe(at)gmail(dot)com>, Amber <guxiaobo1982(at)hotmail(dot)com>, pgsql-general(at)postgresql(dot)org |
Subject: | Re: What's size of your PostgreSQL Database? |
Date: | 2008-08-19 11:34:19 |
Message-ID: | 20080819073419.902ed832.wmoran@collaborativefusion.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
In response to Ow Mun Heng <Ow(dot)Mun(dot)Heng(at)wdc(dot)com>:
> On Mon, 2008-08-18 at 11:01 -0400, justin wrote:
> > Ow Mun Heng wrote:
> > > -----Original Message-----
> > > From: Scott Marlowe <scott(dot)marlowe(at)gmail(dot)com>
> > >
> > > > If you're looking at read only / read
> > > > mostly, then RAID5 or 6 might be a better choice than RAID-10. But
> > > > RAID 10 is my default choice unless testing shows RAID-5/6 can beat
> > > > it.
> > > >
> > >
> > > I'm loading my slave server with RAID-0 based on 3 IDE 7200 Drives.
> > > Is this worst off than a RAID 5 implementation?
> > >
> > I see no problem using Raid-0 on a purely read only database where
> > there is a copy of the data somewhere else. RAID 0 gives performance.
> > If one of the 3 drives dies it takes the server down and lost of data
> > will happen. The idea behind RAID 1/5/6/10 is if a drive does fail
> > the system can keep going. Giving you time to shut down and replace
> > the bad disk or if you have hot swappable just pull and replace.
>
> I'm looking for purely read-only performance and since I didn't have the
> bandwidth to do extensive testing, I didn't know whether a RAID1 or a
> Raid 0 will do the better job. In the end, I decided to go with RAID 0
> and now, I'm thinking if RAID1 will do a better job.
When talking about pure read performance, the basic rule is the more
spindles you can have active simultaneously, the better. By that rule,
RAID 0 is the best, but you have to balance that with reliability. If
you have 10 disks in a RAID 0, the chance of the entire system going
down because of a disk failure is 10x that of a single disk system --
is that acceptable?
In theory, you can have so many disks that the bottleneck moves to some
other location, such as the IO bus or memory or the CPU, but I've never
heard of that happening to anyone. Also, you want to get fast, high-
quality disks, as 10 15,000 RPM disks are going to perform better than
10 7,200 RPM disks.
Another solution is RAM, if you can get enough RAM in the system to hold
your working set of data, then the speed of the disk is not really
relevant. Of course, that's tough to do if you've got 3TB of data,
which I don't know if that's your case or not.
--
Bill Moran
Collaborative Fusion Inc.
http://people.collaborativefusion.com/~wmoran/
wmoran(at)collaborativefusion(dot)com
Phone: 412-422-3463x4023
From | Date | Subject | |
---|---|---|---|
Next Message | Bill Moran | 2008-08-19 11:43:24 | Re: Re: can't get UPDATE ... RETURNING ... INTO ... to compile successfully |
Previous Message | Pavel Stehule | 2008-08-19 11:01:30 | Re: CASE |