pg.ddx.io  pgsql-admin@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Robin Iddon <robin@edesix.com>
To: pgsql-admin@postgresql.org
Subject: Re: New system recommendations
Date: Thu, 27 Apr 2006 08:58:17 +0100
Message-ID: <44507999.9080400@edesix.com> (raw)
In-Reply-To: <8511B4970E0D124898E973DF496F9B432515CF@stash.stackdump.local>
References: <8511B4970E0D124898E973DF496F9B432515CF@stash.stackdump.local>

Hi,
> It is currently running on a 3GHz Xeon HT, 2GB RAM, dual 72GB disks
> running RAID 1.  This server is a 1U without only 2 drive bays, so I
> have a potential issue with drive space.
>
> As a result, I will be moving the db server to a Dell 1650 with 3 146GB
> SCSI drives running RAID 0.  System is a dual processor, 1.2GHz, with
> 4GB RAM.
>
>  
>   

Do you mean RAID0?  You face total data loss when the first disk breaks 
- a 1.2GHz machine is not going to be young, so I assume neither are the 
disks - a recipe for disaster I think.

If your backup strategy is such that you can cope with total data loss 
and you are more interested in performance I believe you will get better 
performance by logical splitting of the database across multiple 
spindles rather than striping.  For example if you put the 
write-ahead-log onto a spindle by itself you will get better performance 
than running everything on one jumbo RAID0 drive.  Databases are usually 
sensitive to head latency rather than raw transfer rate.  Dedicating one 
drive to the log (which must be written for each and every 
insert/update) minimizes the latency on writing the log. 

Is it perhaps an option to take the 146GB drives out of the 1650 and 
putting them into the Xeon machine?  That might get you the best of both 
worlds (faster CPU, bigger hard disks, RAID1).

One final word or warning, whichever O/S you use HT doesn't always work 
well for postgres.  It's hard to pin down exactly what happens, but on a 
P4 HT system (2.6 kernel again, PG 8.0) we found that we had some very 
odd performance problems that were hard to reproduce in the lab but 
occurred in the live environment a lot.  Turning off HT made them go away.

Hope this helps,
Robin




view thread (15+ messages)  latest in thread

Message-ID: <44507999.9080400@edesix.com>
Permalink:  ../44507999.9080400@edesix.com/
Also on:    postgresql.org/message-id/44507999.9080400@edesix.com

 · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients from your mail client; the link fills in
  the headers below: reply via email
  Some clients, webmail especially, drop In-Reply-To and References from
  a link, and the reply then starts a new thread: use the mbox below.

  To: pgsql-admin@postgresql.org
  Cc: robin@edesix.com
  Subject: Re: New system recommendations
  In-Reply-To: <44507999.9080400@edesix.com>
  References: <8511B4970E0D124898E973DF496F9B432515CF@stash.stackdump.local> <44507999.9080400@edesix.com>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox