linux - server configuration for small database - Mailing list pgsql-performance

From Paolo Negri
Subject linux - server configuration for small database
Date
Msg-id b242f69c0703230548i13fcc967i6a7211ed450d945@mail.gmail.com
Whole thread Raw
Responses Re: linux - server configuration for small database  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-performance
Hi

I'm going to install a new server and I'm looking for some advice
about how I can help my database performing at its best. I'll be using
a redhat ES 4 on a dual core opteron with 2 SCSI 10.000rpm disks in
RAID1.

The first question I have is, since I'd strongly prefer to use
postgresql 8.1 or 8.2 instead of the 7.4 that comes with redhat, if
someone else is using with success third party rpms with success in
business critical applications.

Database size:  under 1 G
Load: reaches 100 r/w queries per second peak time but the average is lower
Filesystem: I'll probably have ext3 because of the redhat support.
RAM: I will have enough to let postgresql use a quantity about the
size of the db itself.

I'll have just one pair of disks so Linux will live on the same disk
array of the database, but obviously I can make any choice I like
about the partitioning and my idea was to put postgresql data on a
dedicated partition. As processes I'll have some other webby things
running on the server but the I/O impact at disk level of these should
be not that bad.

Now aside of the filesystem/partitioning choice, I have experience
with LVM, but I've never tried to take snapshots of postgresql
databases and I'd be interested in knowing how well this works/perform
and if there's any downside. I'd like even to read which procedures
are common to restore a db from the LVM snapshot.
As an alternative i was thinking about using the wal incremental
backup strategy.
In any case I'll take a pg_dump daily.

I would appreciate even just interesting web links (I've already
googled a bit so I'm quite aware of the most common stuff)

Thanks

Paolo

pgsql-performance by date:

Previous
From: Tino Wildenhain
Date:
Subject: Re: Performance of count(*)
Next
From: Michael Stone
Date:
Subject: Re: Performance of count(*)