Re: SATA drives performance - Mailing list pgsql-performance

From Greg Smith
Subject Re: SATA drives performance
Date
Msg-id 4B339F66.7020808@2ndquadrant.com
Whole thread Raw
In response to Re: SATA drives performance  (Mark Mielke <mark@mark.mielke.cc>)
List pgsql-performance
Mark Mielke wrote:
> Can you be more specific? I am using ext4 without problems than I have
> discerned - but mostly for smaller databases (~10 databases, one
> almost about 1 Gbyte, most under 500 Mbytes).

Every time I do hear about ext4, so far it's always in the context of
something that doesn't work well--not hearing about improvements yet.
For example, there was a thread on this list earlier this month titled
"8.4.1 ubuntu karmic slow createdb" that had a number of people chime
saying they weren't happy with ext4 for various reasons.

Also, I have zero faith in the ability of the Linux kernel development
process to produce stable code anymore, they're just messing with too
many things every single day.  Any major new features that come out of
there I assume need a year or two to stabilize before I'll put a
production server on them and feel safe, because that this point a
"stable release" means nothing in terms of kernel QA.  Something major
like a filesystem introduction would be closer to the two year estimate
side.  We're not even remotely close to stable yet with ext4 when stuff
like http://bugzilla.kernel.org/show_bug.cgi?id=14354 is still going
on.  My rough estimate is that ext4 becomes usable and free of major
bugs in late 2010, best case.  At this point anyone who deploys it is
still playing with fire.

> File systems failures have been pretty rare for me lately, so it's
> hard to say for sure whether my setup is really running well until it
> does fail one day and I find out.

All of the ext4 issues I've heard of that worry me are either a)
performance related and due to the barrier code not doing what was
expected, or b) crash related.  No number of anecdotal "it works for me"
reports can make up for those classes of issue because you will only see
both under very specific circumstances.  I'm glad you have a good backup
plan though.

--
Greg Smith    2ndQuadrant   Baltimore, MD
PostgreSQL Training, Services and Support
greg@2ndQuadrant.com  www.2ndQuadrant.com


pgsql-performance by date:

Previous
From: Mark Mielke
Date:
Subject: Re: SATA drives performance
Next
From: Richard Neill
Date:
Subject: Re: SATA drives performance