Re: Slow count(*) again... - Mailing list pgsql-performance

From Neil Whelchel
Subject Re: Slow count(*) again...
Date
Msg-id 201010131308.27760.neil.whelchel@gmail.com
Whole thread Raw
In response to Re: Slow count(*) again...  (Mladen Gogala <mladen.gogala@vmsinfo.com>)
List pgsql-performance
On Wednesday 13 October 2010 05:33:28 Mladen Gogala wrote:
>   On 10/13/2010 8:12 AM, Greg Smith wrote:
> > The work incorporating a more stable XFS into RHEL started with xfsprogs
> > 3.0.1-6 going into Fedora 11, and 3.1.X would represent a current
> > release.  So your Ubuntu kernel is two major improvement releases
> > behind, 3.0 and 3.1 were the upgrades to xfsprogs where things really
> > got going again making that code modern and solid.  Ubuntu Lucid
> > switched to 3.1.0, RHEL6 will probably ship 3.1.0 too.
>
> I am afraid that my management will not let me use anything that doesn't
> exist as a RPM package in the current Red Hat distribution. No Ubuntu,
> no Fedora, no manual linking.  There will always be that ominous
> question: how many other companies are using XFS? From the business
> perspective, questions like that make perfect sense.

XFS sees extensive use in the billing departments of many phone and utility
companies. Maybe not the code that you see in Linux, but the on-disk format,
which I think is unchanged since its original release. (You can use the modern
XFS code in Linux to mount a filesystem from an older SGI machine that used
XFS.) The code in Linux is based on the code that SGI released some time in
2000, which worked at that time very well for the SGI machine. At the time
that SGI came up with XFS, they had realtime in mind. They added specific
features to the filesystem to guarantee IO at a specific rate, this was
intended for database and other realtime applications. I have not looked at
the Linux version to see if it contains these extensions. I will be doing this
soon, however as my next big project will require a true realtime system.
-Neil-

pgsql-performance by date:

Previous
From: david@lang.hm
Date:
Subject: Re: Slow count(*) again...
Next
From: Neil Whelchel
Date:
Subject: Re: Slow count(*) again...