Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!decwrl!sgi!shinobu!odin!bh
From: bh@sgi.com (Bent Hagemark)
Newsgroups: comp.sys.sgi
Subject: Re: Do IRIS files systems fragment?
Message-ID: <1991Mar14.220120.13856@odin.corp.sgi.com>
Date: 14 Mar 91 22:01:20 GMT
References: <1991Mar13.211738.9885@abbott.com>
Sender: news@odin.corp.sgi.com (Net News)
Organization: Silicon Graphics, Inc., Mountain View, CA
Lines: 29

In article <1991Mar13.211738.9885@abbott.com> larry@abtlabs (Larry Pajakowski) writes:
>I've been running news on a PI for serveral months now.  Now when I run
>gr_osview while news is running the io wait for disk stays pegged at or
>near 100%.
>
>Questions:  How badly does the IRIS fs fragment?  Should I bother to
>save/mkfs/restore the thing?  Anybody know of a defragmenter?
>
>Many thanks.
>
>Larry Pajakowski
>Abbott Labs
>uunet!abtlabs!larry

First off, I'd let more direct measurement or even personal feel decide if
your system is not performing up to snuff.  gr_osview can give a nice
vivid picture of only part of the story.  What do you mean by "news
is running"?  New articles coming in?  Creating lots of little files is
generally harder on the system than the same amount of data in few large files.
Inodes are updated synchronously on close(2), etc...

In general the EFS file system does a nice job of keeping file system layout
optimal for both reading and writing.  I've really only seen poor layout
on very full file systems (>90%) which see a lot of activity.

The next major release of IRIX (4.0) includes an automatic file system
reorganizer which is designed to keep file blocks and free space contiguous.

Bent