zaki писал(а):Dimerson писал(а):Описано буржуями 100 лет как
не нашел где было описано...
Если здесь то поиском ReadAhead ищутся 2 thread'a которые содержат это. На буржуях это например письмо в форуме novell.support.open-enterprise-server.linux.storage-and-backup
==== кут хере =========
<KAnderson@REMOVEbcwgroup.com> wrote in message
news:a_Myf.33$Pa2.26@prv-forum2.provo.novell.com...
> (Second attempt at posting this - apologies if it appears multiple times)
>
> I've built a new HP BL20p as an OES (SP2) server. I've created two logical
> drives, one for a Reiser
> boot partition and the other for NSS.
>
> The NSS volume works fine for the GroupWise 7 post office on it, but
> general file sharing
> performance is truely awful.
>
> For example, I've got a 1.2Gb tarball. Creating a copy of it from the
> shell on the Reiser volume takes
> around 30 seconds. On the NSS volume, the same operation takes over *6
> minutes*!
>
> I know not to expect the same performance from NSS under Linux as I'd get
> on NetWare ... but 12
> times slower? How can I tune NSS or troubleshoot what's going wrong?
I've just started looking and OES Linux and I read somewhere that you will
probably want to adjust the
"Read Ahead Count in Blocks" value on the volume.
I believe it is set at a default of 2 (4k blocks)
and the other guys suggested uprating this to 128.
The setting goes up to 1024
The secret is balance according to the style of data you have on the volume.
GroupWise might even perform better with a slightly larger figure but not
much more.
Hope this helps
Karim C Kronfli
CNE 6
==== кут хере =========
от начала февраля сего года.