Amanda-Users

Re: RE Tuning for performance

2006-08-24 10:51:35
Subject: Re: RE Tuning for performance
From: Joshua Baker-LePain <jlb17 AT duke DOT edu>
To: Cyrille Bollu <Cyrille.Bollu AT fedasil DOT be>
Date: Thu, 24 Aug 2006 10:44:35 -0400 (EDT)
On Thu, 24 Aug 2006 at 4:36pm, Cyrille Bollu wrote

root@srv-fhq-bkp bonnie++-1.03a]# ./bonnie++ -u 0
(snip)
Version  1.03       ------Sequential Output------ --Sequential Input-
--Random-
                   -Per Chr- --Block-- -Rewrite- -Per Chr- --Block--
--Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec
%CP
srv-fhq-bkp.fed 16G 36522  95 120980  60 58116  21 17956  49 56374  27
431.1   0

I see bad sequential input with getc but good (better than what I get with
Amanda) block sequential input.

The getc/putc perfomance is a measure of glibc, not your disks. The only numbers I'm really interested in are the block output (i.e. reads) and block input (i.e. writes) to the array. You can read from the array at
100MB/s, so that is *not* what is limiting your bandwidth to the tape
drive (unless the array is otherwise busy when you're trying to run backups). Have you tried increasing amanda's blocksize and/or testing with 'tar -b'?

--
Joshua Baker-LePain
Department of Biomedical Engineering
Duke University

<Prev in Thread] Current Thread [Next in Thread>