BackupPC-users

Re: [BackupPC-users] Tuning for disk contention

2009-09-04 23:12:10
Subject: Re: [BackupPC-users] Tuning for disk contention
From: higuita <higuita AT GMX DOT net>
To: backuppc-users AT lists.sourceforge DOT net
Date: Sat, 5 Sep 2009 04:08:22 +0100
Hi

On Fri, 4 Sep 2009 14:39:22 -0700, James Ward <jeward AT torzo DOT com> wrote:
> What's a good IO/WAIT average to shoot for?  Load average?  Other  
> things I might not be thinking of?

        the "perfect" max load should be the number of cpus you have,
        so a quad-core server can sustain a load of 4 without any
        problem... after that number, the higher the load, the higher
        will be the performance lost

        a high load and still having cpu idle is a sign of IO load.
        high system cpu usage is a sign of the kernel spending too 
        much time managing the IO

        for IO load can be 99/100%, but only as long the IO wait 
        doesnt increase alot... this values depends of the HD type,
        raid layout, number of disks, bus, etc

        in modern HDs (sata, scsi, sas) IO waits between 5-20 are
        normal... 100 is high load...if you see times above 150, 
        that HD is probably getting too much load and is slowing 
        down a lot the rest of the system (but again, this values 
        depend of your OS/config/FS/HD layout)

        dont forget that writes usually put more load on a HD than
        reads, the HDs cache helps a lot, but usually write cache is
        not enabled

        use iostat (like iostat -kx 30 ) to monitor the HD loads 
        and see how the io load and wait is going... sporadic 
        high loads/times are normal, specially in small check times,
        but sustained high loads/times are a sign of problems

        if you want to try to get a little more from your HDs and
        the data isnt critical in case of a powerloss (usually 
        backups arent, next backup would "fix" the bad data), you
        can enable the write cache.

        be aware that at least xfs doesnt like to lose data with
        powerloss+write cache, its format assumes that data hits
        the HD when it tells and can miss behave if the cache
        just decided to only really write half of the data before 
        the powerloss. that is why is highly recommended to use xfs
        with a UPS and do a controlled shutdown in case of powerloss

good luck
higuita
-- 
Naturally the common people don't want war... but after all it is the
leaders of a country who determine the policy, and it is always a 
simple matter to drag the people along, whether it is a democracy, or
a fascist dictatorship, or a parliament, or a communist dictatorship.
Voice or no voice, the people can always be brought to the bidding of
the leaders. That is easy. All you have to do is tell them they are 
being attacked, and denounce the pacifists for lack of patriotism and
exposing the country to danger.  It works the same in every country.
           -- Hermann Goering, Nazi and war criminal, 1883-1946

Attachment: signature.asc
Description: PGP signature

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
trial. Simplify your report design, integration and deployment - and focus on 
what you do best, core application coding. Discover what's new with 
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
BackupPC-users mailing list
BackupPC-users AT lists.sourceforge DOT net
List:    https://lists.sourceforge.net/lists/listinfo/backuppc-users
Wiki:    http://backuppc.wiki.sourceforge.net
Project: http://backuppc.sourceforge.net/