Bacula-users

Re: [Bacula-users] minimize iowait

2008-12-11 09:47:48
Subject: Re: [Bacula-users] minimize iowait
From: "Lukasz Szybalski" <szybalski AT gmail DOT com>
To: "Pasi Kärkkäinen" <pasik AT iki DOT fi>
Date: Thu, 11 Dec 2008 08:45:21 -0600
On Thu, Dec 11, 2008 at 3:51 AM, Pasi Kärkkäinen <pasik AT iki DOT fi> wrote:
> On Wed, Dec 10, 2008 at 04:38:58PM -0600, Lukasz Szybalski wrote:
>> On Wed, Dec 10, 2008 at 4:06 PM, John Drescher <drescherjm AT gmail DOT com> 
>> wrote:
>> > On Wed, Dec 10, 2008 at 4:10 PM, Lukasz Szybalski <szybalski AT gmail DOT 
>> > com> wrote:
>> >> Hello,
>> >> My bacula server is pretty busy, and I notice that at times the IOwait
>> >> reaches 40%.
>> >>
>> >> Currently I use: top, free, iostat -k 5 /dev/md5
>> >>
>> >> Are there any tools/commands on linux that can tell me what is the
>> >> status of Input output. I would like to know:
>> >> - How much IO are my hard drives capable of?
>> >> - What is an average? max? min?
>> >> - Is IO a bottleneck on my performance?
>> >> - Monitoring tool names and exact commands would be appriciated.
>> >>
>> >> Sorry to bring these questions here but I don't know any other list
>> >> that would have the technical people that have experience and are
>> >> knowledgeable on the topic.
>> >>
>> >
>> > Is this a software raid 5 or 6?
>>
>> Software raid5 using mdadm
>>
>
> What kind of drives? how many of them?
>
> Single SATA 7200 rpm drive is capable of around 100-150 random IOs per 
> second..
>
> You can check http://www.storagereview.com and go to "Performance Database"
> and select for example "IOMeter File Server - 128 I/O" and click "Sort" to
> see some results..
>
> SSD drives seem to beat the crap out of SAS and especially SATA drives:)
>
> 15k rpm U320 SCSI drives seem to be over 400 IOs per second..
> a lot faster than SATA disks.
>
> Anyway, the point was that even when a single SATA drive might have a nice
> sequential throughput, it won't do very well with random IOs.. If you're
> running multiple backup jobs at the same time you're pretty much having
> random IO patterns..
>
> Try benchmarking your md-raid-array and see how much IOs you can get out
> from it with random IO patterns? Try for example with LTP disktest.
>
> With LTP disktest you can try different IO sizes, different amount of threads,
> different read/write ratios etc..


So one way is to benchmark disks (4x 500GB WD raid5 ) . Is LTP
disktest the name of the software? Is it part of linux already?
How can I measure or benchmark the tps (transfers per second?) does that matter?
How can I measure or benchmark the IO from the system vs application ?
does that matter?

Thanks,
Lucas
------------------------------------------------------------------------------
SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada.
The future of the web can't happen without you.  Join us at MIX09 to help
pave the way to the Next Web now. Learn more and register at
http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users