Brian Chivers wrote:
> Simon Capstick wrote:
>> Brian Chivers wrote:
>>> Simon Capstick wrote:
>>>> Brian Chivers wrote:
>>>>> I have setup a little test network with two machines connected together with iSCSI on gigabit
>>>>> ethernet. I'd like to test the filesystem performance, the iSCSI target is an OpenFiler box & the
>>>>> client / initiator is a Centos 5.3 machine with the iSCSI mounted as ext3.
>>>>>
>>>>> Has anyone got any recommendation for testing software or has anyone played around with this sort of
>>>>> setup?
>>>>>
>>>>> I've looked at Iozone and it looks very good but a little too complicated for me at the moment at
>>>>> this early stage.
>>>>>
>>>>> Thanks
>>>>> Brian
>>>>>
>>>> Bonnie++ ?
>>>>
>>>> Simon
>>>>
>>> Thanks for the pointer.
>>>
>>> I've run the tests just got to try & understand the results now :-)
>>>
>>> Have you ever run bonnie ? any pointers as to how to interpret the results ??
>>>
>>> Brian
>>>
>>> ------------------------------------------------------------------------------------------------
>>> The views expressed here are my own and not necessarily
>>>
>>> the views of Portsmouth College
>>>
>> It's been a while, but I would start by benchmarking several storage
>> devices including a bog standard single drive. Then put the data into
>> a spreadsheet or table so you can actually read the data, so you have
>> something like this...
>>
>> http://www.coker.com.au/bonnie++/res0.90.html
>>
>> Simon
>>
>>
>
> I used the bon_csv2html2 script to generate some tables and combined them into the page below. The
> first two tables are from the same machine & the second is one of our main servers which has a PERC4
> SCSI card but the trouble is I don't know what I'm looking at, is this good bad or indifferent :-)
>
> http://ww2.portsmouth-college.ac.uk/brian/bonnie.html
>
> Brian
>
Hi Brian,
Take a look at the random seeks per second. This will mainly determine
file server (small files) and database performance. If you're looking
at recording video then the sequential create figures are more relevant.
You'll generally find that the more drives you have the more the
random seeks per second you'll have. Different RAID levels will make a
difference here too.
Both iSCSI ext3 partition/iscsi-test.portsmouth-college.ac.uk and
ilpdb09.portsmouth-college.ac.uk/Dell local ext3 SCSI drive have similar
performance, over double that of the standalone drive.
You'll need to give the list some idea of the services being run to
decide which figures are most relavent to you.
Simon
