Storage Area Network as a tlug meeting topic

[Mark Lane]

Christopher Browne wrote:
> On Wed, Apr 29, 2009 at 6:29 PM, William Muriithi
> <william.muriithi-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org> wrote:
>   
>>  I am wondering if anyone among us have been lucky to work extensively
>> with SAN and would be willing to share the knowledge. I think we could
>> all gain by exchanging experience on the subject, for example on:
>>
>> - Good practical reference we have come across -  be either books or web links.
>> -  Share experience on hardware performance, especially when using
>> multipathd and Linux in general. Yeah, I could get that information
>> from the white page, but I will be frank and say I do not trust
>> marketing department.
>> -  Any interesting way of gaining experience on the subject without
>> parting with thousands of dollars.
>>     
>
> Sounds interesting...
>
> The one challenge I'd foresee is that there are two perspectives on
> this that are quite different in their slants:
>
> a) Those that are doing "hard-core" SAN work, where maximizing numbers
> of spindles is worth paying extra money, and where multichannel
> FibreChannel is worthwhile.
>
> I saw the following recently that explains that (though doesn't
> particulary explain *why*) expensive disks can give 10x better
> performance than the cheaper disks.
>
> http://www.dbms2.com/2009/04/28/data-warehouse-storage-options-cheap-expensive-or-solid-state-disk-drives/
>   
The article does not give support for the claim by Oliver Ratzesberger 
that 15K drives are 10x faster than 7200RPM drives. Maybe they are but 
it doesn't really matter when dealing with a lot of drives. You only get 
an advantage of a faster disk when you are running single or small 
arrays of disks. If you have a large array, the bandwidth of your bus or 
storage network will be the limiting factor and not the drives 
themselves. When I was designing NAS boxes, we found that it would only 
take 4 7200RPM SATA 1 drives to max out a PCI-X bus using RAID 5 and 
3ware card. So if I had 4 15RPM drives you might get better seek times 
but the overall transfer rate will not be any better than the 4 7200RPM 
drives. I guess if you have an application that requires seeking a lot 
of small files and not a lot of bandwidth, the 15K RPM will still be 
faster but for most applications it won't be.

Also the article talked about cheap drives being optimized for power 
consumption and not performance and reliability. While this is true for 
desktop drives, the article fails to mention that for not a lot of more 
money you can get RAID edition SATA drives from WD and Seagate that are 
actually rated for performance and reliability.

Ultimately, the decision will come down to the executives. They assume 
that SCSI, Fiber Channel are enterprise class hardware and SATA isn't. I 
was at a Western Digital presentation once where they were showing of 
drives that were equivalent in performance and reliability to SCSI 
offering yet the still refrained from referring to them as enterprise 
class. Yet Google has no problem using commodity hard drives.

If you need a ton of storage, cheaper SATA drives are definitely much 
more cost effective. You can even get them with 5 year warranties.

Fiber Channel based SANs are not even the fastest storage networks out 
there. They are just very well established in the enterprise. Running 
iSCSI over 10Gig Ethernet or Myrinet will give you a much fastest 
Storage Network.



--
Mark Lane
lmlane-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org




--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists