Hi All,
Thanks for the input thus far.
From what we can tell, the 36-bay solution we've been offered has two
SAS backplanes, one with 12 bays, the other with 24. Quite how these
are connected to the RAID card isn't entirely clear to us yet, though
we have theories.
We understand that the bandwidth-per-disk to one of the backplanes
(paradoxically the 12-bay one, it seems) will be less than to the
other, though we're seeking confirmation of this.
As Brian suggests, we are likely to host the system disks on the
motherboard disk controller, rather than the RAID card. This frees up
2 bays and means we have to buy one less server (16 rather than 17)
for the ~same total storage capacity.
Cheers,
Mike.
On 4 June 2010 15:40, Alessandra Forti <[log in to unmask]> wrote:
>> Not have the system disk connected through the raid. Ie only use raid
>> cards for data parttions/filesystems.
>
> not on these units.
>
> [log in to unmask] wrote:
>>
>> Other things you might tweant to consider are:
>> Splitting into two separate raids.
>>
>>
>> -----Original Message-----
>> From: GRIDPP2: Deployment and support of SRM and local storage
>> management [mailto:[log in to unmask]] On Behalf Of
>> Alessandra Forti
>> Sent: 04 June 2010 12:12
>> To: [log in to unmask]
>> Subject: Re: 36-bay storage nodes
>>
>> Hi Mike,
>>
>> They are in use at CERN since January. Manchester just bought them but
>> we don't have them on site yet. The usable space is 30x2TB(3x2TB for the
>> raid) + 2 slots for the OS. Beyond that the raid card efficiency
>> degrades according to the vendor.
>>
>> cheers
>> alessandra
>>
>> Mike Kenyon wrote:
>>
>>>
>>> Hi All,
>>>
>>> As most of you probably know, we're going through the procurement process
>>> at Glasgow, and are looking to buy ~1TB of disk.
>>>
>>> One solution we've been offered uses a 36-bay server (yes, that'd be
>>> 36 x 2TB disks per server). Does anyone out there have experience of
>>> these beasts? We're trying to figure out if they've been successfully (or
>>> otherwise) deployed within GridPP...clearly, suppliers will only mention
>>> favourable case studies.
>>>
>>>
>>> Cheers,
>>> Mike
>>>
>
> --
> The most effective way to do it, is to do it. (Amelia Earhart)
> Northgrid Tier2 Technical Coordinator
> http://www.hep.manchester.ac.uk/computing/tier2
>
|