[Bioclusters] mid range storage solutions

[Joe Landman]

Hi Bonnie:

Bonnie Hurwitz wrote:
> Hi all,
> 
> I was just wondering if anyone has a recommendation for mid-range
> storage.  We need to purchase a storage server for our cluster to act as
> a mySQL database server that has around 32 terabytes of disk space
> utilizing Raid10.  Also, we are looking for fast disks and a 10-20Gb
> card since this is meant for a database server and we want to try to
> minimize resource contention from writing to the db server from the
> nodes.  We currently have 500 nodes on our cluster.

First off, we build and sell these sorts of things in addition to other
related components.

I presume you mean that you need 32 TB of usable RAID10 space per 7U or
less.

> What are people currently using for similar database servers?  What has
> the performance been like when writing to the databases from compute nodes?
> 
> What other options are out there?
> 
> Also, we need to minimize rack space.  We currently have about 7U of
> space we can use at the moment.

Ok, this is not meant to be a commercial, email/contact me offline for
more information.

Our JackRabbit JR48 unit has 48x 1TB drives, which provides 9.6 TB/U of
raw capacity.  You can use them as any sort of RAID you wish
(1,5,6,10,50,60,...).

Some sizing calculations:  With RAID 10 you would be limited to (at
maximum with no hot spares) 24TB in this platform.  You would need 2
units to get 32 TB of usable RAID10 space.  With RAID 60 you would be
limited to (at maximum with no hot spares) 44TB of usable space.  Adding
2 hot spares per controller (4 hot spares total), you would have 40TB
usable space.

These disks are 7200 RPM SATA II NCQ's, and you can see an older
benchmark report on our site (in the .sig below).

As for size, each unit is 5U high.  Your requirements are 4.57 TB/U, as
far as I understand them.

10 GbE/IB (SDR/DDR) is not an issue for these units, the 48TB units come
with an IB card by default, which could be swapped out for a 10 GbE.

I would like to note that if you have 500 nodes reading and writing at
once, the network will be only one of your points of resource contention.

> -Bonnie

Feel free to contact me offline.  We have a smaller unit we have run
some tests for customers with.

Joe

-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web  : http://www.scalableinformatics.com
       http://jackrabbit.scalableinformatics.com
phone: +1 734 786 8423
fax  : +1 866 888 3112
cell : +1 734 612 4615