They can have multiple day downtime with no problem. They do no 
clustering. Academia is quite often this way..

If you wanted to build this more reliable, you could go sas with 
multipathing or even run multiple sata controllers and do manual 
import.. I usually recommend 2 head nodes and 2 paths to each storage 
chain.. But this still could be done under probably 150k.

On 12/2/14, 9:18 AM, Carl Wilhelm Soderstrom wrote:
> On 12/02 08:55 , Linda Kateley wrote:
>> I recently wrote a blog about a local petabyte system. They use the WD red
>> nas drives.
>>
>> http://kateleyco.com/?p=815
> I'm impressed.
> Can you describe more how this works? "They can have scheduled and
> unscheduled downtime without complaints."
> What kind of clustering/failover are they doing?
>