[Bioclusters] Urgent advice on RAID design requested

[Christopher Dwan]

Honestly, most of my objection to the use of this device comes from  
having helped several groups go through the uncomfortable transition  
from "wow, we have a whole lot of firewire disks hanging off our  
workstations, our data is all over the place, and it's clearly not  
sustainable" to a central storage solution.  I understand that you  
can rig it up as an NFS server, but it's the same idea:  Little  
consumer boxes that sit on shelves.  Buying one of these devices is  
the first step down a road that ends up with 10 of them sitting in a  
little row on a shelf, and the users having to sort through which  
volume they happened to have their stuff on.

This device could be different from all of those that I've seen  
before, but I doubt it.

The canonical answers to "why should we spend more per TB" are:

* Reliability:  There is a big difference in the vibration,  
temperature, and general use characteristics of a device intended for  
intermittent, home use, vs. one intended to be slammed 24/7.  Sure,  
RAID protects from individual disk failures ... but there are more  
components than just the disk.

* Scalability, both in I/O bandwidth and storage volume.  Data needs  
never shrink.  So whatever solution is selected needs to be able to  
grow as a small number of pieces, rather than multiply and wind up as  
dozens of small volumes, of different sizes and provenance, all over  
the network.

* Maintenance overhead:  Number of parts in the solution,and the  
number of 3rd party vendors who are the Only Ones In The World who  
can help you if one of those parts goes sideways.

If $ per TB were the only criterion, I would go with the 1TB firewire  
disks that are all over the market today.  $0.50/TB.  On the other  
hand, just this week I was working with a cluster whose owner had  
tried to share one of those firewire devices out over NFS to a couple  
of dozen compute nodes.  Not only was the performance atrocious  
(firewire is best for streaming data, not the frenetic reads and  
writes characteristic of a compute cluster), but the NFS server  
itself kept falling over because of the latencies in getting to the  
device.

That said:

* You can get it populated from Lacie for $1299.  $500 is the chassis.

http://www.lacie.com/products/product.htm?pid=10877

* As I read through the manual, they claim that:

"Macintosh and Linux users can access the network storage device, but  
Windows is necessary for setup and administration of the device."

- Chris Dwan

On Jan 17, 2007, at 6:49 PM, Chris Dagdigian wrote:

>
> Reliable? With a single power supply?  That may be a good home NAS  
> box or perhaps scratch space for a production server but I'd  
> certainly not list it in the "extremely reliable" category.
>
> I can't be sure but I think the $499 price listed in your URL does  
> not include any actual SATA drives. Other online sites list this  
> device (populated 2TB raw) for $1200. For that kind of money I'd be  
> looking at Infrant instead of Lacie.
>
> My $.02
>
> -Chris
>
> On Jan 17, 2007, at 12:50 PM, Angulo, David wrote:
>
>> I disagree with you.  Look at http://www.bestbuybusiness.com/bbfb/ 
>> en/US/adirect/bestbuy? 
>> cmd=catProductDetail&productID=BB10725830&operation=showDetails   
>> This is RAID, so the mean time between failure will be extremely  
>> large.  This is extremely reliable.
>
> _______________________________________________
> Bioclusters maillist  -  Bioclusters at bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters