[ubuntu-za] What NAS device(s) do you use? And why?

[Rudi Ahlers]

On Mon, Dec 13, 2010 at 8:18 AM, Lee Sharp <leesharp at hal-pc.org> wrote:
> On 12/13/2010 12:45 AM, Andre Rossouw wrote:
>>> Fiber attached and SAS is a lot of money for little benefit. SATA is
>>> very fast, and very cheap.  For the cost savings you can buy two
>>> boxes,
>>> and have a logical mirror.
>>
>> Well - yes and no. It really does all depend on what you're putting on
>> those disks, and how quickly your application need to get to it. If you
>> are using a web-based application, or small db's, or need lots of
>> storage that seldom gets touched - then I agree, SATA is the way. If,
>> however, you are working with a relational db 1TB+,  with all the logs
>> and hundreds of users hitting the application servers throughout the
>> business day ... then I totally disagree. Fibre and SAS is well worth
>> the money.
>
> Frankly, in that case I would build an array with SATA solid state
> drives, or a giant RAM disk with lazy writes to an array.  Generally,
> the attachment point is not the bottle neck, it is the spinning
> platters.  Especially with SATA 3.
>
>                        Lee
>
> --


Thanx for all the replies to far, it's been very helpful and I
gathered a lot of info.


Right now I'm still researching, but my goals as as follows:

1. For a (possible new) client who needs to archive CCTV footage in a
security company for an NPO. I need the cheapest possible way of doing
this, while still being able to grow the storage within one host. The
PVR can only access one single NAS (via FTP) so I need to keep it
simple. Their budget is obviously limited so I need to keep it costly.

2. To change our backup strategy. Currently we backup all our client's
data (websites, email, databases) to multiple backup servers (at no
extra cost to the client) and simply purchase new 2U servers with 8
drive bays, slap in the largest SATA HDD's at the time, install
CentOS, setup RAID10 and use SSH / SFTP / RSYNC for backups. This
whole setup gets mirrored since one copy of a file on RAID hardware
isn't a backup. So, for every hosting package we sell, we need to
budget for 3x the storage (on the web server & 2x backup servers) and
need to keep costs down on this one. The problem comes with client's
who suddenly grow their storage needs and that particular server may
not handle the date, then we need to move the client's data to a new
server and tell him to use a new host, or configure his software with
the new hostname. It would be nice to combine these servers in one
large array and everyone just connects to one host.

3. I would like to move our virtual machines to central storage and
try and achieve higher uptime. I honestly can't afford a NetApp or EMC
for this, just to see if it works. Achieving high availability doesn't
just happen cause you have a nice expensive device. You also need to
prove, over a long period of time (without making a profit on it yet)
that it actually works well.



Ideally, I would like to stick to one solution / brand / vendor for
all these scenarios.

I don't like vendor lock-ins
I don't like solutions with factory fitted HDD's - if I need to
swap-out failed drives with spares that we already have, I want to. I
don't want to wait hours-at-end for a support tech to bring a
replacement HDD module.
I also don't like spending cash on high market hype. Sure, if a
product is good, it's good. But too many vendors / retailers /
suppliers hike up the prices due to market hype.

For now I think I'll stick to using bare-bone SuperMicro servers, and
then look at using something like Lustre / Gluster / Nexenta /
We use OpenFiler on a few of our current "NAS" servers, but it's
outdated and doesn't scale well so I need something new.
-- 
Kind Regards
Rudi Ahlers
SoftDux

Website: http://www.SoftDux.com
Technical Blog: http://Blog.SoftDux.com
Office: 087 805 9573
Cell: 082 554 7532