[gradsusr] Recommended hardware for GrADS

[Arlindo da Silva]

On Thu, Oct 20, 2011 at 8:07 PM, Joe Moore <joe at wxjoe.com> wrote:

> Hello fellow users and GrADS developers!
>
> I'm researching to purchase a new machine to run GrADS. I'm buying a new
> machine because I don't keep my computers on 24x7 and I would like a more
> reliable method of creating real-time plots than my cheap shared web
> hosting. (By using my own machine I will also be able to use GrADS-DODS,
> since my web hosting blocks port 9090.)
>
> My plan is to purchase a "nettop" computer such as this<http://www.newegg.com/Product/Product.aspx?Item=N82E16883220072>or
> this <http://www.newegg.com/Product/Product.aspx?Item=N82E16883103297>;
> I'm looking for a bottom-of-the-line, energy efficient machine to run 24/7.
> I will be installing some variant of Linux on it.
>
>
I know it costs a bit more, but if it is within your budget Mac Mini's are
really quiet and energy efficient:


http://www.amazon.com/Apple-MC815LL-Desktop-NEWEST-VERSION/dp/B004YLCLM6/ref=sr_1_1?ie=UTF8&qid=1319159557&sr=8-1

It can also serve as an entertainment system... I am not sure you can add
SSD disks to Mac Minis...


> My question is: As far as GrADS and speed goes, what would be the best
> components to invest in? I'm no computer scientist, so I'm not sure if, for
> instance, purchasing an SSD hard drive would increase the speed more than
> extra RAM. (At the start, GrADS will be the only software running - maybe
> multiple instances at the same time, though.) I also found this message from
> Jennifer that stated that GrADS only runs on one processor; is this still
> true?
>
>
Most of those smaller machines have a limit on the amount of  memory it can
take, so check the specs very carefully. The listed 2GB will get dated
really soon. You can run multiple instances of grads with no problem (each
one runs in 1 processor). However, they both must fit in memory and with 2GB
you will be pushing it; make sure you can have at least 4GB; 8GB would be
better. If you have a way of testing this now, check how much memory is
taken by your typical grads process (on linux, use "top" to monitor memory
consumption). Remember, running multiple instances only makes sense if you
have multiple cores.

Having a SSD disk (lower capacity ones such as 64 GB are getting affordable)
would greatly improve I/O (both for gridded data as well as for reading map
database during plotting). Since your application is I/O bound, this could
greatly improve performance.

So the answer is both: buy enough memory to fit multiple instances of grads,
a small SSD disk to hold the OS and gridded datasets.

   Arlindo

-- 
Arlindo da Silva
dasilva at alum.mit.edu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://gradsusr.org/pipermail/gradsusr/attachments/20111020/c3ffd7eb/attachment-0003.html