An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-hpc/attachments/20110929/2ae99779/attachment.pl>
Suggestions for dedicated server for simulations
9 messages · Johannes Edinger, Whit Armstrong, Carl Boettiger +3 more
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-hpc/attachments/20110929/d69fd3b1/attachment.pl>
you can get a lot out of EC2 these days if you know how to leverage it. I would be very reluctant to buy a new system these days unless you really need your data to be local (perhaps for security reasons). check out the cluster compute instances on ec2: http://aws.amazon.com/ec2/#instance -Whit On Thu, Sep 29, 2011 at 12:36 PM, Johannes Edinger
<hannesedinger at gmail.com> wrote:
I've been asked by our IT department what sort of server (including specifications) our team requires to run our simulations. ?Our models aren't terribly complex, but they are too demanding to run on the main server and we anticipate that we will require more capacity in the future. ?I have no idea where to start, we are new and I've been running these simulations on my laptop with 4gb ram and dual core processor. ?I run far fewer simulations that I would like to so my laptop doesn't burn a hole in my desk and now we have the opportunity to send the jobs to a dedicated server. Where should I start? Thank-you -- B. Johannes Edinger ? ? ? ?[[alternative HTML version deleted]]
_______________________________________________ R-sig-hpc mailing list R-sig-hpc at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-hpc
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-hpc/attachments/20110929/56455fca/attachment.pl>
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-hpc/attachments/20110929/9ee9170b/attachment.pl>
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-hpc/attachments/20110929/d051889a/attachment.pl>
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-hpc/attachments/20110929/7b4d46ed/attachment.pl>
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-hpc/attachments/20110930/7e18d1a2/attachment.pl>
Hi Considering storage and speed one option looking it in my mind is SSD disks. For intersections with large rasters on desktops I have positive experiences with that. I can imagine that for genome work where large datasets can't be read into memory it also works. I guess it is less relevant for the specific case here but it is worth considering in general. Bart
On 09/30/2011 10:21 AM, Rainer M Krug wrote:
On Thu, Sep 29, 2011 at 9:46 PM, Michael Hannon<jm_hannon at yahoo.com> wrote:
I generally agree with the analysis given so far in this thread, but let me point out:
...and if data will not be backed up elsewhere, consider RAID mirroring
hard disks. ... If you accidentally delete all of last week's work on one mirrored drive, it is immediately gone from the other drive. Mirroring does guard against hardware errors but not wetware errors. I'd suggest backing up anything that's of importance to you.
To say it in R code: raid != backup To be safe, you need both. Raid has another advantage, which does not seem to be important for you: with the right RAID level, hdd access will be faster. Rainer
-- Mike