Re: [Nelsonlab-analysis] [Nelsonlab-devel] Cluster Usage

[Nils H. <nil...@uc...>]

Presumably, if I am on compute-3-4 and another user is on compute-3-4 and we
are both writing to disk (or reading) then the bottleneck is the Ethernet
connection we are sharing on compute-3-4.  ³-pe serial 8² may be another way
of saying I don¹t want people rsyncing/reading/writing while I am using this
node.

Nils


On 2/10/10 5:00 PM, "Jordan Mendler" <jme...@uc...> wrote:

> Not sure SGE/Lustre are that complex. The closer thing will be once
> Inifiniband is working, we will be able to request IB nodes which have a much
> faster connection to Lustre.
> 
> Jordan
> 
> On Wed, Feb 10, 2010 at 4:02 PM, Nils Homer <nil...@uc...> wrote:
>> 
>> What about reserving I/O?  If a job is rsyncing etc. we don¹t want others to
>> use bandwidth.
>> 
>> 
>> Nils
>> 
>> On 2/10/10 3:38 PM, "Jordan Mendler" <jme...@uc...
>> <http://jmendler@ucla.edu> > wrote:
>> 
>>> I have noticed that many people have been submitting jobs with -pe serial 8,
>>> when their job only uses one thread. Please only use -pe serial #, if your
>>> job is multi-threaded and using # threads.
>>> 
>>> If you are simply trying to reserve memory, use -l vf=#G instead. Please
>>> note that this is on a per core basis, so -l vf=4G -pe serial 8 will request
>>> an 8-core machine with 32G of RAM. -l vf=4G -pe serial 2 will request 8G and
>>> 2 cpu's. -l vf=4G alone will request 4GB and 1 cpu.
>>> 
>>> Thanks,
>>> Jordan
>>> 
>>> 
>>> ----------------------------------------------------------------------------
>>> --
>>> SOLARIS 10 is the OS for Data Centers - provides features such as DTrace,
>>> Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW
>>> http://p.sf.net/sfu/solaris-dev2dev
>>> 
>>> _______________________________________________
>>> Nelsonlab-devel mailing list
>>> Nel...@li...
>>> <http://Nelsonlab-d...@li...>
>>> https://lists.sourceforge.net/lists/listinfo/nelsonlab-devel
> 
>