[Swift-devel] Re: replication vs site score

[Ioan Raicu]

Add another constraint, that you only have 6 jobs that you can submit to 
the LRM queue, are you still as optimistic about using dynamic resource 
provisioning?

Mihael Hategan wrote:
> On Thu, 2009-04-09 at 09:56 -0700, Ioan Raicu wrote:
>   
>> Assuming you have more work to do, than the static resources
>> allocated, you are not wasting any resources. The workflow will run
>> until the resources are de-allocated, and whatever was not completed,
>> will get rescheduled on the next round of static resources allocated.
>>     
>
> Right. As opposed to the system figuring out that there is more work and
> having workers ready appropriately.
>
>   
>>  As far as I know, this is the usage pattern of the static resource
>> allocation on the BG/P for the few regular users, that are running
>> several jobs per day, where each job is a static resource allocation
>> of 1K~10K processors for several hours each. Their parameter space is
>> large enough that they keep doing this over and over again, and they
>> still more work to do!
>>     
>
> Which, seems to show that, for static provisioning to work efficiently,
> you need to fit work exactly into the pre-allocated resources, or have
> pre-allocated resources to exactly fit your work load.
>
> I'm still not sure why you would choose this instead of "allocate
> resources on demand and de-allocate resources when you stop needing
> them".
>
>
>
>   

-- 
===================================================
Ioan Raicu, Ph.D.
===================================================
Distributed Systems Laboratory
Computer Science Department
University of Chicago
1100 E. 58th Street, Ryerson Hall
Chicago, IL 60637
===================================================
Email: iraicu at cs.uchicago.edu
Web:   http://www.cs.uchicago.edu/~iraicu
http://dev.globus.org/wiki/Incubator/Falkon
http://dsl-wiki.cs.uchicago.edu/index.php/Main_Page
===================================================
===================================================

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/swift-devel/attachments/20090409/954e205f/attachment.html>