Hi,
I have a question that, if I have a program which is a parallel program and
work
properly in cluster with NFS sharing disk between computing nodes. But I
don't
know if it could work on two clusters (as two grids) with different NFS
sharing disks ?
On Thu, Nov 13, 2008 at 12:40 AM, Nicholas Karonis <[EMAIL PROTECTED]> wrote:
> mpich-g2 is far more battle-tested. MPIg is a complete re-write
> with performance enhancements, especially when running on compute
> nodes where there is more than one CPU.
>
> my advice would be to start with MPICH-G2 and migrate to MPIg
> as you need more performance.
>
> nick
>
>
> On Nov 12, 2008, at 9:19 AM, Yann JOBIC wrote:
>
> Hello,
>>
>> Thanks for the fast answer !
>>
>> So i have to choose between MPICH-G2 ans MPIg.
>> Which one is the most relevant to use ?
>> It seems that MPIg works with the last WS-GRAM of GT4. Is it better to use
>> this one ?
>>
>> Many thanks,
>>
>> Yann
>>
>>
>>
>> Nicholas Karonis wrote:
>>
>>> Dear Yann,
>>>
>>> Yes, it is possible to do that. Both MPICH-G2 and MPIg are MPI
>>> libraries designed to do precisely that - including automatically
>>> automatically converting data in messages sent between clusters
>>> with different architectures.
>>>
>>> You can learn more about MPICH-G2 at www.niu.edu/mpi .
>>> MPICH-G2 works only with pre-web services GRAM from the Globus Toolkit
>>> (i.e., GT2).
>>>
>>>
>>> You can learn more about MPIg (MPICH-G2's successor) via
>>> anonymous ftp:
>>>
>>> The tarballs and instructions to build are all available on
>>> NIU's anon ftp server ftp.cs.niu.edu: pub/karonis/MPIg.
>>> There you can find mpig-1.0.6p1.1.tar.{bz2,gz} (same source,
>>> compressed differently) and index.html (build instructions).
>>>
>>> MPIg works with the same pre-web services GRAM that MPICH-G2
>>> does and it also works with WS-GRAM found in GT4.
>>>
>>> Nick
>>>
>>> On Nov 12, 2008, at 6:18 AM, Yann JOBIC wrote:
>>>
>>> Hello,
>>>>
>>>> I would like to know if, with globus, i can run a parallel program over
>>>> 2 clusters (running gridengine) at the same time, with mpich communications
>>>> between the 2 clusters (and if the architecture is different too).
>>>> I made a figure which expose what i want to do. Cluster A and Cluster B
>>>> may have different architectures.
>>>>
>>>> Thanks,
>>>>
>>>> Yann
>>>>
>>>>
>>
>> --
>> ___________________________
>>
>> Yann JOBIC
>> HPC engineer
>> Polytech Marseille DME
>> IUSTI-CNRS UMR 6595
>> Technopôle de Château Gombert
>> 5 rue Enrico Fermi
>> 13453 Marseille cedex 13
>> Tel : (33) 4 91 10 69 39
>> ou (33) 4 91 10 69 43
>> Fax : (33) 4 91 10 69 69
>>
>
>
--
Le Trung Kien.
