[mpich-discuss] A Windows 7 implementation question
Hiatt, Dave M
dave.m.hiatt at citi.com
Fri Feb 19 15:56:47 CST 2010
I wonder if there is anyone else who might have a similar issue/need out there and if so, what you might be doing to try and address it.
We have a kind of unique need / limitation that our "test grid" is also our personal workstations. During the day when we want to do basic testing, we actually need to run in the background. When we were under XP, we had a single UserID that was used and valid on all workstations. We've since upgraded to Windows 7 with the advent of new much more powerful workstations. But, with Windows 7 comes with some limitations, in that it doesn't want to allow background runs when someone is signed on in the foreground unless the userid is the same. So, I was wondering if anyone else out there has a similar topology that they have to deal with, normal user workstations by day with users running, cluster by night, but run as a cluster in either environment. I know it's not the best arrangement, but one has to make due. And all this as Windows 7.
There are some patches that will allow Windows 7 to support background execution. But I'm hoping to not have to run that way if I can. So
1 - Is there an easy obvious method with MPICH2 that I just don't know about that will allow it to support runs from one workstation on others by using the various run id's on each workstation. I'm assuming this really doesn't exist, but for completeness I'm asking.
2 - I know we can wait till afterhours and run with the single dedicated version, but
3 - Has anyone customized a local smpd that would force runs under the local runId and then return to the starting workstation. That's what I'm hypothesizing is the approach I'm going to need to take to get this to run. Anyone tried to make local changes to try and deal with that kind of solution?
4 - Am I overlooking some other obvious solution that would be easier and solve all these problems beyond having to buy a whole new set of dedicated systems to run as a full time test cluster (right now I'll never get funding for that but I realize that is the solution that SHOULD be done, but I don't have anyone to bankroll it).
Thanks
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20100219/2adce057/attachment-0001.htm>
More information about the mpich-discuss
mailing list