[HTCondor-users] Questions regarding MW https://research.cs.wisc.edu/htcondor/mw/

[Vikrant Aggarwal <ervikrant06@xxxxxxxxx>]

Hello Expert,

I was reading about the topic mentioned in subject. I really don't understand the purpose of it I know about MPI jobs which are using multiple machines with dedicated scheduler configuration.Â

Queries regarding MW:

- What kind of benefits this library will bring and how it can impact the normal functionalityÂof condor?Â

- Also we don't use checkpoint functionality so probably I need to rely on socket based communication out of 4 methods mentioned in [1]

- Is this actively maintained I can see last update of 2006.

- Do we need to compile it on submitÂnode and job will do the negotiation using condor master only?

Text from condor wiki:

In the world of parallel jobs, HTCondor supports the standardÂMessage Passing InterfaceÂandÂParallel Virtual MachineÂ(Goux, et al. 2000) in addition to its own Master Worker "MW" library for extremely parallel tasks.

[1]Âhttps://research.cs.wisc.edu/htcondor/mw/usersguide.pdf

Any inputs are highly appreciated.

Thanks & Regards,

Vikrant Aggarwal