Please forgive me the naivety of this question, but - for the first time in my many years of running AMI on a single workstation - I'm trying to set up a new, more efficient environment for my work. Particularly, I'd like to be able to run analyses on 2 networked computers using the Distribution Queue of Job Manager for DOE.
This works OK when just a single study is run, launched from any of my 2 computers. But - with all this power of multi-core CPU and GPU - it's not very efficient, with the consecutive DOE jobs executed one after another (which of course must be so by the very design). Especially with Dual Domain models, it'd be nice to be able to launch several studies at the same time...
However, when I try to launch another analysis sequence (also including DOE), a very strange thing happens: both are launched, and even the solvers involved seem to be running (as per the Task Manager) - but in the Job Manager itself, their progress ceases to advance after a while (at varying stages - sometimes as soon as at 15%, sometimes in the middle of the first 2 iterations (1 per each workstation))... I tried this scenario several times, each time with the same result.
What's more, when this happens the Job Manager on one or both machines cannot be connected to, even locally - and the computers need complete rebooting.
Now, my question is:
- should it be possible to execute 2 (or more) DOE analyses for different studies at the same time, using 2 (or more) networked computers?
If so, what may be to blame for such a scenario not working here:
- the Job Manager service in Beta 3 ?
- the DOE algorithm in Beta 3?
- something else (perhaps I'm doing something wrong)?
I'll appreciate all comments; thanks a lot
Piotr
PS. Sorry for cross-posting, but to get the widest response possible I'm also posting it in the Beta testing area forum