Is it possible to dispatch my PAT runs to another machine on my local network in a similar way one would use Amazon EC2?
If so, how?
Edit: I've added a user voice post here. Vote for it if you agree this would make a nice feature.
The short and sweet answer is that this is not currently supported. We favored the EC2 functionality because it can expand indefinitely and everyone has access to it (while not everyone has access to several machines on their local network). It is possible to take a PAT project, move it to other machines, then delete design alternatives. That way you could run a subset of jobs on each machine. However, you'd have to do this manually and there is not any way currently to stitch those back together into a single result set. The good news is that PAT is due for a refresh to better align the local version with the large scale analysis capabilities. I'd suggest adding feature requests on user voice, we'll take those into account when designing the next version.
One option would be to use the spreadsheet export functionality in PAT to generate an OpenStudio Analysis Spreadsheet. The github repo for how to run and use the spreadsheet is found here. This will allow you to run your PAT project and use the algorithmic capability of OS-Server on EC2, however you wont be in the PAT world anymore.
Another option for running OS-Server jobs is to create local OS-Servers on your other machines using Vagrant, VirtualBox and Chef. Instructions on how to do this are found on the github page. This will allow you to submit model runs to your own OS-Server instances instead of EC2. This workflow is how we test and develop new OS-Server code, however it is a bit more involved than just running on EC2.
At my office, we have a 32-core machine that we all remotely connect to and run PATs on. We do it just by copying and pasting the PAT file while connecting remotely to the 32-core machine. Not the most elegant solution, but it works!
Asked: 2015-09-02 05:59:14 -0600
Seen: 412 times
Last updated: Sep 02 '15
I can't be the first one to come up with that question, but I couldn't find anything online... Apologies if I just missed it
I edited your question to avoid confusion about running OSP projects on network drives
@Julien, do you use a popular cluster scheduler like condor or SGE or do you have your own scripted/manual process for this?
Right now I've got squat, but I'm thinking about it considering I have 2-3 computers on standby I could use, and I'm knee deep into PAT runs right now. (You just got a shoot-out in the Autotune email, congrats)