Hi all, |
I was hopping some of you may be able to answer a question I have about Condor.
We are looking to implement grid computing on or site, and ideally we would look to begin with a dedicated cluster of PC's (10 or 20) as a proof of concept. But at a later date I am very interested in looking to move the client side of the deployment to users desktop PC's.
I have in mind two system, CONDOR and BOINC, and from what I can see both have positive and negatives. Condor seems a simpler set up and more focuced towards in house grid/distributed computing. However the area I see that BOINC seems to score on for me, is that the BOINC client can continue running as a background process, while a user is still on there PC working on non CPU intensive task, rather than stop an have to pass it over to another client device?
We are looking at using it for batch running of applications that may take several hours to complete a run, so the benifit gained from switching tasks from client to client is not as high.
Also we are looking at running along side this project some thing like Gain-Grid for true processes level grid computing, is this something that will over lap with condor, or some thing condor can already do?
I have played around with BOINC a little and another member of staff has looked in to Gain-Grid. But we both like the look of CONDOR, however as time is short I would rather know before hand what benefits is would offer over either of these other two solutions before spending to much time setting it up.
So in summer my questions are,
Can condor run as a client (on windows machines) and make use of wasted CPU cycles while a user is active?
and what level of granularity is there to the processing models? application, process, thread?
Thank you, and if you have any other info you think would be useful please let me know.
Got a cool Hotmail story? Tell us now