Preferred Run Time?

Message boards : Number crunching : Preferred Run Time?

To post messages, you must log in.

AuthorMessage
zombie67 [MM]
Avatar

Send message
Joined: 8 Aug 06
Posts: 70
Credit: 1,847,169
RAC: 0
Message 2491 - Posted: 7 Nov 2006, 3:17:41 UTC

With Rosetta, I understand that there is no \"better\" run time from a science perspective.

But what is the best for RALPH? Is the goal to just to see if a WU works? In which case, would a minimum run time be best, to process the largest variety of WUs? Or is it better to try to run as many models from the same WU?
Dublin, CA
SETI.USA - Stats - My stuff - BOINC IRC chat
ID: 2491 · Report as offensive    Reply Quote
Profile feet1st

Send message
Joined: 7 Mar 06
Posts: 313
Credit: 113,747
RAC: 0
Message 2500 - Posted: 7 Nov 2006, 22:57:38 UTC

Well, the question has come up before and the project team was rather set on quick turn around. Highly focused apparently on proving the new release has a low error rate, and getting it out to Rosetta.

I pointed out that this is not how true testing should be done. Indeed, presently, the point is being brought up that R@H is having screen saver problems, and many Ralph users probably don\'t USE the screen saver... so right there, we\'ve not run the program in the mannar it will be run on Rosetta. And this difference means a difference in Ralph predicted error rate, and the Rosetta experienced error rate (my guess).

Indeed after I pointed this out last Spring, there were issues were some WUs were logging extensive msgs to stdout and only if you ran the WU long enough would you get an error about the file being large enough that it exceeded a maximum limit that is set for the output (to prevent programs from looping out of control and consuming all of your disk space by mistake).

So, MY advice, for what it is worth, is simply to run Ralph with the same preferences as you do Rosetta. I feel this is the truest testbed for Ralph.
ID: 2500 · Report as offensive    Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 8 Aug 06
Posts: 70
Credit: 1,847,169
RAC: 0
Message 2501 - Posted: 7 Nov 2006, 23:57:58 UTC

Thanks for your thoughts. Your approach seems reasonable to me.
Dublin, CA
SETI.USA - Stats - My stuff - BOINC IRC chat
ID: 2501 · Report as offensive    Reply Quote

Message boards : Number crunching : Preferred Run Time?



©2018 University of Washington
http://www.bakerlab.org