Message boards : Feedback : Why so little testing on Ralph?
Author | Message |
---|---|
tralala Send message Joined: 12 Apr 06 Posts: 52 Credit: 15,257 RAC: 0 |
Hi, I just read on Rosetta that 5.01 has problems with some WUs. I wonder why you didn't create more jobs for 5.01 here on Ralph? I wanted to check out 5.01 on Ralph before you released it to Rosetta but there was already no work available. For Ralph there is more computing power present than you seem to use. Soem errors show only up in big samples so I suggest to increase testing here on Ralph. Or is the science returned on Ralph not comparable to that returned on Rosetta? |
Moderator9 Volunteer moderator Send message Joined: 16 Feb 06 Posts: 251 Credit: 0 RAC: 0 |
Hi, There really is no science use made of the RAPLH protein results beyond testing the application itself. If a RAALPH application generates errors then it will not be deployed in Rosetta. Also RALPH is used to test Work Unit stability, and to test bug fixes. So the actual protein results generated by RALPH processing have almost no use as science results beyond the fact that they run to success or they don't. As for testing 5.01, if you are running Rosetta it is currently running 5.01, but that will change soon. Moderator9 RALPH@home FAQs RALPH@home Guidelines Moderator Contact |
tralala Send message Joined: 12 Apr 06 Posts: 52 Credit: 15,257 RAC: 0 |
I wonder why. Aren't the proteins and the application the same which are eventually deployed to Rosetta? The results should be of the same quality as the results from Rosetta. Are there technical reasons why the results can't be used? |
Moderator9 Volunteer moderator Send message Joined: 16 Feb 06 Posts: 251 Credit: 0 RAC: 0 |
Well you have to remember, RALPH is where the bugs get worked out. This includes things like new methods to running the models. In some cases the math is not yet really worked out. To some extent the results are valuable, but there really are not enough of them generated in test runs to put a dent in the 10,000 to 100,000 models required to really do the science. The RALPH plan is to get the application and the Work Units tested and get then out to the calculation power of the main project as soon as they can validate them here. Also Ralph is testing things like Work Unit aborting, and other things that cut the model runs short. Last but most important, all the Proteins run on RALPH for testing are small so the results come in faster. This is true most of the time, unless they are testing a new Work Unit type for deployment on Rosetta, in which case the protein could be quite large. If that is the case they want to run them against a good sample of machine types to be sure they work, but still get them out to the main project. That said some of the models are probably kept and added to the science data but that is not the true focus of this part of the project, and they would require very close examination first. Moderator9 RALPH@home FAQs RALPH@home Guidelines Moderator Contact |
Message boards :
Feedback :
Why so little testing on Ralph?
©2024 University of Washington
http://www.bakerlab.org