Message boards : Number crunching : Rosetta keeps preempting
Previous · 1 · 2
Author | Message |
---|---|
Rusty Lafavour Send message Joined: 12 Apr 06 Posts: 4 Credit: 26,391 RAC: 0 |
Thanks for the responses. Even though time is of the essence for me on the CPN. I will let what happens, happen for a couple days to see the final results of both projects. I feel good helping out on both projects, running 24/7 to help |
Rebirther Send message Joined: 17 Sep 05 Posts: 116 Credit: 41,315 RAC: 0 |
The problem is downloading new files. So Boinc is switching projects (WUs). My experiences are to download WUs for some days and stop getting new work. So you can see that CPDN is running all the time on one CPU and Rosetta on the other. You can try this. It helped me ;) |
Rollo Send message Joined: 2 Jan 06 Posts: 21 Credit: 106,369 RAC: 0 |
What about setting no new work for CPDN, so you have only one model running, and than increase the resource share of CPDN to more than 50%, lets say 60%. Since you have only one CPDN model, boinc cannot run that model on both cores simultaneously. So you can run CPDN effectivly at maximum of 50% share. This is lower than your setting of 60%. Therefore CPDN should always run and prevent rosetta from starting on both cores. |
Rusty Lafavour Send message Joined: 12 Apr 06 Posts: 4 Credit: 26,391 RAC: 0 |
|
David@home Send message Joined: 7 Oct 05 Posts: 29 Credit: 185,330 RAC: 0 |
Just trying to figure out why my Rosetta WUs only run for about three hours rather than the default of four hours. Using the default settings for Target CPU run time each WU seems to run for about three hours. This raises a couple of questions: 1) For the model steps that have not been run are these assigned to another WU so all combinations are covered or are they simply not available in the results database? 2) Why does Rosetta decide not to run the fourth hour? Is it because it thinks it can not complete the fourth hour in the available Switch between applications every 60 minutes slot? Thanks |
Astro Send message Joined: 2 Oct 05 Posts: 987 Credit: 500,253 RAC: 0 |
I use the default "not selected" and I see the same thing on 3 puters. Don't know why. my amd64 3700's will usually to 20 models/hour, so that shouldn't make it stop at 3 hours. |
Feet1st Send message Joined: 30 Dec 05 Posts: 1755 Credit: 4,690,520 RAC: 0 |
Just trying to figure out why my Rosetta WUs only run for about three hours rather than the default of four hours. The switch between applications is basically a guideline more than a rule. The "switch time" is really just telling BOINC how frequently you want it to reconsider what should be running. Often times it decides to continue running what it has going already. But I believe it ALSO reconsiders when a WU completes. As for why they complete is 3 hours timeframe rather than 4... check how many models you got crunched in what time. Divide it out and figure out how many minutes per model, and if you add one more model to the existing runtime, if you cross over the 4hr preference, then you've got it with #2, "it thinks it can not complete" another full model in under your 4hr objective. So, it ends now instead. The client application (Rosetta in our case-at-hand) does not KNOW when BOINC may rip the CPU away from it. It is blissfully unaware of the "switch between application every" setting. This is why the new checkpoint is such a huge great deal! As for point #1, it's really a huge space, the number of models POSSIBLE on a WU. And so they randomize it. The project isn't depending upon you to crunch these 20 given models or whatever. It's trying to get the install-base to crunch on the order of a million models at random points in a space where there are at least hundreds of millions of models possible. As you might expect, with this randomness there is a small amount of duplication where two clients strike upon the same starting value and essentially crunch the same exact model. But the space it so vast that this duplication is still less than 5%. If they had 100x more crunching power and were going for 100,000,000 models instead, then they might have to assign it out to eliminate duplication, but we're not there yet. (all that info. is based on other posts I've read. I don't really have any other knowledge about it). Add this signature to your EMail: Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might! https://boinc.bakerlab.org/rosetta/ |
AMD_is_logical Send message Joined: 20 Dec 05 Posts: 299 Credit: 31,460,681 RAC: 0 |
Using the default settings for Target CPU run time each WU seems to run for about three hours. Those WUs probably have their default set to three hours. |
Message boards :
Number crunching :
Rosetta keeps preempting
©2025 University of Washington
https://www.bakerlab.org