Message boards : Number crunching : What to do with 5.13 WUs?
Author | Message |
---|---|
David@home Send message Joined: 7 Oct 05 Posts: 29 Credit: 185,330 RAC: 0 |
With version 5.16 being released should we continue with the 5.13 WUs held in cache or should we abort these to download new 5.16 WUs? Thanks |
Kerwin Send message Joined: 19 Sep 05 Posts: 10 Credit: 1,773,393 RAC: 0 |
For the last few versions, the project has asked that you do not abort units and just let them finish. This was stated for versions 5.12 and 5.13. Thought I didn't see it explicitly stated for 5.16, I'm sure the same holds true. |
Astro Send message Joined: 2 Oct 05 Posts: 987 Credit: 500,253 RAC: 0 |
I just reported a 5.07 this morning and have a bunch of 5.12 I'm still planning on crunching. tony |
David@home Send message Joined: 7 Oct 05 Posts: 29 Credit: 185,330 RAC: 0 |
OK, I will leave my cache as is. Sods law came into play and I downloaded 12 WUs for version 5.13 only for version 5.16 to be released a few hours later. I have noticed that Rosetta seems to handle the cache differently to Seti@home. Rosetta seems to download a large number of WUs and then drain them right down to the last WU being completed before downloading new work whilst S@H seems to download fewer WUs but more frequently to keep the cache topped up. This has happened several times now with Rosetta since I rejoined the project. Is this normal behaviour for the project or is it something happening within the Core Client scheduler? |
Astro Send message Joined: 2 Oct 05 Posts: 987 Credit: 500,253 RAC: 0 |
Rosetta seems to download a large number of WUs and then drain them right down to the last WU being completed before downloading new work whilst S@H seems to download fewer WUs but more frequently to keep the cache topped up. This has happened several times now with Rosetta since I rejoined the project. Is this normal behaviour for the project or is it something happening within the Core Client scheduler? It's the boinc Work Scheduler doing that. Once it gets a better balance on resource share you should see it be the same. It has to do with the LTD (long term debt) numbers and Work fetch. |
Moderator9 Volunteer moderator Send message Joined: 22 Jan 06 Posts: 1014 Credit: 0 RAC: 0 |
|
David@home Send message Joined: 7 Oct 05 Posts: 29 Credit: 185,330 RAC: 0 |
The word is you should let the work units for prior versions run to completion Cool. For future news items re new versions can a line always be added to the effect Please allow currently downloaded work units to run to completion. or Please abort any Work Units for version x.x This would stop any confusion on what action to take. Thanks |
Moderator9 Volunteer moderator Send message Joined: 22 Jan 06 Posts: 1014 Credit: 0 RAC: 0 |
The word is you should let the work units for prior versions run to completion Well the standing condition is to not abort Work units unless there is a problem reported in this thread. You might want to subscribe to that thread so you will get an e-mail if something is posted there. In particular during CASP people should not abort Work Units if they can avoid doing so. Moderator9 ROSETTA@home FAQ Moderator Contact |
David@home Send message Joined: 7 Oct 05 Posts: 29 Credit: 185,330 RAC: 0 |
The word is you should let the work units for prior versions run to completion I get the news via my RSS reader and I would guess this is a popular method for many other users as well. It is great to see news about new versions etc; some other projects leave you totally in the dark. Trouble is when you see news that a new version is out first thing you wonder is what about my cache of old WUs; are they just going to waste CPU time? A quick line in the news item would make this clear there is no need to take any action. Being quite busy I do not always have the time to visit the project web sites which is one reason I like RSS so much. |
Moderator9 Volunteer moderator Send message Joined: 22 Jan 06 Posts: 1014 Credit: 0 RAC: 0 |
The word is you should let the work units for prior versions run to completion I would be surprised of the project ever put messages to abort work units on the home page. The news from the home page feeds the RSS so that is why the thread mentioned above was established. But in any case the simple change of a version will not cause a need to abort older work. All the results are still valid. So the project position is to never abort work unless there is a specific problem, and that would be posted in the thread. Moderator9 ROSETTA@home FAQ Moderator Contact |
Message boards :
Number crunching :
What to do with 5.13 WUs?
©2024 University of Washington
https://www.bakerlab.org