Message boards :
Number crunching :
Overall progress on server status?
Message board moderation
Author | Message |
---|---|
Send message Joined: 5 Jul 11 Posts: 990 Credit: 376,143,149 RAC: 3 |
On Einstein at Home, we can see overall progress of each application on the server status page. Is such a thing here or possible to create? |
Send message Joined: 16 Mar 10 Posts: 210 Credit: 106,115,573 RAC: 23,924 |
Peter, The Server Status page here only reflects work that hasn't been assimilated yet; I suspect that's the BOINC server default, and I don't know how much extra work would be needed for something like the Einstein output! Bear in mind that, unlike Einstein, there is no exact number of work-units required to complete a given analysis, so progress measurement would be tricky! New jobs are produced based on the outcome of previous jobs until some sort of convergence criterion is met. Perhaps Tom can tell us how close the current stripes seem to be to that point? Cheers - Al. P.S. I notice MilkyWay appears to have a new "intern" who might be in a position to look into this once he's learnt the ropes! However, I can think of more important things to fix, such as the 7-job work-unit issue... [Edit - re-wording] |
Send message Joined: 5 Jul 11 Posts: 990 Credit: 376,143,149 RAC: 3 |
What's the 7 job work unit issue? Two other things I can think of are: 1) the impossible task of making the server hand out new GPU tasks while taking ones back - note to intern, don't bother, Eric and Tom both failed. 2) changing the settings so I can put N-body on a CPU and Seperation on a GPU on the same computer. Currently the CPU also gets seperation, which the GPU could be doing much quicker. |
Send message Joined: 16 Mar 10 Posts: 210 Credit: 106,115,573 RAC: 23,924 |
What's the 7 job work unit issue? There have been multiple times when significant numbers of work units got created that appeared to contain 7 jobs. Trying to deal with the results returned for these tasks could crash the validator, and some of the jobs weren't genuine anyway. There's lots of threads in various places related to the issue; Tom's explanation can be found in the thread Validator Outage in the News section... Two other things I can think of are: I didn't mention the first one because it is probably related to BOINC Server configuration and the issue may or may not disappear if/when the latest BOINC software is installed as per Dylan's introductory message. Your second one is a valid point -- if a science application has both CPU and GPU versions, one should be able to pick one or both... (There's a similar issue at WCG with OPN1/OPNG, though there are workarounds suggested there...) And if we're on the subject of work downloads, those of us who do multiple projects would probably welcome a "Send me no more than..." facility like the one at WCG. I'd especially appreciate that at Einstein!!! At least with Dylan in place there's a chance some of the issues can/may/will be addressed at last :-) Cheers - Al. |
Send message Joined: 24 Jan 11 Posts: 708 Credit: 544,093,763 RAC: 125,066 |
I added a request for the bugs he is intending to work on in my welcome post. I agree, the 7-job task bundles are top of the list to squash. |
Send message Joined: 10 Apr 19 Posts: 408 Credit: 120,203,200 RAC: 0 |
Hey everyone, I'm actually in the process of preparing new data to put up. I want to take down the runs that are currently up and replace them with other stripes in the southern sky. We are also in the middle of some extensive analysis of the previous results, but it's slow going because we have many projects. There was a talk that I gave on this project at a conference in 2020 (https://ui.adsabs.harvard.edu/abs/2021DDA....5240303D/abstract), although this information is somewhat outdated and I am not able to find the recording of the talk. I might dig around further if I have some spare time. |
Send message Joined: 24 Jan 11 Posts: 708 Credit: 544,093,763 RAC: 125,066 |
Thanks for some project update news Tom. Will be nice to look at new stripe data. Hope it all works out. |
©2024 Astroinformatics Group