Message boards :
Number crunching :
Application Switching Algorithm
Message board moderation
Author | Message |
---|---|
Darren Send message Joined: 18 May 99 Posts: 12 Credit: 1,642,195 RAC: 0 |
I am running three projects: Seti@Home (25% resource share) CPDN (50% resource share) Einstien (25% resource share) I have an Intel dual core processor. CPDN runs all the time on one core. I understand that, as it tries to meet the deadline. I have the switching between apps set to default at 60 minutes so I would assume it would switch between Einstein and SETI at 1 hour intervals but it doesn't. Seems to let one get to 95% then switch. Also, at the moment it is only doing Einstein. I had a SETI WU 'stick' so it worked on that for about 12 hours. Is it trying to catch up with Einstein work? Will it go back to SETI eventually? |
mikey Send message Joined: 17 Dec 99 Posts: 4215 Credit: 3,474,603 RAC: 0 |
I am running three projects: Yes this is called Long Term Debt, it will even out eventually. |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
I am running three projects: BOINC keeps track of something called DEBT, and over the long term, it is debt that keeps your machine doing 25% SETI, 25% EINSTEIN, and 50% CPDN. In the short run, BOINC watches debt (short-term debt, actually) and also keeps an eye on deadlines. If it sees that a work unit might miss deadlines, it will give extra time to that work unit. The extra time accumulates debt, and that debt has to be paid back some day. I'm guessing that BOINC did some extra SETI (because of the short deadlines) and is now "paying back" time to Einstein. Don't expect to always see CPDN running either -- BOINC may use both cores for SETI or EINSTEIN if it needs to (i.e. for deadlines). |
Odysseus Send message Joined: 26 Jul 99 Posts: 1808 Credit: 6,701,347 RAC: 6 |
Also, at the moment it is only doing Einstein. I had a SETI WU 'stick' so it worked on that for about 12 hours. Is it trying to catch up with Einstein work? Will it go back to SETI eventually? Long-term debt is used to decide what project to ask for work. But it’s short-term debt that determines which task(s) will be crunched when an opportunity to switch arises. See the BOINC Wiki [[Work Scheduler]] article. |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
BOINC keeps track of something called DEBT, and over the long term, it is debt that keeps your machine doing 25% SETI, 25% EINSTEIN, and 50% CPDN. It also looks like the host is running under a new Host ID, so it's hard to see any trends at this point. Also, it looks like Darren is new to crunching for EAH, so I guess it's possible that there was a latent LTD/STD issue which was being masked by having a dual core machine. When EAH came into the picture it would have popped in between the LTD and STD values for the CPDN and SAH which could explain why it's getting the lion's share of the second core at the moment. Since it doesn't look like it's malfunctioning in any way, probably the best thing to do is just leave it alone until it has more track record to work with. Most likely it's doing just what it's suppose to. Alinator |
Darren Send message Joined: 18 May 99 Posts: 12 Credit: 1,642,195 RAC: 0 |
Thanks for the answers folks. Whilst reading the thread it downloaded a Seti unit and started crunching. Panic over. I'm a bit of a worrier. |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
FWIW, the Combined Hosts stats for your account over on BoincStats tracks the work done for each project on a host on a daily, weekly, and monthly basis. It's a good way to keep an eye on what's going on if you want to verify your resource share is being followed over time. Alinator |
W-K 666 Send message Joined: 18 May 99 Posts: 19062 Credit: 40,757,560 RAC: 67 |
The switching, I believe, is done on wall clock time. The shorter time shown in the CPU time column is because the cpu has to run other things such as BOINC and any housekeeping tasks, and of course any programs you might be running. Andy |
Odysseus Send message Joined: 26 Jul 99 Posts: 1808 Credit: 6,701,347 RAC: 6 |
FWIW, the Combined Hosts stats for your account over on BoincStats tracks the work done for each project on a host on a daily, weekly, and monthly basis. It's a good way to keep an eye on what's going on if you want to verify your resource share is being followed over time. Given some kind of baseline, that is: any two projects with the same resource share are unlikely to produce credit at the same rate. Not only do the apps & WUs vary in terms of the type of processing required (memory-intensive vs FPU-intensive, &c.), making even benchmark-based productivity vary accordingly, but projects also vary in uptime, availability of work, and promptness of validation. I had a look at few of the hosts on my account to compare the past week’s production (according to BOINCstats) from projects with equal resource shares: iMac: S@h (opt) 57.3%; Einstein@home 31.5%, SiMaP 11.2% Dual Mac G4: S@h (opt) 50.2%; E@h 49.8% Mac G4: S@h (opt) 38.2%/58.3% = 65.5%; E@h 20.1%/58.3% = 34.5%; Pirates@home 0% Celeron/Win98: S@h (opt) 40.1%/63% = 63.7%; E@h 22.9%/63% = 36.3% I have no reason to believe BOINC is failing to honour my resource shares as best it can in any of these cases, even though only one of the hosts’s production appears to accord with equal shares. |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
Good point. Even though the projects have worked hard together to make sure there is rate parity between them all for the most part, there still is variation for a number reasons. Therefore, as you point out, you need to be aware of what the effective relative credit percentages should be normally to use BoincStats host data for this purpose. Alinator |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.