Questions and Answers :
Preferences :
How to get proper service with 2 projects?
Message board moderation
Author | Message |
---|---|
Cruncher-American Send message Joined: 25 Mar 02 Posts: 1513 Credit: 370,893,186 RAC: 340 |
I was running SETI, but had the same problems everyone else was having recently - getting no work. Having finished all SETI WUs I had, I attached to Einstein, and got a bunch of WU. (I have an 8-core machine, so I set the processing shares as Einstein - 100, SETI - 700, thinking that when I got some SETI WU, I would gradually take over 7 cores as Einstein WUs finished). And sometime thereafter, the Heavens opened and I got 18 SETI WUs downloaded. That hasn't happened - I have had 2 Einstein WUs finish, but BOINC started up two more E WUs, ignoring SETI, even though, according to BOINC Mgr Projects page, the Resource shares are E/SETI 100/700. BTW, all the E WUs I have are approx. 8-10 hours, and have deadlines of 7/17; the SETI WUs are 1-2.5 hours and have deadlines of 7/26. Have I done something wrong, or missed something? Why won't it start even one SETI WU? Thanks for your help! |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
I was running SETI, but had the same problems everyone else was having recently - getting no work. Having finished all SETI WUs I had, I attached to Einstein, and got a bunch of WU. (I have an 8-core machine, so I set the processing shares as Einstein - 100, SETI - 700, thinking that when I got some SETI WU, I would gradually take over 7 cores as Einstein WUs finished). And sometime thereafter, the Heavens opened and I got 18 SETI WUs downloaded. The round robin simulator that determines when tasks will finish if the CPU scheduler runs normally factors in the resource fraction (proj resource share / sum of all resource shares). Set "Connect every X" to reality, i.e. how often you really let BOINC attach to the network to request, report, upload, and download work. If always connected, you might want to set this to something small (like 0.05). Set Extra work to a day or two. When Einstein gets work downloaded, it will likely run its cache out because of the low resource share. S@H will run most of the time though. A 7 to one ratio does not guarantee that 7/8 of the CPUs will be used for that project, it guarantees an attempt to make the resource shares balance over the long term (a few weeks in this case - if CPDN were in the mix, the balance would come after many months). BOINC WIKI |
Cruncher-American Send message Joined: 25 Mar 02 Posts: 1513 Credit: 370,893,186 RAC: 340 |
Thanks for that info. I forgot to mention that after E d/l the WUs to compute I had set E to "no new tasks" so it would not d/l any more (it has about 85 - approaching the 4-day queue that I had the WU cache set to (assuming using all 8 coreas for E)). Unfortunately, that happened before I boosted SETI to 700 resource share (from 100). Does that play into this scenario? I have a cable modem, so I always leave BOINC at "network activity always available". Is that wrong in this case? (I don't quite get what you said about net activity above). |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
Thanks for that info. Ok, you have an always on connection. There are two settings that determine the full queue length. 1) Connect every x - this should be set to be approximately what the duration between connections is. In this case, it needs to be set near 0. BOINC uses this to determine the minimum amount of work, and to set the computation deadline (which is earlier than the report deadline). 2) The extra work setting. This is added to Connect every X, but is NOT used in the calculation of the computation deadline. In your case, I would suggest settings of 0.05 for Connect every X, and somewhat larger (a couple of days) for extra work. Havint a large queue of work download from Einstein, and then having the resource fraction for Einstein cut will indeed have Einstein be scheduled for some time, that time will be made back later by not allowing download of Einstein for a bit. jm7 2) Extra work. This is BOINC WIKI |
Alexander Aas Send message Joined: 13 Mar 06 Posts: 17 Credit: 2,003,170 RAC: 0 |
Intersting, this is somewhat realted to what I have going on. I got a computer always connected to the net, and I gto S@H as well as rosetta@HOME. I gave rosetta 100 and SETI 900 in priority. So far its fine, but today I have not gotten any new work units for the STI. I still got 13 hours left of teh astro pulse, but still, nothing more. and only 2 samll packs of rosetta. And I selced 4 days buffer ^^. So if I get this right, you suggest setting this: Computer is connected to the Internet about every (Leave blank or 0 if always connected. BOINC will try to maintain at least this much work.) to 0.05 Am I right? Even tho I am connected 100% of the time to the net? I might be asking a silly question, but I just want to confirm it ;) |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
Intersting, this is somewhat realted to what I have going on. I got a computer always connected to the net, and I gto S@H as well as rosetta@HOME. I gave rosetta 100 and SETI 900 in priority. The setting of 0 allows a CPU to go completely dry in some circumstances before work fetch is initiated. A setting of 0.05 (an hour+) means that the computer will always start looking for work before the CPU is completely dry. BOINC WIKI |
©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.