Message boards :
Number crunching :
IF "no work available" THEN resend ?
Message board moderation
Author | Message |
---|---|
Keith T. Send message Joined: 23 Aug 99 Posts: 962 Credit: 537,293 RAC: 9 |
I don't know how much load this would place on the schedulers and other servers, but with the current lack of new tasks, and the high number of lost tasks, could early resends be created. I would suggest resending tasks older than 10 or 14 days after they are first sent, when one or both of the replies have not been received. I am aware that the original task would have to be kept until the deadline had passed, but in many cases this will just time out. I know this will occasionally result in unnecessary duplication of tasks, but we used to have a replication of 3 or 4, so there would not be much waste. Comments and suggestions please. [edit] changed quorum to replication |
Francesco Forti Send message Joined: 24 May 00 Posts: 334 Credit: 204,421,005 RAC: 15 |
By my opinion, it's a GOOD idea. Francesco |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
I don't know how much load this would place on the schedulers and other servers, but with the current lack of new tasks, and the high number of lost tasks, could early resends be created. "Out of work" is caused by a number of different things. A common one is related to angle range: some process slowly, and some are "shorties." If there is a more than 10:1 ratio between "fast" and "slow" work, and the work creation rate is constant, it seems like we can consistently "outrun" the splitters if there is a run of shorties. I'm not sure what the limiting factor is, but if it's the feeder or the database, the odds are that reissues wouldn't be any better than new work. |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
There is another reason: If SETI started releasing "make work" work units to artificially keep us busy, that is volunteer CPU cycles that would not be available for other projects. It'd make those who crunch only SETI happier, but not for the multi-project crunchers. If you really want to stay busy, the answer is to find another project that is mildly interesting, and give them a 5% or 10% resource share. |
tullio Send message Joined: 9 Apr 04 Posts: 8797 Credit: 2,930,782 RAC: 1 |
I am running 5 projects with equal share (plus LHC) and I am never out of work. Recently, after upgrading my BOINC to 6.6.36 on Linux, Einstein and only Einstein insists that I will not finish in time (false). I then suspend all other projects and get Einstein work. One must beat the machine with brain. Tullio |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
I am running 5 projects with equal share (plus LHC) and I am never out of work. Recently, after upgrading my BOINC to 6.6.36 on Linux, Einstein and only Einstein insists that I will not finish in time (false). I then suspend all other projects and get Einstein work. One must beat the machine with brain. Tullio, Have you tried just letting it run? I think you might be surprised. -- Ned |
tullio Send message Joined: 9 Apr 04 Posts: 8797 Credit: 2,930,782 RAC: 1 |
I am still getting SETI work (MB).To Ned: yes, I've let it run but it has the same behavior. |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
ewver since i went to 6.6.36 on my mac milkyways says the same thing wont finish in time. Sop i just set aseti at NNT and when the cahe gets low enough i get milkyway tasks again. [/quote] Old James |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
I am still getting SETI work (MB).To Ned: yes, I've let it run but it has the same behavior. I suspect the "won't finish in time" is caused by the other work already in the cache, and at some point BOINC will let the cache for other projects run down. ... and at that point, it'll fetch the "won't finish" project. It may take a while for that to happen, however -- as in more than a week. |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
I am still getting SETI work (MB).To Ned: yes, I've let it run but it has the same behavior. It is caused by a bug in 6.6.36 that caused a value for estimated delay that was much too large. BOINC WIKI |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
I am still getting SETI work (MB).To Ned: yes, I've let it run but it has the same behavior. Ah, I'm not running anything that old. |
tullio Send message Joined: 9 Apr 04 Posts: 8797 Credit: 2,930,782 RAC: 1 |
I've got a 6.10.2 BOINC client and can easily install it. But is it reliable on Linux? Tullio |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
I've got a 6.10.2 BOINC client and can easily install it. But is it reliable on Linux? It goes crash during detection of some GPUs. If you are going to run Alpha test versions, you should really join the BOINC Alpha test email list. BOINC WIKI |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
I don't know how much load this would place on the schedulers and other servers, but with the current lack of new tasks, and the high number of lost tasks, could early resends be created. Adding a new feature like that to BOINC isn't easy, though of course it's technically feasible. To a certain extent it would mean hosts which are actually slow machines would be doing redundant processing every time the project runs short of work. Presumably fast hosts with large caches will be turning around work faster during such episodes, so would not be affected much. From the subject, I initially thought you were referring to the <resend_lost_results> project option. I'm hoping that mork might be able to support that mode, and when the project doesn't have much work the "no jobs available" reply to many hosts may be reducing database load anyhow. Certainly resending those ghosted WUs would make sense when possible. Joe |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
I don't know how much load this would place on the schedulers and other servers, but with the current lack of new tasks, and the high number of lost tasks, could early resends be created. Unfortunately, this runs into other projects that have work available. Part of the point of BOINC is for the client to pick up work from other projects when a project is out of work. Since there more than 50 active projects, everyone should be able to find some other projects that strike their fancy. BOINC WIKI |
Heflin Send message Joined: 22 Sep 99 Posts: 81 Credit: 640,242 RAC: 0 |
If you really want to stay busy, the answer is to find another project that is mildly interesting, and give them a 5% or 10% resource share. I try to setup other people's machines with their PRIMARY Project @ 100% Backup Project @ 10% All else empty Project @ 1% They never run out of work. And people seem to become interested in other projects as well. SETI@home since 1999 "Set it, and Forget it!" |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
If you really want to stay busy, the answer is to find another project that is mildly interesting, and give them a 5% or 10% resource share. In my opinion, you can give a project too low a resource share. The "all else, empty project" is going to get 80 hours a year, and on the rare occasions that it gets work, it'll go straight to "high priority." |
[AF>Libristes] erik Send message Joined: 30 Jul 07 Posts: 19 Credit: 4,016,114 RAC: 0 |
It's not the only bug in 6.6.36 for linux. This version is apparently incompatible with ubuntu 9.04 version. I run boinc 6.4.5 version again and it's running. I haven't any deadline problems because seti don't download tasks for my computer. Around one year ago, I crunshed astropulse taskes with success and now I don't have astropulse taskes whereas my computer is highly efficient. I actually crunsh another projects. just a poet |
tullio Send message Joined: 9 Apr 04 Posts: 8797 Credit: 2,930,782 RAC: 1 |
It seems to work on my SuSE Linux 10.3. It just downloaded an Einstein unit. All SETI units and QMC units are regularly downloaded after the preceding one has finished. But I must start the BOINC client and then the BOINC manager. The manager won't start the client. Tullio |
©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.