Pre-empted

作者	消息
Kieron Walsh 发送消息已加入：2 Mar 00 贴子：72 积分：43,502,325 近期平均积分：112	消息 316711 - 发表于：25 May 2006, 19:31:44 UTC 最近的修改日期：25 May 2006, 19:33:26 UTC Thank goodness for this forum/thread, and my decision to read it before acting: I was on the verge of aborting a 'preempted' WU as I didn't know what this meant and assumed it had somehow gone wrong, luckily I decided to do a search on here first. Makes you wonder how many WU's have been needlessly aborted because of this misunderstanding? I know the WU's would get re-issued but it does mean that people are wasting several hours processing by aborting them and probably feeling pretty negatively towards SETI as a result. Anyway, thanks again for the words of wisdom on here: 8.5 hours saved in my case, feels much more positive! ID: 316711 ·

Steve Cressman 志愿者测试人员发送消息已加入：6 Jun 02 贴子：583 积分：65,644 近期平均积分：0	消息 310615 - 发表于：19 May 2006, 18:15:38 UTC - 回复消息 310321. this pre-empting is getting crazy.. presently have 6 preempted WU's out of 30. All WU's were downloaded less than 2 hours ago. and since it is a Pentium D, it crunches 2 at a time. what is crazy though, 2 of the preempted WU's have an earlier deadline than the ones it is working on right now? and I am only running SETI, no other projects. Don't worry, the scheduler will get around to doing them again. The scheduler has many rules to follow to make sure all work completes before their deadlines. Pre-Empted = Paused 98SE XP2500+ @ 2.1 GHz Boinc v5.8.8 And God said"Let there be light."But then the program crashed because he was trying to access the 'light' property of a NULL universe pointer. ID: 310615 ·

nutcase 志愿者测试人员发送消息已加入：13 Jun 05 贴子：19 积分：6,589,801 近期平均积分：0	消息 310321 - 发表于：19 May 2006, 11:46:29 UTC this pre-empting is getting crazy.. presently have 6 preempted WU's out of 30. All WU's were downloaded less than 2 hours ago. and since it is a Pentium D, it crunches 2 at a time. what is crazy though, 2 of the preempted WU's have an earlier deadline than the ones it is working on right now? and I am only running SETI, no other projects. ID: 310321 ·

Pooh Bear 27 志愿者测试人员发送消息已加入：14 Jul 03 贴子：3222 积分：4,603,826 近期平均积分：0	消息 310308 - 发表于：19 May 2006, 11:14:53 UTC - 回复消息 310298. chugging away on a p4 ht 3.0 ghz cpu?? Are you running more than one project? This will happen. Your machine can only do 2 results at a time, and the way the scheduler works, it will sometimes do 2 of the same project, even if set at 50/50. ID: 310308 ·

lynxtra 发送消息已加入：3 Sep 04 贴子：137 积分：273,636 近期平均积分：0	消息 310298 - 发表于：19 May 2006, 10:57:15 UTC - 回复消息 303245. Preempted means that a workunit has been stopped so that other workunits can proceed. It's (usually) a normal thing. Most likely you've set your cache too large, and your computer has been chugging away trying to get beat the deadlines. chugging away on a p4 ht 3.0 ghz cpu?? the truth is out there ID: 310298 ·

John McLeod VII 志愿者开发人员志愿者测试人员发送消息已加入：15 Jul 99 贴子：24806 积分：790,712 近期平均积分：0	消息 305582 - 发表于：14 May 2006, 20:38:30 UTC - 回复消息 305462. I just look for the one that is RUNNING and if it has not been running long, Abort it. That way the Preempted one will start back up. But why? If I ever have a WU that has been setting around long enough that it has to preempt another WU to prevent missing a deadline, then there is almost never any reason not to Abort it, as there has usually already been a quorum and credit granted and it would be a waste of my processors time to bother to process that WU and be the "useless forth" to report. There is almost no reason to abort it. Aborting it will insure that you don't get credit. Also, I don't know the current statistic, but when the project chose to send out four work units, they did so because the odds were good that they wouldn't get three results by the deadline. If you are in fact the third, then aborting insures that the project will have to carry that one WU around just that much longer. If I recall, at the time nearly 35% of the results never made back as complete, error free and valid. The Berkeley team has also stated that they were going to revisit this sometime after Enhanced was completely rolled out. If the failure rate (not returned + error + invalid) is low enough, then the initial send will be lowered to 3. (and no, I don't know what they would consider low enough). BOINC WIKI ID: 305582 ·

1mp0£173 志愿者测试人员发送消息已加入：3 Apr 99 贴子：8423 积分：356,897 近期平均积分：0	消息 305462 - 发表于：14 May 2006, 17:27:53 UTC - 回复消息 304735. I just look for the one that is RUNNING and if it has not been running long, Abort it. That way the Preempted one will start back up. But why? If I ever have a WU that has been setting around long enough that it has to preempt another WU to prevent missing a deadline, then there is almost never any reason not to Abort it, as there has usually already been a quorum and credit granted and it would be a waste of my processors time to bother to process that WU and be the "useless forth" to report. There is almost no reason to abort it. Aborting it will insure that you don't get credit. Also, I don't know the current statistic, but when the project chose to send out four work units, they did so because the odds were good that they wouldn't get three results by the deadline. If you are in fact the third, then aborting insures that the project will have to carry that one WU around just that much longer. ID: 305462 ·

Mray 发送消息已加入：16 May 99 贴子：124 积分：35,848,796 近期平均积分：23	消息 305316 - 发表于：14 May 2006, 13:54:29 UTC - 回复消息 304735. 最近的修改日期：14 May 2006, 13:54:46 UTC If I ever have a WU that has been setting around long enough that it has to preempt another WU to prevent missing a deadline, then there is almost never any reason not to Abort it, as there has usually already been a quorum and credit granted and it would be a waste of my processors time to bother to process that WU and be the "useless forth" to report. Sometimes I find a computer was off for a few days or a week or more and hit this condition. In that case I check the results page to see how many results were received so far. It's surprising how many only have one or two valid results (or none). So coming in towards the end of the deadline doesn't neccessarily mean you're the "useless fourth". ID: 305316 ·

Grant (SSSF) 志愿者测试人员发送消息已加入：19 Aug 99 贴子：13012 积分：208,696,464 近期平均积分：304	消息 305271 - 发表于：14 May 2006, 10:36:33 UTC - 回复消息 305242. I just look for the one that is RUNNING and if it has not been running long, Abort it. That way the Preempted one will start back up. But why? If I ever have a WU that has been setting around long enough that it has to preempt another WU to prevent missing a deadline, then there is almost never any reason not to Abort it, as there has usually already been a quorum and credit granted and it would be a waste of my processors time to bother to process that WU and be the "useless forth" to report. You will still get your credit for it. Aborting it means you've just blown the time that has been spent crunching it. I have a 4 day cache, sometimes i make it into the 3, other times it's the 4th. Every time i get the Credit due to me. Grant Darwin NT ID: 305271 ·

Eewec 发送消息已加入：28 Nov 05 贴子：19 积分：190,633 近期平均积分：0	消息 305270 - 发表于：14 May 2006, 10:35:12 UTC I have 2 WU's that pre-empted the unit I was working on, both WU's have now completed and I'll be the first to report for either workunit... looks like I now know why. ID: 305270 ·

Richard Haselgrove 志愿者测试人员发送消息已加入：4 Jul 99 贴子：14142 积分：200,643,578 近期平均积分：874	消息 305242 - 发表于：14 May 2006, 9:31:47 UTC - 回复消息 304735. I just look for the one that is RUNNING and if it has not been running long, Abort it. That way the Preempted one will start back up. But why? If I ever have a WU that has been setting around long enough that it has to preempt another WU to prevent missing a deadline, then there is almost never any reason not to Abort it, as there has usually already been a quorum and credit granted and it would be a waste of my processors time to bother to process that WU and be the "useless forth" to report. I don't know your cache/share settings, of course, but is that still so with Enhanced? We're now getting WUs with 4.34 day deadlines. If you run a 2 day cache (not previously regarded as too far over the top), those WUs will go into EDF just over 4 hours after downloading. You might still be the fourth, but I don't think it's as inevitable as you imply. ID: 305242 ·

Jack Gulley 发送消息已加入：4 Mar 03 贴子：423 积分：526,566 近期平均积分：0	消息 304735 - 发表于：14 May 2006, 5:53:54 UTC - 回复消息 303908. I just look for the one that is RUNNING and if it has not been running long, Abort it. That way the Preempted one will start back up. But why? If I ever have a WU that has been setting around long enough that it has to preempt another WU to prevent missing a deadline, then there is almost never any reason not to Abort it, as there has usually already been a quorum and credit granted and it would be a waste of my processors time to bother to process that WU and be the "useless forth" to report. ID: 304735 ·

1mp0£173 志愿者测试人员发送消息已加入：3 Apr 99 贴子：8423 积分：356,897 近期平均积分：0	消息 304632 - 发表于：14 May 2006, 2:42:34 UTC - 回复消息 304071. I am crunching for 10 projects using a 1 day cache (I increased it from its usual .1 to get few extra LHC while they are available). I currently have work from 8 of them on my slow system. I always have at least 7 WU pre-empted and that's how it's supposed to work. Short deadline projects are often given priority but my resource share are scrupulously followed in the long term. With more projects coming on line, we will see more projects with WU with different length of due date, as is now the case with Seti-enhance. Pre-empting of WU within a project is now to be expected. Boinc was design with this in mind and so far it is performing admirably by juggling all the needs of various projects, follow our resource share and still return WU on time. Sorry to sound like a cheerleader. I'm just a Boinc fan. ;) Never argue with anything that works. BOINC takes a whole bunch of conflicting requirements and does a really good job shuffling work to meet deadlines -- even when you go out of your way to make things more difficult. With my 0.1 day cache, I have a couple Einstein (the short ones), two SETI-Enhanced, one work unit from Beta, and one LHC. Three of them are pre-empted, but they'll be finished eventually. ID: 304632 ·

Aurora Borealis 志愿者测试人员发送消息已加入：14 Jan 01 贴子：3075 积分：5,631,463 近期平均积分：0	消息 304071 - 发表于：13 May 2006, 15:15:58 UTC I am crunching for 10 projects using a 1 day cache (I increased it from its usual .1 to get few extra LHC while they are available). I currently have work from 8 of them on my slow system. I always have at least 7 WU pre-empted and that's how it's supposed to work. Short deadline projects are often given priority but my resource share are scrupulously followed in the long term. With more projects coming on line, we will see more projects with WU with different length of due date, as is now the case with Seti-enhance. Pre-empting of WU within a project is now to be expected. Boinc was design with this in mind and so far it is performing admirably by juggling all the needs of various projects, follow our resource share and still return WU on time. Sorry to sound like a cheerleader. I'm just a Boinc fan. ;) Boinc V7.2.42 Win7 i5 3.33G 4GB, GTX470 ID: 304071 ·

Grant (SSSF) 志愿者测试人员发送消息已加入：19 Aug 99 贴子：13012 积分：208,696,464 近期平均积分：304	消息 303922 - 发表于：13 May 2006, 9:28:46 UTC - 回复消息 303908. I just look for the one that is RUNNING and if it has not been running long, Abort it. That way the Prempted one will start back up. But why? My thoughts exactly. Work Unit is due back soon, get it out of the way & do the one with the later due by date later. Been working for me with no problems at all. Grant Darwin NT ID: 303922 ·

Richard Haselgrove 志愿者测试人员发送消息已加入：4 Jul 99 贴子：14142 积分：200,643,578 近期平均积分：874	消息 303908 - 发表于：13 May 2006, 8:13:20 UTC - 回复消息 303900. I just look for the one that is RUNNING and if it has not been running long, Abort it. That way the Prempted one will start back up. But why? My machine (Trux CC and Crunch3r 5.12) seems to work through the WUs in the order they were issued by the scheduler. So it started on a slow one that it got a while ago, with a deadline sometime in June. In the meantime, it downloaded a mix of ARs for my 2-day cache, including some short ones with deadlines of 16 May. Oops, it said, that's too close for comfort - so it pre-empted the long, slow one, and nipped a short one in quickly to be on the safe side. I have every confidence that once the quickie is out of the way, it'll turn its attention back to the longer one. Or it might do a bit of the Einstein WU which is also currently pre-empted. Don't know, don't care: it's all perfectly normal. Just sit back and watch the fun. ID: 303908 ·

W-K 666 志愿者测试人员发送消息已加入：18 May 99 贴子：13875 积分：40,757,560 近期平均积分：67	消息 303906 - 发表于：13 May 2006, 8:07:57 UTC - 回复消息 303900. I just look for the one that is RUNNING and if it has not been running long, Abort it. That way the Prempted one will start back up. Why, this is the intended design of the scheduler, if you start a long task, then download a short task, at some point it may preempt the long task so that the short task can meet its deadline. Andy ID: 303906 ·

Jack Gulley 发送消息已加入：4 Mar 03 贴子：423 积分：526,566 近期平均积分：0	消息 303900 - 发表于：13 May 2006, 7:49:46 UTC I just look for the one that is RUNNING and if it has not been running long, Abort it. That way the Prempted one will start back up. ID: 303900 ·

Fuzzy Hollynoodles 志愿者测试人员发送消息已加入：3 Apr 99 贴子：9659 积分：251,998 近期平均积分：0	消息 303603 - 发表于：12 May 2006, 23:27:53 UTC - 回复消息 303442. 最近的修改日期：12 May 2006, 23:28:44 UTC What's even worse is I had Leave applications in memory while preempted? = NO So everything before today is lost! Today was just like starting over! No checkpoints created, huh?! That sucks! :-( EDIT: Reminds me of one of the bugs over at Rosetta and Ralph... "I'm trying to maintain a shred of dignity in this world." - Me ID: 303603 ·

Geek@Play 志愿者测试人员发送消息已加入：31 Jul 01 贴子：2467 积分：86,146,931 近期平均积分：0	消息 303453 - 发表于：12 May 2006, 18:08:56 UTC - 回复消息 303445. You make new friends every day, huh? :) You got that right! Boinc....Boinc....Boinc....Boinc.... ID: 303453 ·

©2020 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.