Message boards :
Number crunching :
why oh why
Message board moderation
Author | Message |
---|---|
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
I've noticed that the # of WU's I have pending has increased greatly as of late. When checking as to why that would be the case, I found several people had cached 100's of wu's that their reported cpu's couldn't possibly finish in a 2 week period. 2 2.8ghz p4 cpus just cannot process 566 WU's in that kind of time. each of this persons pc's would have to process 20+ WU's a day just to get these finished. Highly unlikely and most likely a better part of the WU's will end up returned unfinished at the deadline. Perhaps in the future BOINC will integrate a process that will prevent individuals from intentionally or inadvertently from D/Ling excess WU's that can't be finished. Perhaps a benchmark and a calculation prior to any D/L or better yet govern the boinc process to never allow more than a 2 week supply of WUs no matter how much somebody wants to have a huge cache. I can only guess that the people that are creating these huge caches used oldseti and dont realize the time restraints involved with BOINC |
ghstwolf Send message Joined: 14 Oct 04 Posts: 322 Credit: 55,806 RAC: 0 |
> Perhaps in the future BOINC will integrate a process that will prevent > individuals from intentionally or inadvertently from D/Ling excess WU's that > can't be finished. Perhaps a benchmark and a calculation prior to any D/L or > better yet govern the boinc process to never allow more than a 2 week supply > of WUs no matter how much somebody wants to have a huge cache. > > I can only guess that the people that are creating these huge caches used > oldseti and dont realize the time restraints involved with BOINC > A Benchmark... Brilliant. A limit of say 6 units (about 20 hours on a 1yo machine) cached per machine for the first 20 returned units, at that point open it up for a 7 day supply. Resent WU's should be restricted to machines using >1 day for caching. Waiting on credit is very annoying, and will be an obstical to growing the community. I do believe that some form of this is in the works Still looking for something profound or inspirational to place here. |
Richard M Send message Joined: 24 May 99 Posts: 64 Credit: 265,847 RAC: 0 |
Doing 27+ WU's/day. Created: 23 Nov 2004 10:47:16 UTC Total Credit: 14,512.41 Recent average credit: 670.45 CPU type: Intel(R) Xeon(TM) CPU 2.80GHz Pentium Number of CPUs: 4 Operating System: Microsoft Windows XP Professional Edition, Service Pack 2, (05.01.2600.00) Memory: 2047.27 MB Cache: 976.56 KB Measured floating point speed: 1214.68 million ops/sec Measured integer speed: 2241.75 million ops/sec ;) Click the Sig! BOINC Wiki |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
thats very nice but you'd need to divide you #'s in half since the guy mentioned only has 2 processors running. so if you are getting 27 a day from 4 xeon cpus then he might get 13-14. still a tad bit shy of getting anywhere near the 500+ that he's cached |
. Send message Joined: 3 Apr 99 Posts: 410 Credit: 16,559 RAC: 0 |
> I've noticed that the # of WU's I have pending has increased greatly as of > late. When checking as to why that would be the case, I found several people > had cached 100's of wu's that their reported cpu's couldn't possibly finish in > a 2 week period. 2 2.8ghz p4 cpus just cannot process 566 WU's in that kind > of time. each of this persons pc's would have to process 20+ WU's a day just > to get these finished. Highly unlikely and most likely a better part of the > WU's will end up returned unfinished at the deadline. > > Perhaps in the future BOINC will integrate a process that will prevent > individuals from intentionally or inadvertently from D/Ling excess WU's that > can't be finished. Perhaps a benchmark and a calculation prior to any D/L or > better yet govern the boinc process to never allow more than a 2 week supply > of WUs no matter how much somebody wants to have a huge cache. > > I can only guess that the people that are creating these huge caches used > oldseti and dont realize the time restraints involved with BOINC > Yes, this has irritated me also! I tried to make the staff aware of this problem and another problem in http://setiweb.ssl.berkeley.edu/forum_thread.php?id=7231 and got an answer from Rom about it. But the more people react on this, the higher chance it will be corrected. |
Seventh Serenity Send message Joined: 14 Dec 04 Posts: 17 Credit: 38,864 RAC: 0 |
I think a limit is in order, I only allow my systems to cache 2 WU's at a time, before it downloads another 2 so I don't hold up people with credit. The only problem is, there holding me and some one else up on credit due to there huge caches, and sometimes don't get the WU's sent back by the dead line resulting in us being delayed even further while it has to be processed by another system. |
STE\/E Send message Joined: 29 Mar 03 Posts: 1137 Credit: 5,334,063 RAC: 0 |
@ Scott I find it quit humorous that somebody that just Attached his computer to the Project on Dec 15, 2004 & has a Total of 0.36 Credits feels he's being held up ... Give me a break will you, you don't know the first thing about having your credits held up. I have a constant 350-400 Pending Wu's in my Account at all times and there are other people I'm sure with a lot more than that. Now thats being held up on the credits. Thats the way it is and we all just live with it. If anything it will only get worse once they shut down Seti Classic and all those people or how ever many of them finally jump over to the BOINC Platform. Thats when your going to see a big increase in your Pending Wu's because theres going to be so much Detaching & Resetting of the Project that it'll make your head swim. |
Seventh Serenity Send message Joined: 14 Dec 04 Posts: 17 Credit: 38,864 RAC: 0 |
> @ Scott > > I find it quit humorous that somebody that just Attached his computer to the > Project on Dec 15, 2004 & has a Total of 0.36 Credits feels he's being > held up ... Give me a break will you, you don't know the first thing about > having your credits held up. Excuse me, I have taken part in SETI@Home since July, I had to create a new account due to email address problems. I have taken part in LHC@Home with a build up of about 2000 credit, so don't you DARE start on me! |
STE\/E Send message Joined: 29 Mar 03 Posts: 1137 Credit: 5,334,063 RAC: 0 |
Start, Start, Start, Start, Stop ... ;) |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
> Resent WU's should be restricted to machines using >1 day for caching. > Waiting on credit is very annoying, and will be an obstical to growing the > community. I do believe that some form of this is in the works > Yes, it is in the works. The new "turnaround" value on your host's webpage is the beginning of the "prioritization" of WU issuing system. Goal 1: Resends will go to Turnaround systems at or near 1 day. This means even a slow 8 hour or 10 hour system could be given a resend, as long as their "connect every xx days" was 1 day. A resend, BTW, will (assuming everything validates ok) be granted credit almost as soon as it is returned. Goal 2: WUs will attempt to be issued to system with similar turnaround times. Thus all 3 recipients should return their results at approximately the same time, and get quick granting (again assuming good validate). Speed: You can get a Pentium IV 2.8Gig HT system to crunch 22-23 WUs a day (well I can anyway ;) sample 6500 to 8300 secs/WU) |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
excellent reply benher. What about the restriction on D/ls that exceed the timelimit now in place? This may help with what appears to be an increasing backlog of WU's not returned. Will the Oldseti crunchers understand the limiting factors used for caches and not attempt to get a months worth of WU's just so they wont have to connect. I realize you'd basically haveto slap every single about the head to get them to Not over indulge in caching. Perhaps a "new" client limit of 2-3 days to start and when they start returning things quickly let the door open for a larger cache. Seems possible for te process to take the clients state(new, unable to run, not returning WU's ontime) into account since BOINC checks the cpu benchmarks and routinely uploads and d/ls WU's |
Pascal, K G Send message Joined: 3 Apr 99 Posts: 2343 Credit: 150,491 RAC: 0 |
> Doing 27+ WU's/day. > > Created: 23 Nov 2004 10:47:16 UTC > > Total Credit: 14,512.41 > > Recent average credit: 670.45 > > CPU type: Intel(R) Xeon(TM) CPU 2.80GHz Pentium > > Number of CPUs: 4 > > Operating System: Microsoft Windows XP Professional > > Edition, Service Pack 2, (05.01.2600.00) > > Memory: 2047.27 MB > > Cache: 976.56 KB > > Measured floating point speed: 1214.68 million ops/sec > > Measured integer speed: 2241.75 million ops/sec > > ;) > > Last time I looked 27x14=378 Semper Eadem So long Paul, it has been a hell of a ride. Park your ego's, fire up the computers, Science YES, Credits No. |
Ingleside Send message Joined: 4 Feb 03 Posts: 1546 Credit: 15,832,022 RAC: 13 |
Just added to the cvs: David 16 Dec 2004 - User web: put a cap of 10 days on network connect period - User web: organize prefs page layout so that venues are grouped more clearly. Still needs some work. |
Steve Dundes Send message Joined: 6 Sep 04 Posts: 43 Credit: 159,057 RAC: 0 |
> Speed: You can get a Pentium IV 2.8Gig HT system to crunch 22-23 WUs a day > (well I can anyway ;) <a> href="http://setiweb.ssl.berkeley.edu/show_host_detail.php?hostid=316624">sample[/url] > 6500 to 8300 secs/WU) > > @Ben Since you brought it up, how are you able to crunch 22-23 WUs a day? I have a P4 2.8Gig with HT and I am only getting about 12-14 a day. Please let me in on your little secret. |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
>David 16 Dec 2004 >- User web: put a cap of 10 days on network connect period David (Anderson) also posted a new task to the taskbase: Task #1490 Secret - Well you can buy more blades and motherboards/cpus, or if you get bored you can learn SIMD assembly language as it works with C++ and rewrite time critical parts of Seti. |
Heaphus Send message Joined: 1 Apr 03 Posts: 96 Credit: 4,148,549 RAC: 0 |
> > Secret - Well you can buy more blades and motherboards/cpus, or if you get > bored you can learn SIMD assembly language as it works with C++ and rewrite > time critical parts of Seti. > Or, perhaps we could download what someone has apparently already done. ;-) |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
> > Doing 27+ WU's/day. > > > > Created: 23 Nov 2004 10:47:16 UTC > > > > Total Credit: 14,512.41 > > > > Recent average credit: 670.45 > > > > CPU type: Intel(R) Xeon(TM) CPU 2.80GHz Pentium > > > > Number of CPUs: 4 > > > > Operating System: Microsoft Windows XP Professional > > > > Edition, Service Pack 2, (05.01.2600.00) > > > > Memory: 2047.27 MB > > > > > > Cache: 976.56 KB > > > > Measured floating point speed: 1214.68 million ops/sec > > > > Measured integer speed: 2241.75 million ops/sec > > > > ;) > > > > > > > > Last time I looked 27x14=378 >please note AGAIN: the individual mentioned has 500+ WU's cached. so yes this person will in all likelihood NOT finish the previously mentioned WU's. Thank you for the quick multiplication though and if you will also note 500>378 making at least 122 WU's to be resent. that and can we guarantee this individual will keep his pc's on 24/7 for the next 2 weeks(14 day) |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
> > > Speed: You can get a Pentium IV 2.8Gig HT system to crunch 22-23 WUs a > day > > (well I can anyway ;) <a> > href="http://setiweb.ssl.berkeley.edu/show_host_detail.php?hostid=316624">sample[/url] > > 6500 to 8300 secs/WU) > > > > > @Ben > Since you brought it up, how are you able to crunch 22-23 WUs a day? I have a > P4 2.8Gig with HT and I am only getting about 12-14 a day. Please let me in on > your little secret. >I believe he is getting that # using xeon procesors |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
> >I believe he is getting that # using xeon procesors Nope :) We've posted the source on sourceforge. |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
> > >I believe he is getting that # using xeon procesors > > Nope :) > > We've posted the source on sourceforge. >ummmm.... huh? and thanks for the link or explanation to what you mean? |
©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.