Message boards :
Number crunching :
20 workunit limit
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 . . . 8 · Next
Author | Message |
---|---|
Send message Joined: 9 Nov 07 Posts: 131 Credit: 180,454 RAC: 0 |
|
Send message Joined: 28 Aug 07 Posts: 31 Credit: 86,152,236 RAC: 0 |
Perhaps wasting work, or even the lax deadlines aren't that much of a problem. The server could send out WUs in progress again to new users if it hasn't gotten results after a certain amount of time. The result that arrives later will simply be ignored, but credit still granted. Since computers are fickle and the WUs are short, this could mean the difference between a few hours and a few days... BOINC has feature to abort task that are no longer needed (quorum met for example). It is a server side feature, not client (like cancel those beyond deadline). A smart way would be to feed fast hosts with more work a slow onces with less amount. A fast host is the one with low turn-around time, slow one is the one that sents results back late. Note that slow one may a be quad that gets turned off or is overworked with other tasks meeting deadline. It is not only about FLOPs. EDIT: What is a general turn-around time per Wu? |
Send message Joined: 6 Sep 07 Posts: 1 Credit: 87,753 RAC: 0 |
Maybe if each WU carried work from a number of different genetic seeds and ran for a while longer? Seems like a good idea! Would it be possible for the project to be split into several "strands" starting from different genetic seeds, and these strands be processed side by side so that progress within a strand is slower and the time-window within which a returned result is still meaningful/useful is thereby extended? You could also distribute small batches of WUs with one WU coming from each strand. |
Send message Joined: 10 Feb 08 Posts: 32 Credit: 13,227,191 RAC: 0 |
If these WUs are time sensitive, then why do they have a 5 day deadline? The server has been out of work for nearly 24 hours now, and there are still over 5000 WUs out in the wind. Seems to me that if new work generation depends on the old results, the deadline should be shortened. Maybe try 12 hours and see how that goes, at least if the WU ends up on a duffer, it would time out and possibly get sent to a faster, more reliable host. I couldnt agree more - nothing that bothers me more then when I have to read that im computing for the trash. If the project needs a shoter deadline then thats the way it has to be, I think most will agree here. |
Send message Joined: 27 Aug 07 Posts: 647 Credit: 27,592,547 RAC: 0 |
I think the deadlines were increased just because we had some longer server outages before. ;-) When the servers went down for more than 12 hours it was no fun not being able to report finished results in time. *grin* EDIT: when the server is stable enough the deadlines can be decreased again of course, but 12 hours is really short if you participate in more than a few projects. Lovely greetings, Cori |
Send message Joined: 21 Nov 07 Posts: 6 Credit: 154,900 RAC: 0 |
I agree on this, but then it is simple - for those with numerous projects in which they are participating simultaniously Milky Way simply is the wrong project. The same for slower machines. Deadlines should be set to the need of the project's science, not the convienience of users. Science first. Greetings, Groundhog |
Send message Joined: 27 Aug 07 Posts: 647 Credit: 27,592,547 RAC: 0 |
Yup. But you'll always need at least one backup project when running out of WUs here. :-) Lovely greetings, Cori |
Send message Joined: 9 Nov 07 Posts: 131 Credit: 180,454 RAC: 0 |
I have several backup projects :) CLICK TO HELP BUILD |
Send message Joined: 27 Aug 07 Posts: 647 Credit: 27,592,547 RAC: 0 |
Me too. :-) That's why I thought a 12 hrs deadline might be a little tight. *grin* Lovely greetings, Cori |
Send message Joined: 18 Nov 07 Posts: 280 Credit: 2,442,757 RAC: 0 |
Me too. :-) That's why I thought a 12 hrs deadline might be a little tight. *grin* It would be nice if you could tell Boinc to do a certain amount of WUs before switching to another project (unless it fails to get any, of course), rather than just relying on a time share and deadlines :) I don't suppose there's an advanced option, or an up-to-date custom build that allows you to do that, is there? |
Send message Joined: 9 Nov 07 Posts: 131 Credit: 180,454 RAC: 0 |
Very true. Doesn't BOINC give WU's getting ready to expire priority over other WU's regardless of the project? CLICK TO HELP BUILD |
Send message Joined: 27 Aug 07 Posts: 647 Credit: 27,592,547 RAC: 0 |
It does. So-called "panic mode". *grin* But it would be bad if BOINC would go into constant panic mode and only crunch Milkyways and you don't get your other WUs finished. But I guess it's a bit early to figure out now what might be in some days/weeks. ;-) Lovely greetings, Cori |
Send message Joined: 17 Feb 08 Posts: 363 Credit: 258,227,990 RAC: 0 |
Sorry i don't agree here. It's all a personal preference if you'd like to attach to all available boinc projects on earth or wan't to crunch MW and leave only a few like 2 or 4 projects for backup. It all depends on what priority you give to MW... i for myself am ready to crunch only MW and ABC+TSP as backup. So, cut it down to 4h deadline... no just kidding 12h are fine :P Join Support science! Joinc Team BOINC United now! |
Send message Joined: 31 Aug 07 Posts: 66 Credit: 1,002,668 RAC: 0 |
It does. So-called "panic mode". *grin* Exactly the point I raised yesterday. The current computing model doesn't really suit BOINC, it relies on rapid reporting. Options: 1) Abandon BOINC and go stand-alone. 2) As 1 above, but farm out 'child' genetic threads to BOINC. 3) Make WUs a mix of parallel genetic seeds, increasing crunching time, but feeding more improved start points back into the matrix. I'm actually thinking that option 1 might be best. My slow host is turning over WUs in 5 hours. This is too slow. I will probably set this host to NNT as I suspect more 'old' science is jsut that ... OLD. Al. |
Send message Joined: 17 Feb 08 Posts: 363 Credit: 258,227,990 RAC: 0 |
.... to 1) no ...why ? ....one could use a different boinc client that reports results immediately and single out that std. boinc clients... now how about that one ? 5h return time ? what's that a PII@300 MHz ? or do you mean the time that pases befor your boinc client contacts the server ? Join Support science! Joinc Team BOINC United now! |
Send message Joined: 28 Aug 07 Posts: 31 Credit: 86,152,236 RAC: 0 |
My quads (stock speed) are doing 0.03 days of turn-around time. If there were no down times and 20 minutes limit for another scheduler connection, turn-around time would be ~8 minutes. That's it - 8 minutes or 0.006 days. I've raised question of several islands in GECCO2008 paper accepted, so far without answer. |
Send message Joined: 31 Aug 07 Posts: 66 Credit: 1,002,668 RAC: 0 |
5h return time ? what's that a PII@300 MHz ? or do you mean the time that pases befor your boinc client contacts the server ? Sorry, must have explained this on another thread that you missed. 1.8G machine, WU time 15m. 20WU max. Last WU in will take a minimum of 5h to get crunched and reported. FIFO and all that. So every WU that machine crunches will take 5h from arrival to reporting. Agreed? Solution: Set a buffer of 0.001 hours, so that only one WU is ever active. But that throttles the faster hosts. Also, given the 'reliability' of MW, that ain't a good strategy ;) Now, a BOINC client that can limit the number of WUs per project (say 2 for MW), give them priority over all else, but respect the deadlines of backup projects (and we need BU projects when MW is still flaky) to avoid starving other projects would be good. I'm really thinking that genetic models don't suit BOINC. BOINC is great for boring, grinding, tedious number-crunching, but evolving models won't work. BTW, this has nothing to d0o with MW. MW is one of my fave projects. But, as such, I'd rather see it run for the science than for the user. Al. |
Send message Joined: 8 Oct 07 Posts: 289 Credit: 3,690,838 RAC: 0 |
5h return time ? what's that a PII@300 MHz ? or do you mean the time that pases befor your boinc client contacts the server ? But Al ...all that being said and true ....Travis has also said we are beating the BlueGene supercomputer in both the amount of results returned and the short turnaround time so how can Boinc not be at least an adequate platform? |
Send message Joined: 27 Aug 07 Posts: 647 Credit: 27,592,547 RAC: 0 |
But Al ...all that being said and true ....Travis has also said we are beating the BlueGene supercomputer in both the amount of results returned and the short turnaround time so how can Boinc not be at least an adequate platform? I hope that BOINC is an adequate platform! I would certainly miss Milkyway if it was no BOINC project anymore. ;-) Btw: my turnaround time for 20 WUs is about 80 to 85 minutes on my new lappy. So a 12 hrs deadline would be no problem if there were almost no outages and mostly always WUs available. That way I'd be happy to crunch along Milkyway exclusively on that box 24/7. *grin* Lovely greetings, Cori |
Send message Joined: 29 Aug 07 Posts: 486 Credit: 576,548,048 RAC: 5,478 |
My turnaround is 50 Minutes for 20 WU's running 50/50 with another Project, less than 30 running 100% but then I invariably run out of work so I stay with 50/50 ... :) |
©2024 Astroinformatics Group