Message boards :
Number crunching :
FEED ME MORE - FEED ME MORE!
Message board moderation
Author | Message |
---|---|
Cliff Harding Send message Joined: 18 Aug 99 Posts: 1432 Credit: 110,967,840 RAC: 67 |
If this old man's mind has not gone completely senile, the servers have been feeding my 2 hungry GTX750Ti FTW GPUs with a steady diet of AP WUs since we came back up from the weekly maintenance period on Tuesday. That's going on 5+ days now and I'm certainly not complaining and I hope I don't put a jinx on it. The question is - Do any of you old heads out there in the Setiverse remember the longest we've been served such a rich diet? I don't buy computers, I build them!! |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
In the past I seem to recall AP work being much more common. With the few AP only crunchers only mentioning running out of work every few months. - With the introduction of v7 of MB a large difference in credit between MB & AP lead more people to run AP only. - We now have a Nvidia AP app. So the crunchers that have beefy GPU systems can tear though the AP work faster than ever. - With people favoring AP work over MB the MB tasks are not getting burned through as quickly as they otherwise would. - Moving to the colo may also play a part in things. As they move the data for splitting to the servers and back over the network. I'm not sure if that is how it was done before or if they were just plugging drives into the system to dump the data. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Sleepy Send message Joined: 21 May 99 Posts: 219 Credit: 98,947,784 RAC: 28,360 |
The more tapes they put in one go, the longer they will last, but then the longer will be the starvation period for APs. The best, which depends on each computer crunching capabilities, would be that they put as many tapes so that with 200 WUs in cache you can survive the period during which the AP splitters are disabled while MB Wus are still being split. The more powerful your PC is, the lower number of tapes you would wish to be loaded at any one time if you want to have a continuos supply of AP WUs. Paradoxically, this present bonanza can mean a long starvation when the AP splitters will stop for medium to high capacity crunchers. Which may not be in line with your happyness. Though things are actually generally going much better (I would say very well) since colo transfer. All the best, Sleepy |
Cliff Harding Send message Joined: 18 Aug 99 Posts: 1432 Credit: 110,967,840 RAC: 67 |
Paradoxically, this present bonanza can mean a long starvation when the AP splitters will stop for medium to high capacity crunchers. Which may not be in line with your happyness. Though things are actually generally going much better (I would say very well) since colo transfer. Yeah, there is that dreadful drought that keeps popping up, but with the addition of my new toy, it make those periods more bearable. I don't buy computers, I build them!! |
betreger Send message Joined: 29 Jun 99 Posts: 11360 Credit: 29,581,041 RAC: 66 |
APs are not being split and it looks like it will be a long time before they start up again. |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
APs being much more available in the past was also do to the feeder proportions, as well. Used to be 97 MBs and 3 APs in every load of 100 tasks for the feeder. So that limited the number of APs that were available to the scheduler every 2 (?) seconds to being 3. So.. 10,000 APs in RTS queue would last a while. Now it's just kind of a free-for-all and they go fast. Plus all of the advancements in processing speed doesn't help, either. I still think if we cut the number of AP splitters down to one or two instead of 6, AP splitting would slow down closer to the speed of MB splitting and then we wouldn't end up with these huge backlogs of waiting for MB to catch up...but I guess the staff doesn't see a problem with the way things are and it seems to work fine..it just kind of annoys some of us crunchers, but it makes no difference to the science. *shrug* Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
Blurf Send message Joined: 2 Sep 06 Posts: 8962 Credit: 12,678,685 RAC: 0 |
Bah! You jinxed us.... :) |
betreger Send message Joined: 29 Jun 99 Posts: 11360 Credit: 29,581,041 RAC: 66 |
APs are gone now we have to live on resends or something else for a real long time. |
Darth Beaver Send message Joined: 20 Aug 99 Posts: 6728 Credit: 21,443,075 RAC: 3 |
Dag nab it just as 1 machine starts to show a real remaning time we run out of AP's grrr FEED US MORE ......HUNGRY .......HUNGRY.......HUNGRY |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
You can have more AP when you finish with all of your MB! SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Darth Beaver Send message Joined: 20 Aug 99 Posts: 6728 Credit: 21,443,075 RAC: 3 |
ops ! i'll have to hide me puters next time forgot everyone can see ....:):) |
Cliff Harding Send message Joined: 18 Aug 99 Posts: 1432 Credit: 110,967,840 RAC: 67 |
I still think if we cut the number of AP splitters down to one or two instead of 6, AP splitting would slow down closer to the speed of MB splitting and then we wouldn't end up with these huge backlogs of waiting for MB to catch up...but I guess the staff doesn't see a problem with the way things are and it seems to work fine..it just kind of annoys some of us crunchers, but it makes no difference to the science. *shrug* At the moment there are only 5 of the 7 MB splitters working with the other 2 being disabled and they have been for some time now. It would improve things some what if/when they were put back online. I don't buy computers, I build them!! |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
I still think if we cut the number of AP splitters down to one or two instead of 6, AP splitting would slow down closer to the speed of MB splitting and then we wouldn't end up with these huge backlogs of waiting for MB to catch up...but I guess the staff doesn't see a problem with the way things are and it seems to work fine..it just kind of annoys some of us crunchers, but it makes no difference to the science. *shrug* The 5 MB splitter are generally more than fast enough to keep up with demand. They actually cycle off every so often as the amount of work RTS is large enough. If there is an error it takes a bit of time to catch up. Some time ago, shortly after moving to the CoLo IIRC, it was mentioned that they limit the speed/amount of the splitters because they were hitting a disk i/o limit with the storage array. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Sleepy Send message Joined: 21 May 99 Posts: 219 Credit: 98,947,784 RAC: 28,360 |
Though I would find it some kind of elegant solution, I fear that many would start complaining about difficulties in downloading AP WUs, which would be generalised and source for headaches for the staff to manage. The way it is done now is a nuisance for those few of us going 100% AP and consistently checking the systems. We know what is happening, we may not like it entirely and complain once in a while, but that's it and everything goes on smoothly. In other words, this is a nuisance for few informed people. The way you suggest may end in problems to many more less informed people who would find themselves in unexpected problems. Therefore, I think the present solution is wiser. Though I always look with terror when I see 600 channels loaded at the same time! :-) ;-) My 2 cents. Sleepy |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
The simple answer to eliminate this problem is fix creditscrew. Then MB & AP work will "pay" the same number of "meaningless" credits and then, only then, the work balance MB vs AP will be restored since will make no diference to do any of both works. Actualy there are few who is working hard to try to do, you could follow them at albert@home. |
Sleepy Send message Joined: 21 May 99 Posts: 219 Credit: 98,947,784 RAC: 28,360 |
Dear Juan, I know from what has been written here what is being tested and tried to fix the credit madness. And my best wishes goes to them. This would obviously be the best and final thing. Granted. What I was discussing (and I think also the others) is what to do and what may be best in the meantime. Meantime that I fear will be rather long. Cheers! Sleepy |
kittyman Send message Joined: 9 Jul 00 Posts: 51468 Credit: 1,018,363,574 RAC: 1,004 |
The kitties have no complaints.... They feast on AP when it is available and then dutifully crunch away at the MB cache to help the servers towards the next batch of AP. It would be nice if the credits were normalized between the two types of work though. "Freedom is just Chaos, with better lighting." Alan Dean Foster |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
The kitties have no complaints.... AstroPulse v7 will probably do that. It is current being tested on Beta. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
arkayn Send message Joined: 14 May 99 Posts: 4438 Credit: 55,006,323 RAC: 0 |
The kitties have no complaints.... Most likely the wrong direction though. |
kittyman Send message Joined: 9 Jul 00 Posts: 51468 Credit: 1,018,363,574 RAC: 1,004 |
The kitties have no complaints.... That's the way I took it too. "Freedom is just Chaos, with better lighting." Alan Dean Foster |
©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.