Pending units
Pending units |
|
| log in |
Message boards : Number crunching : Pending units
Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 8 · Next
| Author | Message |
|---|---|
|
Probably the guys with the jumbomachines don't mind to keep a a big load of pendings. | |
ID: 1000354 ·  | |
Probably the guys with the jumbomachines don't mind to keep a a big load of pendings. I'm not sure that would work, as I and others will crunch 50 units in an hour. Plus many use multiple computers, which adds to the pendings. For my one rig, pendings are around 180,000+. I did get a few timed out units a couple of days ago, and they were the first I had ever seen. I carry a 6 day cache, but came within an hour of depleting my GPU units during the last unscheduled outage. The SETI crew got things working much faster than I was predicting, and my rig re-fed itself before the scheduled Tuesday outage. Steve ____________ Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website | |
| ID: 1000361 · | |
Probably the guys with the jumbomachines don't mind to keep a a big load of pendings. The present limits are just about right, if we have a few days where all the multibeam tasks are VHAR, shorties, then the top computers need about 2000 tasks/day to keep then going if they are on 24/7. | |
| ID: 1000364 · | |
If you want to see how something like this works, check out the Milkyway site. That project limits users to 6 WU per CPU core, so a 4 core system can have 24 at any one time. This means that fast GPU crunchers run out every hour or so. When the project goes down, no one has a cache (except those not processing on GPU), so everyone either goes idle or switches to backup, mainly Collatz for ATI users. Milkyway tends to have better up-time than Seti, and the admin there is pretty attentive, but there are still times my GPUs get cold... Just some info for you. -Dave [edit] Oh, and despite the limited amount of outstanding work, at any one time I still have way more than 4 pendings... Out of far fewer possible tasks. That would indicate that simply limiting tasks would not necessarily satisfy you. [Edit 2] Current pendings there, about 140 out of a total of about 680 tasks, I have 5 GPUs processing there. ____________  | |
| ID: 1000376 · | |
|
Pending : RAC now just under 8.5 - maybe, just maybe beginning to fall. | |
| ID: 1000581 · | |
Pending : RAC now just under 8.5 - maybe, just maybe beginning to fall. Correction. Ratio not falling at all - now up to 8.8 ____________   | |
| ID: 1000698 · | |
|
So after the latest outage pendings are way up to around 700,000 just need wingpersons to report now. LOL | |
| ID: 1001940 · | |
|
Just one idea! | |
| ID: 1005595 · | |
|
My Pending to RAC ratio is 8.58, which is the highest it has ever been. When these latest outages end, the pendings really pile up. Even when this weeks outage began, I had 241,000 credit pending, which was my high total. Is anyone else seeing huge pendings as a result of the outages? | |
| ID: 1016697 · | |
 James Sotherden Volunteer moderator  Send message Joined: 16 May 99 Posts: 6046 Credit: 14,616,626 RAC: 23,787  | |
|
Actualy my pendings have gone down since the 3 day outages. Was normally around 52,000 now down to 36,000. | |
| ID: 1016699 · | |
|
Same here. Before the Outages begun, i have 5.x Times RAC pending Credits. Now | |
| ID: 1016700 · | |
|
Pending credit: 197,197.10 | |
| ID: 1016701 · | |
|
Coming out of a 3 day outage, you upload and report. Your pendings will go 'way up until your wingmen report. On this outage, my pendings went from about 200K to 400K immediately after I reported all my WUs done during the outage. Now, I am down to < 300K pandings, since others returning their results have enabled some of mine to be validated. And so it goes. | |
| ID: 1016711 · | |
|
The Question is - can your wingmen report faster than you can crunch WU's ? ;-) | |
| ID: 1016714 · | |
|
The interesting thing is that my RAC is just about where it was before the outage. It is still increasing at the same rate, but my pendings are way larger than they were, so I know I got some work done. | |
| ID: 1016720 · | |
|
I was wondering about the rising # of pendings for many of us, and the capabilities of the system. I am assuming that the power crunchers for the most part are aware of and pay attention to these boards. This is true for a good number of moderate crunchers. I would think that the vast majority of users are set it and forget it. Has anything been tested on Beta that would include this crowd? What would be happening about ghosts, DCF, and all the other things that have affected us lately, and how many units are out there in that category? Could a large percentage of those be sitting dry? Is there a way to determine how many ghosts there are system wide other than when they time out? I was just thinking of possible reasons for the pendings to be up so high. As I crunch, my RAC is staying about the same, but my pendings continue to rise. It's just curiosity. | |
| ID: 1016877 · | |
|
I really think a lot of it is because so many people have increased their cache to the max. That and quite a few people are giving up because of all the outages. You could be right about the ghosts too. I've had to detach twice now because of them in the last couple of weeks. What makes me think it's increased caches though is that most of the detached WUs are still showing on my tasks page and haven't even been touched yet even though they have been a couple of weeks for some of them. | |
| ID: 1016896 · | |
|
Last time I checked I had 220+ ghost wu's, probably more. I am sorry for the wingmen who are waiting for their results, but before I do something stupid, I'd like to know a little more about the detach thing. Could someone explain how exactly you are getting rid of those ? Isn't detach mess up the history for the comp on the statistic page ? | |
| ID: 1016898 · | |
|
I'm running down my cache again, ready for this week's detach/reattach, ;-) | |
| ID: 1016899 · | |
|
Bearcat, | |
| ID: 1016903 · | |
Message boards : Number crunching : Pending units
| Copyright © 2013 University of California |
