Message boards :
Number crunching :
Pending Credit.
Message board moderation
Author | Message |
---|---|
vantage Send message Joined: 29 Feb 04 Posts: 25 Credit: 74,171 RAC: 0 |
Guess that i´m no´t the first to ask why it´s taking so long to get credit for the WU that i´ve been crunching the past weeks. Here is the amount of WU thats waiting for credit... 596008845 150080455 54.20 597235216 150645199 17.31 598969136 151419949 63.98 600663590 152197903 50.61 600766956 152067123 74.10 600832466 152107703 74.11 601267625 152468420 50.67 601701976 152667351 50.67 601997255 152806414 50.66 603041725 153285603 63.98 603203692 153347465 63.98 603411158 153455013 16.68 603502337 153498081 19.32 603698847 153590848 19.33 599840209 151825105 73.60 Pending credit: 743.20 How long do i have to wait...??? And Why..?? I´m no´t in a hurry, just want to know how things works... |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
A wu is copied and two copies are sent to unique accounts and a host within those accounts. Each one needs to come back before each can be compared to one another. If they both say "nearly" the same thing, or exactly the same thing, then they validate the wu and grant credit. You're just faster than your partner at returning them on those particular tasks. Just like you had, your "co cruncher" has the same amount of time (deadline) to return them. |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
In addition to Astro's explanation of why we need both initial results returned before credit is granted, there's another factor in the increased number and length of pending results. Previously the project was sending out more results by default than was really needed for the science for a number of reasons, all valid at one point. However the downside to doing that was it meant that any result a host returned after the quorum was formed had essentially been wasting their time and your money running the result, since the science was already done for the WU and all you got out of it was credit. What this means in practice is that unless you constantly keep buying faster and faster machines on a regular basis you quickly get to the point where it isn't worth running a host for the project at all, since it rarely is doing anything useful. Likewise, even with a fast modern host, if you run it for only a few hours a day, then even it might not be doing much useful science. The way it is currently, you know for a fact if your host returns a result within the deadline (except for a few rare cases) it has actually contributed to the science and not just wasted your hosts time and your money. IMHO, pending results are a small price to pay for that knowlege Alinator |
Greg Send message Joined: 22 Feb 01 Posts: 1 Credit: 2,909,910 RAC: 0 |
My pending credits continue to climb as well. Maybe my co-cruncher has shut down his/her crunching.......any way to tell who your partner(s) are ? |
Helli_retiered Send message Joined: 15 Dec 99 Posts: 707 Credit: 108,785,585 RAC: 0 |
The main problem is the enormous deadline of 3 Month! At the beginning of the MultiBeam WUs, many Users have problems to get WUs, many WUs stuck at Download state so they canceled these downloads. Now, we have to wait over 3 Month before these WUs where send out again. Helli A loooong time ago: First Credits after SETI@home Restart |
Keith Send message Joined: 19 May 99 Posts: 483 Credit: 938,268 RAC: 0 |
A wu is copied and two copies are sent to unique accounts and a host within those accounts. Each one needs to come back before each can be compared to one another. If they both say "nearly" the same thing, or exactly the same thing, then they validate the wu and grant credit. You're just faster than your partner at returning them on those particular tasks. Just like you had, your "co cruncher" has the same amount of time (deadline) to return them. Astro Surely all accounts are unique. A work unit is created and 2 tasks are created within that work unit. Both should be sent straight away to hosts within different accounts, but frequently only one is sent. Very often there is very considerable delay before the 2nd task is sent to another host host. This seems to be a major contribution to the extremely high volume of pending credits. Keith |
John Clark Send message Joined: 29 Sep 99 Posts: 16515 Credit: 4,418,829 RAC: 0 |
I am not sure whether my pending is stabelising, but it is just below 3 and a half times my current RAC It's good to be back amongst friends and colleagues |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
The main problem is the enormous deadline of 3 Month! Fortunately, not true. If they actually cancel the download and report back (which will happen automatically next time BOINC tries to fetch work), then the servers 'know' that the result isn't going to come back, and put it straight into the queue for re-issue. It's the people who switch off their machines in a huff, or who switch to another project without disconnecting from SETI first, that cause the 3-month waits in pending. @ Keith, As I've said in another thread, there is a delay between the sending of the 'odd' and 'even' results. It seems to have gone up to 16:40 or thereabouts since I posted yesterday - but that isn't a massive wait compared to the three months Helli was talking about - certainly not enough to account for the increase in pending. My suspicion is that increasing numbers of crunchers are increasing their caches to the maximum 10 days allowed - and with the longer general deadlines, more of those 10-day caches are being, if not filled, then at least getting fuller. |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
Keith, To insure scientific accuracy (and cut down on cheats as a side effect), they only send the copies to "unique accounts" (not just hosts) so no ONE user would have the ability to decide the outcome of a wu in regards to the science or credit claim. When it's copied into two they add two digits to the end of the wu name, a _ and a number. The first copy gets an _0, the second a _1. If the transitioner sends out more copies due to failure on the part of the first two, then the numbering scheme continues. The third sent is _2, 4th is _3, etc etc.. After the copies are made they're placed in one of two download servers, one for the odds and one for evens. It's a known and recognized problem (by management) that they aren't always in sync so there can be differences between when the odd one is sent and when the even is. In times (like lately) that other issues exacerbate the lag. Normally, one wouldn't even notice it. |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
... Clarification: The tasks are in odd/even feeder/scheduler pairs, so depending on which of the two schedulers your host reaches, assigned task(s) will be odd or even. The first part of the task which your BOINC client attempts is to download the WU from the single download server. Because there are both odd and even tasks for each WU, the download server is not split. Joe |
Sarge Send message Joined: 25 Aug 99 Posts: 12273 Credit: 8,569,109 RAC: 79 |
Perhaps another issue is, when one looks and sees their pending credit rising, that there is some fluctuation up and down and more importantly it is not the same set of results waiting for validation as seen the previous time one looked? Likelihood of this guess? Capitalize on this good fortune, one word can bring you round ... changes. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
Perhaps another issue is, when one looks and sees their pending credit rising, that there is some fluctuation up and down and more importantly it is not the same set of results waiting for validation as seen the previous time one looked? Likelihood of this guess? Correct on both counts. I've just looked at my list (who put the host IDs on there, and when? Nice. Now all we need is sorting and filtering, LOL). Note that they are in two sequences, each with the oldest at the top. The top 95% (in my case - YMMV) are the ones still waiting for the second quorum member (your 'wingman') to send their reply in. My oldest "pending" was sent to someone who downloaded a massive cache on 11 August, and then seems to have switched the machine off and gone on holiday: the results only started coming in again yesterday. So I would guess that the ones at the top of this section are going to stick there for a very long time, and so give you the impression that nothing's happening. The bottom ~5% are the ones where validation failed, and a new copy had to be sent out as a tie-breaker. Again, these will have to wait for whatever the turn-round time of the new wingman is. Again, nothing may happen for a long time, giving you the impression..... All the action is ~5% up from the bottom, at the end of the first section of the list. This is where the ones you report get added if you get them done before your wingman, and where, I suspect, a high proportion will be deleted again within hours (if not minutes) as wingmen with a similar turnround time to yours report in. Very busy down here - while I've been typing this, numbers 280 and 281 from my list of 305 have been validated, and two new ones from another machine have been added as numbers 226 and 227 - but I wouldn't have noticed if you hadn't asked the question and prompted me to take a couple of snapshots. Edit - 10 minutes later, and three more have been validated: 135, 136 and 137 on the list. Same wingman each time (thank you Michael Buckingham) - must have sent a batch in. |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
Yeah, it's a little weird the CBNC's get broken out of the overall sequence and stuffed at the bottom of the list. The important thing is if you leave your prefs alone long enough you reach equilibrium where the number going to pending is roughly the same as the ones completeing finally and dropping off the list. When that happens the number of results present and the amount credit they represent has no signinifcant effect on your RAC one way or the other. IOW it's almost like there's a 'phantom' host running on your account. The other effect is the more cached work you carry the fewer results you will have pending overall. Alinator |
Sarge Send message Joined: 25 Aug 99 Posts: 12273 Credit: 8,569,109 RAC: 79 |
You can even tell who the wingman is? MB has a similar RAC to yours. Capitalize on this good fortune, one word can bring you round ... changes. |
Odysseus Send message Joined: 26 Jul 99 Posts: 1808 Credit: 6,701,347 RAC: 6 |
You can even tell who the wingman is? Unless his computers are hidden, sure. On any of your Workunit pages, in the Computer column, click on a host ID that doesn’t belong to you: the first entry on the page that takes you to (the public page for that host) will either read “Anonymous†or be a link to the owner’s public account page. |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
In any event, if your trying to estimate how long the WU may stay pending it doesn't matter if the host is anonymous or not. You still get the info you need for that from the detailed host listing for your wingman. Alinator |
Sarge Send message Joined: 25 Aug 99 Posts: 12273 Credit: 8,569,109 RAC: 79 |
In any event, if your trying to estimate how long the WU may stay pending it doesn't matter if the host is anonymous or not. You still get the info you need for that from the detailed host listing for your wingman. Can I PM my wingmates for certain WUs and say hurry the heck up? [Just kidding!] Capitalize on this good fortune, one word can bring you round ... changes. |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
LOL... Well, it can't hurt to try, as long as you don't hold your breath waiting for the owner to respond. :-) Alinator |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
You can even tell who the wingman is? Sure, you see all sorts of weird and wonderful names out there. It reminds you how many SETI crunchers there are who never post in these boards, and, more importantly, probably never read them either. Just seeing Michael's name (or any name I recognised from the boards) was sufficiently unusual to be worth commenting on. |
Sarge Send message Joined: 25 Aug 99 Posts: 12273 Credit: 8,569,109 RAC: 79 |
You can even tell who the wingman is? Yes, Michael is known around here. Former moderator and I believe he also assists on the Help Line, via Skype ... ? Speaking of ... . Capitalize on this good fortune, one word can bring you round ... changes. |
©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.