Message boards :
Number crunching :
minimum quorum: 2, initial replication: 3
Message board moderation
Author | Message |
---|---|
Idefix Send message Joined: 7 Sep 99 Posts: 154 Credit: 482,193 RAC: 0 |
Hello everybody, I've just noticed that the "minimum quorum" and the "initial replication" of a workunit has been reduced. (wu #123723790 is the first "new" one). Does SETI want a higher throughput or what is going on ...? ;-) I looked around a little bit, but I did'nt find any notice. Regards, Carsten |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
wow, where's the news on that tidbit. That's major |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
It was pre-announced in the Staff Blog: Heads Up: Quorum Change I agree, it's a slightly obscure/informal place for a technical announcement, but at least it was posted. It's worth everyone expanding their readling lists, and looking around just that little bit further - there's some interesting stuff in other places on these boards. |
Idefix Send message Joined: 7 Sep 99 Posts: 154 Credit: 482,193 RAC: 0 |
It was pre-announced in the Staff Blog: Heads Up: Quorum Change Thanks, Richard. I only checked the "Technical News" board because this was the place I was expecting the news ... ;-) Regards, Carsten |
keyboards Send message Joined: 14 Jul 00 Posts: 66 Credit: 492,766 RAC: 0 |
Keep in mind that with a quorum of 2, the lowest claimed valid credit will be granted (even if the higher is the canonical result). Hope you don't get paired with old 4.xx BOINC users since that manager doesn't count FPOPs (and usually has a large difference in credits claimed). !!Stupidity should be PAINFUL!! |
michael37 Send message Joined: 23 Jul 99 Posts: 311 Credit: 6,955,447 RAC: 0 |
Keep in mind that with a quorum of 2, the lowest claimed valid credit will be granted (even if the higher is the canonical result). Hope you don't get paired with old 4.xx BOINC users since that manager doesn't count FPOPs (and usually has a large difference in credits claimed). ARGH. Didn't take long for that to happen http://setiathome.berkeley.edu/workunit.php?wuid=123861880 |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
LOL.... Just wait until a few 64+ credit results get granted 10 and then see the howling start! ;-) Alinator |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
ARGH. A brand new (less than three weeks on BOINC) Core 2, running XP SP2. Rather kills the idea that all these 4.xx clients belong to the 'fit, forget, move on to a new job' brigade, doesn't it? |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
Hmmmm, refresh my memory here. Are there still proxy issues with 5x, or have they gotten that resolved? In any event 4.19 is a bit long in the tooth for a Win box. ;-) Alinator |
Odysseus Send message Joined: 26 Jul 99 Posts: 1808 Credit: 6,701,347 RAC: 6 |
Didn't take long for that to happen For the first time in several weeks, I’ve just noticed an ‘underclaim’: here my G5 Mac got paired with a Win XP system, running BOINC v4.45, that claimed 23% low. (Mine was the third result in, so I guess the validator had already compared the other two by then.) While the quorum reduction can be expected to make these events less rare than they used to be, I still doubt they’ll become common enough to make a significant difference over all. |
Jakob Creutzfeld Send message Joined: 13 Oct 00 Posts: 611 Credit: 2,025,000 RAC: 0 |
Didn't take long for that to happen Actually, even for the very first WU #123723790 mentioned in the OP it has been granted far too less. I guess (but don't hope), we'll see this happen more often than before (where we had a quorum of 3)... :-\\ Andy |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
LOL.... Just a thought, This change of minimum quorum & initial replication is partly in preparation for the introduction of the new multibeam data right? Will that mean that some of "ye olde 4.xx boinc clients" will drop off as the science app changes later? or will they get a new science app, and still keep on sending out of whack credit requests? anyone know this ? Jason "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Walla Send message Joined: 14 May 06 Posts: 329 Credit: 177,013 RAC: 0 |
Well this sucks I claimed 67 credit but I only got 38 http://setiathome.berkeley.edu/workunit.php?wuid=124067601 |
Anthony Apfelbeck Send message Joined: 26 May 06 Posts: 72 Credit: 17,425,405 RAC: 31 |
Well this sucks I claimed 67 credit but I only got 38 Is it possible to set up the system so that a quorum is declared to have been achieved only when 2 of the 3 initial replication claimed credits match in addition to the outcome being declared a success? With the above workunit as an example, the quorum would have only been achieved when the two 67 credit workunits reported and matched. Then all three successful results would be awarded the quorum decided credit of 67. Join the Fire Service Team |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
<edit> OOPS scratch that, what you were saying didn't sink in initially. :-O That would be one way to get around it, another is to just not use the claimed from a 4x'er if there is a 5x result in the quorum. Alinator |
Brian Silvers Send message Joined: 11 Jun 99 Posts: 1681 Credit: 492,052 RAC: 0 |
Is it possible to set up the system so that a quorum is declared to have been achieved only when 2 of the 3 initial replication claimed credits match in addition to the outcome being declared a success? On the surface, this would seem to be ok, but there are times when the claimed credit does not "match", but are only "strongly similar". A level of precision would have to be applied. For an example, see this workunit. My Pentium 4 is the bottom result. The two other results that are in now, while they do appear to claim identical amounts, that's because the precision listed is only 2 decimal places. The claimed amounts are: 62.3939049298803 (this is what was granted - standard app - result id 509255971) 62.3939047074736 (only the same out to the 6th decimal - standard app - result id 509255973) 62.3967432526592 (what my Pentium 4 claimed with the optimized app) While you could argue that this is a definite quibble, it does mean that someone will have to decide how many significant digits equates to a "match". Not only that, in the current replication scenario of 3, what if you alone were the 5.x result that got paired with two older apps? Beyond that, what if the two older apps, or all apps for that matter, claimed materially different amounts? Would the validation of the work then be held open for another result to be sent out in the hopes that there would then be a "match" amongst the now 4 (or more) results, thus defeating the purpose of reducing the replication and quorum amounts that led to this in the first place? |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
Except what the validator is comparing is the actual analysis output data. I don't think it even considers the claimed credit as a comparison criteria for validation purposes. Alinator |
Brian Silvers Send message Joined: 11 Jun 99 Posts: 1681 Credit: 492,052 RAC: 0 |
Except what the validator is comparing is the actual analysis output data. I don't think it even considers the claimed credit as a comparison criteria for validation purposes. I was referring to the proposed change by Anthony. His proposal was to not consider a quorum to have been met if there were not at least two matching credit claims. No quorum == no validation Edit: added clarity Edit2: spelling mistake - proposed |
Odysseus Send message Joined: 26 Jul 99 Posts: 1808 Credit: 6,701,347 RAC: 6 |
Just a thought, This change of minimum quorum & initial replication is partly in preparation for the introduction of the new multibeam data right? Will that mean that some of "ye olde 4.xx boinc clients" will drop off as the science app changes later? or will they get a new science app, and still keep on sending out of whack credit requests? anyone know this ? AFAICT there’s nothing in the science apps that requires a specific BOINC version; there are still a fair number of old clients crunching Multibeam WUs on S@h Beta. P.S. Before anyone corrects me ;) I should have said “… old clients managing the crunching of Multibeam WUs …â€Â. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Thanks, I thought it might be wishful thinking "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
©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.