Low credit
16 Sep 2005 10:49:10 UTC
Topic 189894
(moderation:
application All-sky pulsar search
created 13 Sep 2005 4:09:24 UTC
name w1_0251.5__0251.5_0.1_T03_S4hB
canonical result 8718216
granted credit 53.50
8718216 205254 13 Sep 2005 4:09:26 UTC 14 Sep 2005 5:57:02 UTC OSD 43,263.72 61.65 53.50
8718217 90912 13 Sep 2005 19:04:17 UTC 15 Sep 2005 6:08:43 UTC OSD 47,741.16 53.50 53.50
8718218 389415 14 Sep 2005 6:57:31 UTC 15 Sep 2005 16:08:29 UTC OSD 23,289.58 70.85 53.50
8718219 72090 14 Sep 2005 9:34:57 UTC 15 Sep 2005 3:40:19 UTC OSD 30,261.28 89.95 0.00
FAQ:
To determine this, if three or more results
are valid, the high and low credit claims
are dropped, and the remaining credit claims
are averaged.
lesson 2
less than 2 = 1
FAQ:
If less than two results are valid, the LOW
credit claimed is awarded.
Low credit
)
The FAQ is correct. Look at the RECEIVED TIMES of the different results. What happened was that when the first three results (NOT including the last result that claimed 70.85 credits) came in, they were validated. One result was invalid. Of the remaining two, credits granted were from the LOW result.
Then, after the canonical result was found, the four result (third VALID result) arrived, claiming 70.85 credits. It was awarded the credit previously determined by the other three results.
Bruce
Director, Einstein@Home
RE: RE: application All-s
)
The invalid result was sent at last.
The third valid result came in time.
The credit was to low, because
validating was to early.
Thomas
RE: The invalid result was
)
14 Sep 2005 5:57:02 UTC 61.65 53.50
15 Sep 2005 3:40:19 UTC 89.95 0.00
15 Sep 2005 6:08:43 UTC 53.50 53.50
15 Sep 2005 16:08:29 UTC 70.85 53.50
Reordered by arrival times. First WU arrives requests 61.65. Second WU arrives requests 89.95. Third WU arrives requests 53.50. A Quorum is present. Validator runs. First and third WUs match, second does not. Validator looks at 53.50 and 61.65 requested, awards 53.50, as per FAQ. Canonical result is created and sent to the database. First three results are flagged internally as validation-complete. Fourth WU finally arrives, requesting 70.85. Validator sees that it matches two other "good" results and awards it 53.50 as already determined. It is flagged validation-complete. Now entire WU is internally flagged delete-pending.
A Quorum is three. Not four. The validator runs when a quorum is present. A canonical result is assigned when two or more results match. If none of the first three had matched, then the validator would have waited the ten hours on the fourth result and run again. If it still didn't match, then a fifth would be sent out. The fourth result is "insurance", sent out with the first three simply because too high a percentage of results come back either past deadline or with computing errors, etc., and a fourth was being sent out a large number of times anyway, but after the first three were back or expired, delaying validation by days. In the rarer cases where a fifth result was required, it could be weeks.
Having large numbers of results sitting around waiting for a quorum takes up disk space and slows down the whole system. It's "cheaper" to send out the possibly-unneeded fourth result at the same time as the first three - the majority of the time, at the "whole project" level, it winds up giving faster validation and smaller number of results sitting around. Projects that see 99.9% "good returns" can do away with the fourth result.
IF the validator is backlogged and not able to "keep up", and IF the fourth result arrives before the validator gets to this WU (which hopefully means less than 10 hours later than the third), then it makes a four-result "quorum", and it's credit request is included in the calculation. Predictor's validator normally keeps up and validates the quorum within a few minutes of the third result arriving.
The FAQ is a "general overview". What it says is correct, but possibly not in extremely deep detail. For that kind of detail, you have to look at the Wiki.