I'm sure it's not really just luck but rather the law of averages... :).
Let's assume that 5% of returned results suffer from cross-platform validation problems. My boxes return more than 60 results per day so I should see about 3 of these every day.
On average your machine returns a result every 9.2 days, so you could expect to see one such invalid about every six months :). S5R2 has been going for less than three months so you are not quite due for your first invalid just yet :).
Of course, choosing 5% was just a wild guess to make the point look good :).
Digressing for a moment, have you ever considered replacing your veteran PII with a slightly less veteran PIII or P4? I'm sure that all moderately sized cities around the planet would have auction houses which sell off surplus government and business computers for just a few Euros. You could have a lot of fun with something like a 1Gig PIII or a 1.6Gig P4 which would probably double or triple your production rate for about the same electricity cost. in fact if you were to select one of the quite efficient Tualatin PIIIs, you might be able to reduce your electricity consumption at the same time :).
Digressing for a moment, have you ever considered replacing your veteran PII with a slightly less veteran PIII or P4? I'm sure that all moderately sized cities around the planet would have auction houses which sell off surplus government and business computers for just a few Euros. You could have a lot of fun with something like a 1Gig PIII or a 1.6Gig P4 which would probably double or triple your production rate for about the same electricity cost. in fact if you were to select one of the quite efficient Tualatin PIIIs, you might be able to reduce your electricity consumption at the same time :).
It sort of disappeared with the prospect of having to wait 9 days to get a result :).
It's actually very depressing to watch the %done figure and see it moving up at only half the rate of CPU hours :). So I'm making a stand for Tullio's mental health and general well-being. He probably is risking severe stress by watching and waiting for such a long time to see the next 0.001% tick over on the progress meter :).
Thanks for your care about my mental health. But the Pentium II is the fastest box in my stable, which includes a Pentium I with Windows98SE, an AT@T Olivetti Unix PC (10 MHz!) and a TI-99/4A, so the PII is almost a Ferrari compared to the rest. Incidentally, I just installed an Audigy audio board on it, since I am using it to listen to classical music on the Rai3 radio channel, besides watching NASA shuttle missions with Mplayer compiled by me in my Linux environment,
Tullio
Thanks for your care about my mental health. But the Pentium II is the fastest box in my stable, which includes a Pentium I with Windows98SE, an AT@T Olivetti Unix PC (10 MHz!) and a TI-99/4A, so the PII is almost a Ferrari compared to the rest. Incidentally, I just installed an Audigy audio board on it, since I am using it to listen to classical music on the Rai3 radio channel, besides watching NASA shuttle missions with Mplayer compiled by me in my Linux environment,
Tullio
Just to get us off the nostalgia kick and back on topic :) here is some further information on that problem that we used to have way back in the dim dark ages which I think was called the cross-project validation problem :).
The new validator has been running for a couple of days now so I thought I'd do another quick survey of my Linux boxes. This time I looked at over 40 machines and around 200 results in total. I decided to limit the survey to those results that had been returned on or after July 6. This way I was looking at a mixture of both new validator and old validator processed results and more importantly the timeframe was such that it would be hard to imagine that very many results could have been processed and already deleted.
My finding was zero invalid or potentially invalid (CBNC) results.
I did find a total of 7 examples of CBNCs where the two results predated the survey period (often going well back into June or even May) where the "decider" or "multiple deciders" in some cases had not yet been completed. These 7 will be processed by the new validator eventually and it will be interesting to see if all three in each quorum finally get credit. I'll keep monitoring them.
By accident, I also found this example of what had been a CBNC but had now been finalised by the return of the decider. However, you will notice (if you are quick) that there is still a 4th result in progress even though there are three good results, all with credit allocated. To test your understanding of the whole process, see if you can work out why that 4th result was sent in the first place :).
Looking at the whole WU the third and fifth results were issued because the first and fourth ones respectively ran over the deadline and there was no quorum yet. The fourth one was issued because the second and third weren't strongly similar.
2, 3, and 4 all got credit because when the decider came back it matched one of the other two strongly and the remaining one was still within the error tolerance bands of the canonical result (IOW, it was still weakly similar to the canonical). Five will get credit if it's on time and at least weakly similar to the canonical.
I think that pretty much covers the post mortem for this somewhat messy WU! :-)
EDIT: Strewth!! Stone the crows!! He's added even more now!! :)
Yes, you are entirely correct. It's not immediately obvious that the 4th in the list exceeded the deadline because it's no longer red and it also got credit :)
Obviously the old validator must have been unhappy with 2nd and 3rd but the new validator had no such worries, which is very pleasing to see.
Here is another interesting one to test your wits on :). It's a totally different group of machines but once again there are a total of five results. My Linux result was #4 in the full list. My question is why were #2 and #3 both given no credit? :).
This post is not a set of trick questions like my previous two.
This is to report another piece of evidence that the transition from 4.17 to 4.24 did have a positive impact on the validation problem before the new validator was issued.
This WU shows three results with an interesting twist. The 1st result is from one of my Linux boxes. The 2nd result was crunched on a Windows box. Both results were issued early on June 28 just before the transition to 4.24. The Windows box missed the deadline and a third result (Win 4.24) was issued which has not yet returned. Meanwhile the Win 4.17 result has finally been returned and the new validator has turned the Linux/Win 4.17 pair into CBNC.
In other words, the combination of Linux/Win 4.17 has validation problems that the new validator doesn't cope with.
The interest will be when the 3rd result comes in. I think we will find that Linux/Win 4.24 will validate OK, leaving the 4.17 out in the cold.
Then again, the news item on the new validator states that the new validator version is only one part of the fix of the problem. I would not be surprised to see CBNCs also with the new validator, but much less frequent than with the old version.
RE: It may be just
)
Hi Tullio,
I'm sure it's not really just luck but rather the law of averages... :).
Let's assume that 5% of returned results suffer from cross-platform validation problems. My boxes return more than 60 results per day so I should see about 3 of these every day.
On average your machine returns a result every 9.2 days, so you could expect to see one such invalid about every six months :). S5R2 has been going for less than three months so you are not quite due for your first invalid just yet :).
Of course, choosing 5% was just a wild guess to make the point look good :).
Digressing for a moment, have you ever considered replacing your veteran PII with a slightly less veteran PIII or P4? I'm sure that all moderately sized cities around the planet would have auction houses which sell off surplus government and business computers for just a few Euros. You could have a lot of fun with something like a 1Gig PIII or a 1.6Gig P4 which would probably double or triple your production rate for about the same electricity cost. in fact if you were to select one of the quite efficient Tualatin PIIIs, you might be able to reduce your electricity consumption at the same time :).
Cheers,
Gary.
RE: RE: It may be just
)
Ah, Gary. Where's your sense of nostalgia?
RE: Ah, Gary. Where's
)
It sort of disappeared with the prospect of having to wait 9 days to get a result :).
It's actually very depressing to watch the %done figure and see it moving up at only half the rate of CPU hours :). So I'm making a stand for Tullio's mental health and general well-being. He probably is risking severe stress by watching and waiting for such a long time to see the next 0.001% tick over on the progress meter :).
Cheers,
Gary.
Thanks for your care about my
)
Thanks for your care about my mental health. But the Pentium II is the fastest box in my stable, which includes a Pentium I with Windows98SE, an AT@T Olivetti Unix PC (10 MHz!) and a TI-99/4A, so the PII is almost a Ferrari compared to the rest. Incidentally, I just installed an Audigy audio board on it, since I am using it to listen to classical music on the Rai3 radio channel, besides watching NASA shuttle missions with Mplayer compiled by me in my Linux environment,
Tullio
RE: Thanks for your care
)
Wow, this is plenty of nostalgia, I'd say :-).
BRM
Just to get us off the
)
Just to get us off the nostalgia kick and back on topic :) here is some further information on that problem that we used to have way back in the dim dark ages which I think was called the cross-project validation problem :).
The new validator has been running for a couple of days now so I thought I'd do another quick survey of my Linux boxes. This time I looked at over 40 machines and around 200 results in total. I decided to limit the survey to those results that had been returned on or after July 6. This way I was looking at a mixture of both new validator and old validator processed results and more importantly the timeframe was such that it would be hard to imagine that very many results could have been processed and already deleted.
My finding was zero invalid or potentially invalid (CBNC) results.
I did find a total of 7 examples of CBNCs where the two results predated the survey period (often going well back into June or even May) where the "decider" or "multiple deciders" in some cases had not yet been completed. These 7 will be processed by the new validator eventually and it will be interesting to see if all three in each quorum finally get credit. I'll keep monitoring them.
By accident, I also found this example of what had been a CBNC but had now been finalised by the return of the decider. However, you will notice (if you are quick) that there is still a 4th result in progress even though there are three good results, all with credit allocated. To test your understanding of the whole process, see if you can work out why that 4th result was sent in the first place :).
Cheers,
Gary.
Looking at the whole WU the
)
Looking at the whole WU the third and fifth results were issued because the first and fourth ones respectively ran over the deadline and there was no quorum yet. The fourth one was issued because the second and third weren't strongly similar.
2, 3, and 4 all got credit because when the decider came back it matched one of the other two strongly and the remaining one was still within the error tolerance bands of the canonical result (IOW, it was still weakly similar to the canonical). Five will get credit if it's on time and at least weakly similar to the canonical.
I think that pretty much covers the post mortem for this somewhat messy WU! :-)
Alinator
I saw you edit (add in) that
)
I saw you edit (add in) that last bit :).
EDIT: Strewth!! Stone the crows!! He's added even more now!! :)
Yes, you are entirely correct. It's not immediately obvious that the 4th in the list exceeded the deadline because it's no longer red and it also got credit :)
Obviously the old validator must have been unhappy with 2nd and 3rd but the new validator had no such worries, which is very pleasing to see.
Here is another interesting one to test your wits on :). It's a totally different group of machines but once again there are a total of five results. My Linux result was #4 in the full list. My question is why were #2 and #3 both given no credit? :).
Cheers,
Gary.
This post is not a set of
)
This post is not a set of trick questions like my previous two.
This is to report another piece of evidence that the transition from 4.17 to 4.24 did have a positive impact on the validation problem before the new validator was issued.
This WU shows three results with an interesting twist. The 1st result is from one of my Linux boxes. The 2nd result was crunched on a Windows box. Both results were issued early on June 28 just before the transition to 4.24. The Windows box missed the deadline and a third result (Win 4.24) was issued which has not yet returned. Meanwhile the Win 4.17 result has finally been returned and the new validator has turned the Linux/Win 4.17 pair into CBNC.
In other words, the combination of Linux/Win 4.17 has validation problems that the new validator doesn't cope with.
The interest will be when the 3rd result comes in. I think we will find that Linux/Win 4.24 will validate OK, leaving the 4.17 out in the cold.
Cheers,
Gary.
Interesting! Then again,
)
Interesting!
Then again, the news item on the new validator states that the new validator version is only one part of the fix of the problem. I would not be surprised to see CBNCs also with the new validator, but much less frequent than with the old version.
CU
BRM