Is it just me or are we abandoning more?
 

I can't honestly say I have been watching these stats for a long time but lately it sure seems like a LOT of assignments expire, sometimes close to 50% of an assigned range, expecially near the top end of LL testing.

My guess: People see the $100k and give it a try. After a few days of "hard work" they get bored with the lack of results and delete P95.

Maybe it is the I.G.G. that has previously been mentioned in other threads?

I still think that 10% of GIMPS users do 90% of the hard job... :smile:

Luigi

I think you guys are right about the majority of abandoned jobs. But I also see among my PCs some exponents get to 90% and then just get lost for some reason....I try to reassign but it's not easy sometimes.

[quote=petrw1;124858]I can't honestly say I have been watching these stats for a long time but lately it sure seems like a LOT of assignments expire, sometimes close to 50% of an assigned range, expecially near the top end of LL testing.[/quote]
I agree.
Wouldn't it be a better "hardware" test to complete a
doublecheck (<1/4 of the time), thereby having the
confirmation of the final residue and assisting the painfully
slow progress towards proving that "M40" is indeed the
40th Mersenne prime?

David

PS If a test is abandoned halfway through, I don't see why
someone else couldn't take it up from where it left off.

[quote=davieddy;124923]PS If a test is abandoned halfway through, I don't see why
someone else couldn't take it up from where it left off.[/quote]
Wouldn't that require the person who's picking it up to have access to the previous save files?

Here's what you could always do: Make a sticky thread in this PrimeNet forum (or even in the "New To Gimps? Start here!" forum, so that people don't have to register to post in it) where people could post their save files before abandoning a result--thus, anyone who's interested could take the partially-complete save file and run with it, and as such all of that work wouldn't be wasted. :smile:

[QUOTE=ahmerali;124906]But I also see among my PCs some exponents get to 90% and then just get lost for some reason....I try to reassign but it's not easy sometimes.[/QUOTE]

That´s curious. What do you exactly mean by "get lost for some reason"? Some connection problem that keeps the PCs from reporting progress to the server?

[quote=davieddy;124923]I agree.
Wouldn't it be a better "hardware" test to complete a
doublecheck (<1/4 of the time), thereby having the
confirmation of the final residue and assisting the painfully
slow progress towards proving that "M40" is indeed the
40th Mersenne prime?

David

PS If a test is abandoned halfway through, I don't see why
someone else couldn't take it up from where it left off.[/quote]

A doublecheck is I think indeed a good test of hardware and is directly contributing to the project at the same time, but it can take months for many users to complete so I don't think it should replace the current hardware test built into the software which takes only an hour or two. If someone has hardware problems serious enough to surface in a short test then they want to discover that quickly.

Personally I would be unwilling to resume someone else's started test because I wouldn't know that I could trust their hardware and with that uncertainty my motivation would be compromised. Still, if others would not mind doing it then I don't see why this possibility should not be available to them.

[QUOTE=davieddy;124923]PS If a test is abandoned halfway through, I don't see why someone else couldn't take it up from where it left off.[/QUOTE]

Could this be considered a form of legalized "poaching"? Look for the almost complete tests, grab the files, complete the test and get FULL CREDIT?

[quote=petrw1;124988]Could this be considered a form of legalized "poaching"? Look for the almost complete tests, grab the files, complete the test and get FULL CREDIT?[/quote]
Maybe there could be a requirement that all such tests must be completed under the anonymous account?

On second thought...then most people probably wouldn't want to do them. :smile:

[QUOTE=petrw1;124988]Could this be considered a form of legalized "poaching"? Look for the almost complete tests, grab the files, complete the test and get FULL CREDIT?[/QUOTE]

I don't think so. If the user leaves the exponent in the "unreserved" box, with its own savefile, the user gives you permission to complete it instead of wasting all the work done...

Luigi

Well this thread seems to have run off the rails into silly land rather quickly, so far be it for me to break the momentum ...

We should make P95 with a monitoring facility* that allows the server to watch the save files on the disk to check they are being updated at the proper intervals. This facility could also take the precautionary measure of downloading each new save file as it appears just in case the user suddenly decides to delete the whole thing. Then, when someone wants to take over the abandoned numbers the server can issue the latest save file along with the assignment.

Easy really. With each save file ~5meg and ~50000 users that's only 250gig.

* more commonly known as spyware.

A lot of people run prime95 just to check their new hardware. They are only stress testing, yet do not use the stress testing routine. Now that new computers have dual and quatro cores I would expect the number of dropped reservations to increase significant.

What has surprised me for some time is the sudden storm of reservations that take place quite regularly. Hundreds of reservations in just a few hours and then the regular lower pace for a couple of days, followed by just another such storm. Most of the reservations made during such a storm are dropped about 90 days or so later.

[QUOTE=Anonymous;124925]Wouldn't that require the person who's picking it up to have access to the previous save files?

Here's what you could always do: Make a sticky thread in this PrimeNet forum (or even in the "New To Gimps? Start here!" forum, so that people don't have to register to post in it) where people could post their save files before abandoning a result--thus, anyone who's interested could take the partially-complete save file and run with it, and as such all of that work wouldn't be wasted. :smile:[/QUOTE]

Well, then there would be a temptation to get easy double-checking credit by running the double-check from the save file as well. That would be bad for the project because the original machine could have made an error before the point of the save file; and neither the computer that finishes the first-time test nor the double-checker would catch that error.

[quote=jinydu;125056]Well, then there would be a temptation to get easy double-checking credit by running the double-check from the save file as well. That would be bad for the project because the original machine could have made an error before the point of the save file; and neither the computer that finishes the first-time test nor the double-checker would catch that error.[/quote]
Maybe the save file could be taken offline as soon as someone decides to work on it?

Or we could do a system like this: we could have an email address where users could send their abandoned exponents to, and then whoever mans that address can keep an updated list of available, partially completed exponents on the forum (maybe in the LMH or Mersenne-aries forum?). Users could then post in the thread to claim an exponent, and the person in charge of hanging on to all the save files would then email it to them.

Cumbersome, but it might work. :smile: (And no, before anyone asks, I'm not volunteering to keep track of all the save files. :wink:)

I think that if the test is resumed by downloading it from a hardware checking upload(I just can't come up with a good-sounding sentence to express this.) that we should consider it to have an increased likelyhood of having a bad residue. Maybe an increase of anywhere between 5-50%, or higher. So, that fact should be noted in the Sticky, if we were to do this.

In terms of storing the numbers, how about renting space and bandwidth? We could have a site where we could upload the numbers. We could flag stuff based on (1) the size of each uploaded file, (2) the name of each uploaded file, (3) an easily remembered password that could be found on Mersenne Forum but wouldn't be obvious if you accidentally showed up at the login page of upload/download site.

Storage and bandwidth is getting cheaper and cheaper nowadays, and the size of the Mersenne testing files could get smaller and smaller relative to the potential of broadband.