Newer milestone thread

[Madpoo]

Currently, the next milestone with the smallest # of checks to reach it would be the # of first time checks up to M(57885161), with 6273 checks to go.

I'd love to give an ETA on that but there are 5470 exponents in there that haven't been assigned (or were assigned but have expired).

Of the 803 that are assigned, the last one to come in is 56309053 due @ 2016-07-02 06:53:39.230. But hey, it's already 0.6% done after being assigned 10 months ago, so... yeah, we'll get there.

(some of those exponent counts could be off a bit, I rushed through the queries to figure out what was assigned, how many are left, etc)

Methinks lots of folks are getting record breaking assignments, or whatever.

Now this would be interesting to see... some kind of distribution of 1st time LL assignments. What are folks up to?

[TheMawn]

A graphical distribution of assigned LL / DC tests would be cool. Different curves depending on rate of progress?

[Primeinator]

Now November 12th

Countdown to proving M(32582657) is the 44th Mersenne Prime: 15 (Estimated completion : 2014-11-12)

[cuBerBruce]

Madpoo keeps referring to estimated completion dates as "due" dates. I think of due dates as being when assignments must be completed to avoid getting recycled.

These estimated completion dates are simply dates that Prime95/mprime (or other software) calculates and reports to the server. These estimates can be highly inaccurate in some cases. When a user has had an assignment for over 250 days, is less than 70% done, has progressed less than 1 percentage point in the past two weeks, and estimates completing the assignment in only 17 more days, I have to be highly skeptical.

While I suppose it may be of some interest to know what completion date for a milestone will be based upon such estimated completion dates, I think such an estimated value has to be taken with a lot of grain of salt. Especially when we are dealing with a milestone being held up by these old assignments that are grandfathered from the new assignment/recycling rules. These grandfathered assignments should be history by the time we get close to the M45 milestone, fortunately.

[Madpoo]

Originally Posted by Madpoo

...
Methinks lots of folks are getting record breaking assignments, or whatever.

Now this would be interesting to see... some kind of distribution of 1st time LL assignments. What are folks up to?

I ran a little breakdown. Sorry of the formatting is out of whack in the list. These are first time LL assignments that haven't expired (i.e. they *should* be currently assigned, but not all have actually started yet):

exponent range	assignments
  50M-60M        1194
  60M-70M       45897
  70M-80M       32951
  80M-90M         128
 90M-100M           2
100M-110M          29
110M-120M           1
130M-140M          24
150M-160M           1
160M-170M          12
170M-180M           1
200M-300M         105
300M-400M        4461
400M-500M          27
500M-600M          91
600M-700M          25
700M-800M           3
800M-900M           1
900M-1000M         16

[Madpoo]

Originally Posted by cuBerBruce

Madpoo keeps referring to estimated completion dates as "due" dates. I think of due dates as being when assignments must be completed to avoid getting recycled....

Yeah, it's true, they're just the self-reported estimates for completion, but it's all we have to go on. I've been calling it the "ETA" informally which is a little more accurate... or ETC (estimated time of completion) but that's an uncommon abbreviation and ambiguous.

I could call it the "estimated completion date" like the column name in the database infers, but my fingers would get tired of typing that. :)

I did stick with "estimated completion" on that milestone page for M44 though since I new "due date" wouldn't quite be right. Even the estimated completion is likely going to be wrong since at least one of those in the list may have been abondoned by it's assignee "barbedwireisgood". It's only 22 days overdue though, so it's too soon to tell... but it has been a month since it checked in at all. I guess we'll see. Some eager beaver is likely to poach it if nothing else.

[Primeinator]

Originally Posted by Madpoo

Even the estimated completion is likely going to be wrong since at least one of those in the list may have been abondoned by it's assignee "barbedwireisgood".

On this exponent it says the next update is 9-26-2014; however, this has clearly already passed. Does this column denote when the computer was supposed to check in to the server?

[Madpoo]

Originally Posted by Primeinator

On this exponent it says the next update is 9-26-2014; however, this has clearly already passed. Does this column denote when the computer was supposed to check in to the server?

Yup, that's it exactly. I don't remember if it's the client that tells the server how often it'll be checking in, based on the config setting for that. I think that's how it works, so the server knows when to expect the client to check in again.

[chalsall]

Originally Posted by Madpoo

Yeah, it's true, they're just the self-reported estimates for completion, but it's all we have to go on.

Actually, it's NOT all we have to go on.

But, at the end of the day, it doesn't really matter all that much. History will present itself in its own time.

(I hope that's interpreted as being positive; it's meant to be.)

[Madpoo]

Originally Posted by chalsall

Actually, it's NOT all we have to go on.

But, at the end of the day, it doesn't really matter all that much. History will present itself in its own time.

(I hope that's interpreted as being positive; it's meant to be.)

Once a computer starts work, as it does it's regular check-ins it gives it's current % done, and also updates it's "due date". I guess someone/something else could look at the % done and, along with the rolling average, make an educated guess as to whether the estimate is any good, but that's actually what Prime95 itself is doing to come up with the estimate.

Now, when it comes to assignments that haven't started yet, or maybe an LL test that's still doing some pre-TF work or something, the estimates are probably not that good since it's basing it on when all prior work finishes up, which may or may not be any good.

I guess if you have a better idea how to get a more accurate "estimated completion" using available data, we can try to make some of the reports better. I was figuring that using the self-reported dates were probably the best thing to start from even if it turns out to be way off in the end. After all, there's no way to predict if a computer gets turned off over a weekend when it's normally up 24/7, or a computer gets used during the day for something particularly CPU heavy so Prime95 doesn't get as much CPU time as usual for a few hours, throwing off the estimates.

But generally I'd assume that over time the ETA will self-adjust and it's "good enough for government work" as the saying goes.

[Primeinator]

And the list shrinks...

Countdown to proving M(32582657) is the 44th Mersenne Prime: 14

I will be curious to see how many of these are abandoned by the users (if any). Some of them did not check in at their last expected date but there are a hundred reasons for this.