Proposed DC assignment and recycle rules

[Prime95]

Originally Posted by cuBerBruce

It looks to me that the Cat 1/Cat 2 boundary might have been somewhere in that area at the time. Perhaps someone knows what the Cat 1/Cat 2 boundary was on April 3?

Primenet does record this information, but I am out of town until Saturday. I can look into the situation then. The expiration SQL statement could well have some bugs -- thanks for bringing this up.

In the meantime, I would finish the exponent. It is unfortunate this exponent will be triple-checked, but not a major disaster -- especially if we learn something from the event.

[TheMawn]

As a whole though, I must continue to stress who impressed I am with the new rules. The trailing edge looks so much neater and it makes us look more organized that it did.

If this is the first hiccup that anyone has noticed, then congratulations on a job well done.

[cuBerBruce]

Originally Posted by Prime95

In the meantime, I would finish the exponent. It is unfortunate this exponent will be triple-checked, but not a major disaster -- especially if we learn something from the event.

I appreciate your giving me your approval in going ahead with finishing the assignment. Thanks. I note that in moving it to my Sandy Bridge machine, it should be finished in a about two days. If Lazer13 queues up assignments 5 days in advance, I may have the assignment finished before his machine even gets started on it.

The assignment is worth a little over 35 GHz-days total, so all in all not that big of a deal if I were to have lost the work. (I will soon be reaching 10000 GHz-days for my lifetime stats.)

My other assignment that I transferred from that same machine to my Sandy Bridge has just finished. It appears to me now that this assignment would have stayed in the Cat 2 range even if I had left it on the slower machine, so it wasn't really necessary for me to move it. But I wanted to be sure I didn't have another situation like the one with M32062099 so soon.

I agree with TheMawn about overall how well the new assignment rules are working out as in allowing the trailing edge to advance.

[LaurV]

Just a thought: This assignement could not be in cat1 at the time of assignment, because in that case it would not be assigned to that machine, due to the queuing days rule.

A second thought: Prime95 does not abandon assignments which were already in progress, if the key become invalid. Only assignments which are not yet started are abandoned (and deleted from the worktodo file/queue), which is perfectly right, otherwise we will not be able to add "custom" (i.e. not legally assigned) assignments, P95 would delete them any time when it will connect to the server, and where the fun will be in that? If for example I want to test some 10M exponent (for whatever reason I may have) or steal someone's else assignment... hehe

[flagrantflowers]

Originally Posted by TheMawn

… I must continue to stress who impressed I am with the new rules. The trailing edge looks so much neater …

Is there a source that shows how many machines are authorized to perform cat first time LL tests?

How has this number changed since the rules were first implemented? I'm assuming that as more users try to make the times the trailing edge will start to pick up speed. How many computers were there initially? Two weeks ago? Today?

Still a lot of Anonymous assignments around six months old that won't get recycled for a while.

[cuBerBruce]

Originally Posted by Prime95

In the meantime, I would finish the exponent. It is unfortunate this exponent will be triple-checked, but not a major disaster -- especially if we learn something from the event.

My double-check of M32062099 has been (successfully) completed. Lazer13's machine had not reported starting it, though I can't be sure that it hasn't started it.

[flagrantflowers]

Originally Posted by cuBerBruce

My double-check of M32062099 has been (successfully) completed. Lazer13's machine had not reported starting it, though I can't be sure that it hasn't started it.

Well at this point it has become obvious that you have caused Lazer13 some kind of injury/confinement. Maybe you are actually the same person…

[chalsall]

Originally Posted by flagrantflowers

Well at this point it has become obvious that you have caused Lazer13 some kind of injury/confinement. Maybe you are actually the same person…

Actually, no. You are incorrect.

(Facing the "forth wall) Please notice the person with very little weight trying to assert themselves.

Laughter in your general direction is appropriate.

[flagrantflowers]

Originally Posted by chalsall

Actually, no. You are incorrect.

(Facing the "forth wall) Please notice the person with very little weight trying to assert themselves.

Laughter in your general direction is appropriate.

I feel that I have been officially welcomed to the forum now that you have chirped me; where's LaurV when you need them.

There should be a badge for this… You could probably force one and have it link to this thread. Egg all over my face.

[cuBerBruce]

I thought I would share some thoughts and information about the exponent that I felt was proceeding well when it became expired on me.

My guess is that the exponent I had (M32062099) was expired because of something to do with the exponent being low on the list of exponents not yet DC'ed along with either the overall performance of the machine or its progress on the assignment. It also seems to have something to do with the assignment reaching 60 days in age. I note that this machine does not appear to meet the "at least 2 results in the last 90 days for each LL worker thread" criterion for Cat 1 work, as its prior assignments were requiring in the neighborhood of 60 days per exponent per worker.

It is believed it was a Cat 2 exponent when assigned, meaning at least 1500 exponents were lower than it at that time (April 3). Primenet decided to expire the exponent around 60 days later, when it was roughly the 400th lowest exponent for DC work. At that time, the progress on the assignment was somewhere around 77-79%. It was about a day later I think when I noticed an assignment was missing on my assignments page.

Also, I note that the machine was reporting progress daily, except for I think 3 days during that time that it did not.

So anyway, it seems to me until George investigates this, one ahould not be assuming that if you get a Cat 2 DC assignment, that you will necessarily have 100 days to complete the assignment. If the Cat 1/Cat 2 boundary stays below the exponent, you're probably OK, as another exponent I had on that same machine was not recycled after 60 days, but it was still above the Cat 1/Cat 2 boundary.

I note that this machine now has two new assignments that are both Cat 4. So at least for myself, I should be safe from this issue happening again anytime soon.

[Prime95]

Yes, there is a bug in the SQL that expires cat 2 (and higher) assignments too soon. I'm working on a fix, but the SQL is rather ugly and needs optimizing.