That's great. I was wondering who was going to come up with the mythical 99999 WU. I guess that proves that the batches can have a maximum of 100K WUs (but I don't know if all batches have that max). Thanks for sharing.

CJSL

Crunching until my ears fall off...

Then, judging by yesterdays results of 240,538 it should take at least 5 days to complete a batch. This is assuming that the number of replications is dropping as participating computers prove themselves reliable. Having said that, my computer failed on one work unit so I am back to square one.

My heavens !!!... not sure how 100K got devalued to 10K between my brain to my fingers. I'm going to have to investigate this strange occurrence. Oh, and thanks for pointing it out (post has been corrected).

CJSL

Crunching for a.... uh... for a.... darn, I forgot what I was going to say.

Actually, the batch numbers received have been incrementing at a rate of about 1.5-1.75 per day, it varies. Apart from the lone 485, it then fell back to the 450s, and now have on one machine 472-475. They arrive in quasi ordered fashion, at this time. There's also a gap from somewhere after 345 to 450. My BOINCTasks history is incomplete though to be sure what exactly got skipped.

As the official progress is stuck at 3 percent, to which it jumped right at the resume, have reason to think the batches are 100K are the breaks on progress counter. At any rate, some extrapolation suggests there's 15K batches, which I've been using as a base to compute the percentage on the various Start Here charts [3.2% now]. When the Research page has progress jump again, have a new calibration point to refine the estimate [requires looking every day, as not scrapable]. No one tells, no one seems to think we have a need / requirement to know how much work there is. Just the rounded percent that is strangely unmoveable when batches 0-297 was 2 percent, 298-485 is < under 2 percent, though the amount of computing years was lots more. Through batch 297 we did 151.7 years, through now about 472 we've done 429 years.

As for the long running tasks, there was a post by armstrdj yesterday, that says they're currently working to keep the long one's away from the grid. Maybe they sample run these on their own rigs, but given how variable even batch 333 was, hard to see how unless the techs have finally managed to figure out what is what in the task name coding.

From the big thumb.

Crowd Crunch On.

edit: The armstrdj post , in it's entirety:


The checkpoint interval uniformity is good news, makes estimating remaining time as of 12.5% checkpoint quite solid. The client seems to have too figured that out, except for that funny at 100%, when the longest I've seen it being stuck there -AND RUNNING- is 9 hours -AFTER- the 8th checkpoint at 99.99%. That was on the tablet though.

edit 2: Oh, and we had 340K results validating yesterday for 49.3 years. More members possibly opting in, with a fairly steady supply, albeit you have to set an exclusive right now ;D