ecm - long workunits, deadline too short

[jjwhalen]

ecm workunits have always been highly variable in the length of time to process.

In the past two days I have received (and aborted) a number of ecm WUs with a 5 day deadline and Estimated Time to Completion of >100 hours, which guarantees they will run in Panic Mode for all or most of their duration. One example had ETC of 131 hours on a fast machine, which actually exceeds the deadline by 11 hours.

For those of us who crunch multiple projects this sort of scheduling, overriding the project's Resource Share, is unacceptable. What's the hurry? Long workunits must have proportionally longer deadlines. I've turned off ecm in my Preferences until this issue is resolved.

[yoyo]

The time to completion is calculated by Boinc. This seems to be somhow wrong on your computer. The longest WU in the last 24h for ecm was 40h. You can see this here http://www.rechenkraft.net/yoyo/server_status.php on the right side.
yoyo

[jjwhalen]

The time to completion is calculated by Boinc. This seems to be somhow wrong on your computer. The longest WU in the last 24h for ecm was 40h. You can see this here http://www.rechenkraft.net/yoyo/server_status.php on the right side.
yoyo

True, ETC is calculated by the BOINC client based on benchmarks. But reporting deadline is set at the server. One could ask why a task requiring 40 CPUhours (your number) has a 5 day deadline instead of (for example) 10 days or 2 weeks. Much as I like Yoyo, you are still only an 8% Resource Share on my systems. I expect this is true for many of your loyal crunchers.

Sorry, nothing wrong with my computers (all 3). I've been crunching ecm for months without problem, Panic Mode or deadline overrun. I haven't reset the project, so ETC weighting factors in the XML shouldn't have changed. Nor have my processors benchmarked slower recently. ETC for Euler remains about the same. I've deselected ecm and (re)selected OGR & Muon, without problem. The other thirteen BOINC projects I crunch are all running fine. Only the ETC numbers for Yoyo/ecm got drastically larger on 3 different computers at about the same time.

BTW I monitored ecm tasks running on 2 of the 3 hosts (a coreduo_x64 @ 2.26GHz and a quad_x64 @ 2.66GHz): Progress % per hour, running in Panic Mode, was roughly consistent with the ETC at download. This was at a CPUcore load of ~90%. Which I can only interpret as: these tasks really were quite long for a 5 day reporting deadline.

However based on your reply I will reset the project, re-enable ecm and see what happens.

Best wishes.