Is it me, or have workunits gotten bigger in the last 10 days?
23 Aug 2007 12:47:04 UTC
Topic 193068
(moderation:
I have three computers running BOINC and Einstein@home. In the last 10 days, I've noticed that the run time for workunits on all three machines as gone way up.
The slowest of the three machines was running two workunits at a time, each one taking about 58 hours (some a little more, some a little less). About 10 days ago, it picked up two workunits that ran about 140 hours each. When it completed those, it picked up two more at 158 hours, which it is still working on. My other machines are experiencing similar slowdowns -- the machine that was running one at a time, 38 hours per (like clockwork) is all of a sudden running a 63 hour WU.
I'm not complaining about credit (it's been fine), and I don't need to know if it's workunit size or datafile size. I just need to make sure it's not due to a virus or something that I've infected the three machines with. Is there something E@H related that can account for this?
Thanx.
Is it me, or have workunits gotten bigger in the last 10 days?
)
you have gotten new datafiles (for example h1_0543.35_S5R2__387_S5R2c_2 on host 951241). The old one (h1_0489.85_S5R2__36_S5R2c_0) had WUs with 414 credits and the shorter run times, the new one will have WUs with longer run time and more credits per WU (I got WUs with more than 660 credits).
So it is in fact possible to get WUs which take 50% more time to complete....
Udo
RE: I have three computers
)
With Einstein, each host downloads a major datafile, then, typically, it runs a long succession of results carved out of that datafile which have extremely similar runtimes.
When your host runs out of work on that datafile, or for another reason gets assigned another one, you'll see it jump abruptly from one execution time to another, then stay there for weeks more.
In your case, your host 912035 has started running Work Units like
h1_0549.15_S5R2__417_S5R2c
with a requested credit of 666.72 cobblestones.
I think that is about the longest flavor available on the current run.
There does appear to be a bias on the way work is handed out that has saved a disproportionate share of the long ones to near the end.
Thanks, everyone, for your
)
Thanks, everyone, for your input.