Pendings Tasks
6 Dec 2012 22:06:41 UTC
Topic 196671
(moderation:
I notice a high increase in my pendings WU, not they are in the range of 2487 WU and rissing, that´s is normal? Or i do something wrong in my configuration?
And when i go to the pendings credit page shows no info on there, displays just a title of the page/columns.
Pendings Tasks
)
Not everyone around here has a) your number of fast machines, b) your number of fast GPUs. So it's kind of normal. ;-)
I imagine that, but the
)
I imagine that, but the number increase very fast in the last days (for +/- 1500 in the weekend to more than 2400 today) and as the pendings page stops to work i imagine that could be some kind of problem in my configuration.
Thanks
I have two hosts running at
)
I have two hosts running at almost stable state with approximately 35000 daily cobblestone production, hugely dominated by GPU results (I allow only one CPU task, and I use Fred's TThrottle application to slow that one task down as the room temperature rises above 70F).
The requested work queue lengths for these two hosts are between two and three days, so perhaps just a little shorter than the average reply time for BRP4 work here currently (I assume your reply times, Juan, are much less, looks like less than a quarter of a day on a small recent sample).
One of mine currently has 116 tasks pending, the other 174. With your more productive hosts and shorter queues, you'd expect considerably more than that.
If you run much GW work, you'll get an additional element of variability from the very non-random assignment of quorum partners on that work. Einstein employs a scheduling scheme which can eliminate almost all repetitive WU download for GW work (still CPU-only, sadly), by re-using already downloaded files. But this means you will often find the exact same host paired with dozens of your WUs. If your partner responds faster than you, these will get nearly instant credit. If your partner walks away from the project without cancelling work, you wait for extra weeks. These two things happen anyway, but here on Einstein on GW work, they happen in clusters.
Thanks for the info, now i´m
)
Thanks for the info, now i´m quiet, i did nothing wrong... all seem´s normal.