Downloading too many WUs to finish

[Beyond]

I'm running mostly ecm and if I increase the cache size to more than about 2 days far too many WUs will usually be downloaded to finish by the deadline (a far longer time). Currently running BOINC 6.12.28 but this has been happening for a long time. The estimated times for the WUs are actually quite high but they just keep downloading until work fetch is halted (happens on on all machines). Additionally the WUs go into "high priority" mode, run for a while and then a new WU is swapped in at "high priority". This continues until many WUs are suspended in order to drop out of "HP" mode. Is this because of a problem between the older server version and the newer clients? Currently I have so many WUs waiting that there is no way to complete them before the deadline. I suppose the overdue ones will have to be aborted. Hope that's not a problem?

[yoyo]

HEllo,
you can abort wus, which will not keep the deadline. As far as I know newer Boinc versions will do this automatically.
For me it sounds like a Boinc Client problem. The Client requests work for x seconds and the server delivers it. Even the problem, that many wus are started and than switched to an other one is a Client issue. This is also discussed in the Boinc development mailinglist. For this you can increase the time between wu switches. Default is 60 Minutes. You should increase it to e.g. 300 minutes.

yoyo

[Beyond]

HEllo,
you can abort wus, which will not keep the deadline. As far as I know newer Boinc versions will do this automatically.
For me it sounds like a Boinc Client problem. The Client requests work for x seconds and the server delivers it. Even the problem, that many wus are started and than switched to an other one is a Client issue. This is also discussed in the Boinc development mailinglist. For this you can increase the time between wu switches. Default is 60 Minutes. You should increase it to e.g. 300 minutes. yoyo

Thanks for the quick reply. I also thought it sounded like a client issue but it happens on no other projects (and I run a lot of projects on various machines). As mentioned above this has been going on for a long time (long before 6.12.xx). I have gradually increased the switching time and for quite a while the clients have been set to switch between applications every 600 minutes (not a typo, 10 hours). Ecm WUs are still getting switched out en mass long before completion. This can especially be a problem with the high memory usage of some types (even though I recently increased the memory to 8GB on a few boxes expressly for the purpose of running these large WUs). I saw some talk about a similar problem on the BOINC dev list a while ago but haven't noticed anything lately. It's not so big a deal as long as aborting WUs isn't a problem.

Thanks!
Beyond