deltavee, I hope you have a decent size cache (at least 1 day). I personally find that the server runs out of some type of WU at different times in the day, but they are always replenished. With a decent cache, you can crunch only CEP2.

I know Sekerob said that there are plenty of CEP2 WU in the feeder, so if enough people start crunching CEP2 (or the current ones increase their limits) it shouldn't be a problem, but I am not sure. The number of WU crunched per day barely budged yet the time per WU has gone down a lot. I know this is partially due to the limits people set, but they are people like you who set it to unlimited too.

Perhaps explained by several things:
New task production did not increase? But I doubt it, there is more likely to be a glut (only a few people observed an apparent shortage, out of how many that checked?).
Feeder balance, and changes taking time to implement.
Cache settings reduce the impact of WU return, spreading it out over 0 to 10 days. So charts would look raised rather than showing a jump.
And as you suggest, the project limits people set still mean many people are using the default (when opting in) of 1 task per system.
I think the main problem is that it's still an opt-in only project, so only the avid forum readers are even aware of it. Unless the WCG can select appropriate systems some how opt-in only cant be changed; it would mean some users getting tasks that might impede their systems functionality. Perhaps it could be tested on quad cores and up, for some opt into new project selectees that also have Beta selected? Emailing might also help.

Anyway, I have had a reasonably balanced return on the 4 sysetms I have running CEP2 (eash running at least one other project):
12 in 3 days for one system with a CEP2 task setting of 2 and running a mix of 2 other WCG projects including 3 non-WCG. CEP2 tasks vary between 5 and 11h on that i7.
My quad also picked up 12 in that time, but it's only running 2 WCG projects and has a setting of 1 task at a time only.
The other i7 got 7 tasks in that time, and it's running 4 CPU projects and only using 7 threads with a setting of 2 CEP2 tasks at a time. All 24/7 systems with low-ish cache levels.

I think perhaps you’re seeing something rather than missing something.
The system is certainly not perfect and as you say, allows you to have one task on the system, if that’s what you chose, but does not specify that the task will actually run. Boinc works that out from other tasks and priorities. So again I think you’re right, a reduced turn around would help.
anhhai, while a 1 day cache might prevent you running dry, it might also limit the tasks you have and the amount of time they actually spend running. The tasks on your system include running and queued tasks, so there is always likely to be more tasks than cores, but the CEP2 tasks won’t always be running.
Take a 1day cache on an i7-920 for example; with all 8 threads available, and the CEP2 option set to run no more than 4 tasks I think this means that you are limited to receiving 4 tasks, but your queue could have many more other tasks. Say you ran HCC and have an average task run time of 1.5h, you could have 120 HCC tasks in the queue. Boinc does not discern between WCG tasks so with all priorities being equal, CEP2 might not get much of a look in. Partially completed CEP2 tasks would prevent the download of new CEP2 tasks. I normally keep a 0.01 cache and I have set Boinc to switch applications to a high number (900h), so I don’t notice any such issues, and the other projects I run are of similar run times (HFCC).

So my guess is that if you keep a very low cache the actual amount of CEP2 tasks that you are running might be closer to what you want, but then you run the risk of not picking an even number up due to what is in the spooler at any given time. As you will not get more CEP2 tasks than you ask for, on average you will likely get less. If you try to overcompensate for this and say up the number of CEP2 tasks to have it’s probable that at some stage you will be running 4 tasks, but on average you will be running less than 4.
It would be more ideal if the system could allow us to always run the number of tasks you chose, but I think the limitations are more with Boinc here. I keep a very low cache and it seems to work out that I get a reasonable collection of tasks. The only sure way to do this is to separate WCG into projects individually recognised by Boinc as being different, but I guess that would be too difficult.

Are there any differences between the way Boinc 6.10.58 and the 6.12.12 trial version could handle allocation requests/caching/running that might be specifically beneficial to CEP2?

"I set my CEP2 limit to 11" - How do you do this?

I'm trying to focus three dual core computers on CEP2, but with only partial success. All are Intel Macs, all have done CEP2 units mixed with other WCG project tasks. Now it seems hard to keep at least one of them fully engaged on just CEP2 - a MacBook Pro, 2.4GHz Core 2 duo, 2GB SDRAM, OSX 10.5.8.

With 1 CEP2 task running, prefs set to use at most 15 GB of space, 80% (or 100%) of CPU, 1 (or 4) day cache, it will not load another CEP2 task, and the task running uses less than half of available processing power. I get this message:

Wed Feb 16 15:38:00 2011|World Community Grid|Sending scheduler request: Requested by user. Requesting 864783 seconds of work, reporting 0 completed tasks
Wed Feb 16 15:38:05 2011|World Community Grid|Scheduler request succeeded: got 0 new tasks
Wed Feb 16 15:38:05 2011|World Community Grid|Message from server: No work sent
Wed Feb 16 15:38:05 2011|World Community Grid|Message from server: No work available for the applications you have selected. Please check your settings on the website.

No tasks are downloaded, but if I make one change in the controlling profile, to opt in to other projects, and update again, I get 21 new tasks totaling about 96 hours of work, and it launches a c4cw task.

I do not see any other settings to tweak, and I do not understand why it is willing to download 96 hours of other work, but not a second CEP2 unit when I restrict activity to CEP2 only. This sort of thing has been happening for months, and accounts for discrepancy in the c4cw and CEP2 badges - c4cw gets a lot more spots in queues.

What can I do to focus these Macs' attention on CEP2, actually get 2 work units running at once, and occupy the bored Intel processors?

Thanks

Thanks, Paul, sorry to be dense!

Is there a summary of known disk activity susceptibilities somewhere?