Don't think this distribution change has any relation on the time gap problem. My duo laptop just completed a beta on Windows 7 with only 17 minutes loss on 11.5 hours and others also report very low IO slippage, so things are looking up.

WCG has though a prime directive: Don't cause for BOINC to impact usage of a computer... particularly if they run on business / office PCs, where there are tens of thousands actually contributing to the grid.

If hundreds of systems all tried to download a 200MB or 300MB around the same time, this will impact on businesses. Ditto for uploading. This is why it makes more sense to allow a few systems to run a full set of CEP2 tasks than all systems to run a one task. So clearly a different selection process is needed than 1 each. Might also need IP based restrictions for this project when it moves to Windows. Just limiting the number of connections to the feeder from any single IP would help business crunchers avoid network issues.
Of course such crunchers could just leave the CEP2 box unchecked, but then does the WCG want the work done or not?

I reacently read an article about "memcached " that it can reduce disk I/O by caching files that are frequently accessed in shared memory.

Techs, can this be implemented?

I've wondered since starting Linux in May about some things I've never seen happening in Windows such and overall performance grinding to a halt when downloading big ZIP files (not gz files) and stumbled on this: http://ask.slashdot.org/story/10/10/23/182825...cheduling-For-the-Desktop

The name of the patch is awkward, but exists for all different 2.6 kernels http://www.kernel.org/pub/linux/kernel/people/ck/patches/2.6/ . Anyone up to try?

Meantime, I've concluded that swappines=0 does nothing to help CEP2 and am going to do a full swapoff since it seems ALL the science apps just ignore the setting and set up portions in the VM from the start. The swapspace setting in BOINC set to 0 does nothing either. It simply is used to test during task fetch request if there is the minimum memory (RAM + VM= Limit) to at all execute a task.

Intresting. So far my current Linux device that's only crunching CEP2 has been doing well. Swap is turned on, I check it's usage every few days using "top" and it's been at 0k used with 2 GB RAM. Maybe I've got the right combination of hardware and software that works well together. I found this article Finding Performance Bottlenecks in Linux , will run some checks to see what performance load/wait/bottleneck I see and to what degree.