I've got projects that have uneven work. This way it grabs a big bundle when it becomes available.

This is the reason I maintain a large cache, but I see @Dagorath explains it doesn't work that way. Interesting. I usually don't have an issue aborting tasks even with a 10 day cache.

However, having said that I had an odd experience recently. I have an older system and I set it to 10 days and ended up with over 500+ tasks. I don't remember if I've used the 10 day setting on it before so I'm not sure if this is a new issue for that computer. I did have to abort tasks and set the work buffer to 1 day after that.

Another computer is still set to 10 days and never has issues with aborting tasks. I wonder why that is?

Which brings me to a nagging (though slightly OT question). Is there a penalty for aborting tasks? Do projects keep track of cruncher performance and penalize ones that send work back by sending less projects to them? (Am I getting paranoid?)

And while I'm going OT, another performance question:

I've been fiddling with the "switch between applications" setting for some time now. I'm wondering if its more efficient to set that to a larger time block so that projects are completed in one session. Recently I set it to 360 minutes on my dual core thinking that it's more efficient to work on projects from start to finish in a single session.(I've been using 180 mins for a default for two of my cruncher boxes.)

One reason I did this is I noticed I had a lot of projects in the 75-99% level and I figured that just making sure that projects ran to completion would do a better job getting the results returned to the projects ASAP.

---

Ok, here's a couple more crazy theories...

The older computer in question that got the 500 tasks is used by someone other than me. So I've had to tweak the CPU settings, etc. so that user isn't bothered by too much fan noise and perceived slowdowns, etc. In reality, I found that BOINC suspends tasks very quickly based on settings, but they like to blame BOINC as the reason that their web pages loaded slowly...

So, here is my crazy theory. BOINC benchmarks the computer and uses that information when calculating project completion times. But does BOINC also factor in CPU settings or does it just refer to the benchmark data when it downloads new projects?

Even if the project manager is factoring in CPU settings, you'll have a hard time convincing me that actual user behavior gets factored in. That would depend on the types of applications being run, how often the computer is used, etc. So unless the project manager keeps some kind of activity log, user behavior is not being factored in. (Does the project manager keep track of how many hours a day the computer actually runs? etc.)

Of course, that doesn't explain how under any circumstances BOINC would think that computer could complete 500+ tasks. So there is a glitch somewhere.

On the issue of almost completed projects... some sort of AI that looks at an almost complete project and decided to complete it would be a neat problem for the programmers to solve. First, the project manager would have to have an accurate method of predicting completion times. Just having a "go for it" parameter that pushes through projects with 5% (or user set) to completion wouldn't be reliable. I've had large projects with 1% left to go that took over an hour to complete, while other projects can complete and upload in less than an hour.

BTW, if any of the BOINC programmers is listening into this thread my comments are not intended as criticism. I've found the project manager to be a stable, reliable program, so kudos!

---

One reason I upgraded my system memory on three systems was to speed up swap times in Windows. I run 4GB on the two main systems and 2GB on an older laptop. I've read conflicting versions of using the "leave applications in memory while suspended" feature. But as long as I have memory to burn, it can't hurt, right?

The fan running is the feature that annoys my team member the most, so I have to tweak CPU settings and select projects that minimize fan use on her computer. ;)

I looked into BOINC Tasks a few months ago, but I don't remember why I didn't implement it now. I use BAM! as a project manager for the team, so maybe that's why... *mumbles incoherently*

---

I'm running on 32bit systems here so 4GB is as good as it's gonna get for me. You also have a fancier bird than me! I have a Cockateil (who likes to pull his feathers out). He also likes to pull the keys off my laptop, which is an activity that I do not condone. ;)

---

(@Dagorath)...Projects allot you a certain number of tasks per day. When your computer reaches that quota you get no more tasks. Some projects set that quota so high there is no effective quota, other projects set it fairly low. One of the reasons for the quota is to cut off computers that crash one task after another and try to burn up 1,000 tasks per day...

This goes back to my question regarding penalties for aborted projects. Maybe I'm a little paranoid, but before I converted two of my computers to Linux, I aborted the remaining projects and sent them back. I figured this was better than just wiping the system. However in one case I did lose the OS and BOINC files before I could do that. So, if the projects track aborts and lost files... I worry that I'll get fewer tasks from them in the future. (@Peter, I'll check this on a couple project forums)

Since I recently completed BOINC re-installs I guess those computers have to re-build their usage history.

I also just started an experiment where I'll let two computers run 24/7 with a 360 minute interval for task switching.

---