Since Asteroids is my number one Boinc project, I like to keep most of my PCs (especially the quad cores) racing thru the Asteroids units. And I USED TO (heavy on the USED TO) keep about a week of work to do in case of any problems. Then I saw how work backed up in the validation area, and I could lose tasks due to Boinc timeouts - work seemed to be done usually just before timing out when the queue depths are so deep and there would be timeouts if the power went out for a day or 2 - even several times a day from storms. So.. I backed off to only have a .05 day reserve with no extra days, and everything - all task units - fly thru with no backlog. HOWEVER, sometimes I end up waiting for over a week for validation as others (lol) are doing what I used to do. For example, I wondered what had happened to a task waiting to validate, I looked for it, it was assigned validation to a 6 cpu computer that has an "in progress" level of 235 tasks, waited a week there, and when timed out, assigned to another computer with a several day turnaround time. Today I sit and wonder as I have 140 (sometimes nearly 200) tasks waiting for validation, while only a dozen "in progress" and all are actually being worked one. I would just ask that some people set their queue depths a little lower. This keeps tasks from timing out - which requires the unit to be re-assigned, and all the work will all get done at the same speed, and validation won't be such a bottleneck. Thnx. (OK... I opened up a can of worms.. - Let's hear it.)

---

Since Asteroids is my number one Boinc project, I like to keep most of my PCs (especially the quad cores) racing thru the Asteroids units. And I USED TO (heavy on the USED TO) keep about a week of work to do in case of any problems. Then I saw how work backed up in the validation area, and I could lose tasks due to Boinc timeouts - work seemed to be done usually just before timing out when the queue depths are so deep and there would be timeouts if the power went out for a day or 2 - even several times a day from storms. So.. I backed off to only have a .05 day reserve with no extra days, and everything - all task units - fly thru with no backlog. HOWEVER, sometimes I end up waiting for over a week for validation as others (lol) are doing what I used to do. For example, I wondered what had happened to a task waiting to validate, I looked for it, it was assigned validation to a 6 cpu computer that has an "in progress" level of 235 tasks, waited a week there, and when timed out, assigned to another computer with a several day turnaround time. Today I sit and wonder as I have 140 (sometimes nearly 200) tasks waiting for validation, while only a dozen "in progress" and all are actually being worked one. I would just ask that some people set their queue depths a little lower. This keeps tasks from timing out - which requires the unit to be re-assigned, and all the work will all get done at the same speed, and validation won't be such a bottleneck. Thnx. (OK... I opened up a can of worms.. - Let's hear it.)

Part of the problem is projects with down times, ie Seti is a biggie in this, since we can't set our caches for each project independently on any project, you must plan for outages at those projects that frequently, my understanding is it's twice a week for Seti, and set your cache higher. This means a secondary project, like Asteroids could be, still has a 10 day cache and the tasks get delayed to users like you and I that have a much smaller cache. I myself prefer a 0.75 day cache and then keep one or more projects set at a zero percentage setting, this means whenever my main project doesn't give me work my zero percentage project will keep me going thru the outage.