This is great news. The 5.6 version of BOINC will have a scheduler that waits until checkpoints to preempt applications such as Rosetta.

June 15, 2006
BOINC's CPU scheduler has become smarter: it now waits until an application has checkpointed to preempt it and remove it from memory. This can save significant CPU time, especially on multiprocessors. Technical details are here. The new scheduler is in the development version of BOINC, and will appear in the 5.6 release. Volunteer developer John McLeod VII was the driving force behind the new scheduler.

This means no lost work just because you took the defaults when you set everything up and do not keep applications in memory. This will result in more useful work per day from PCs that are not keeping applications in memory when preempted.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

I wonder what will happen with projects that don't have checkpoints in their application? Will it run the wu to completion "waiting" for a checkpoint that never comes?

uFluids, for example, does not checkpoint, and my last wu there was a 20+ hour beast.

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

Well, with the current approach, my understanding is that unless your computer is on for the 20 hrs straight, and unless no other project preempts it, that the work is thrown away when the uFluids thread is removed from memory (which would depend upon the General preference for whether it is removed)... so for a project like that, a new approach can't be any WORSE than the current approach.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

But the default is not to leave in memory, so a newbie downloading a first uF wu would not know that. So instead of resetting, it runs to completion...?

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

Sorry, I don't know any more than is available on the link. It refers to a number of scheduling changes, and to which has most recently reached a checkpoint. Either way, if it crunched straight through 20hrs of uFluids, then debt would be incurred, and paid back as deadlines and future scheduling allow. It wasn't clear to me if it would only check the preemption if applications are not retained in memory or not.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

I wonder what will happen with projects that don't have checkpoints in their application? Will it run the wu to completion "waiting" for a checkpoint that never comes?

uFluids, for example, does not checkpoint, and my last wu there was a 20+ hour beast.

That's a very good question I hope has been taken in to count when the designers spec'd it.

I think uFluids has checkpoints now, doesn't it?

---

Click here to join the #1 Aussie Alliance on Rosetta

I wonder what will happen with projects that don't have checkpoints in their application? Will it run the wu to completion "waiting" for a checkpoint that never comes?

uFluids, for example, does not checkpoint, and my last wu there was a 20+ hour beast.

That's a very good question I hope has been taken in to count when the designers spec'd it.

I think uFluids has checkpoints now, doesn't it?

If the app runs more than double or 1.5 times the switch time it will be forced out. I think they are still working on checkpoints at uFluids, but I havn't checked today so I may be behind.

---

BOINC WIKI

BOINCing since 2002/12/8