I am about to start running BOINC projects on a computer lab that I am installing in a school that I work with. I have encountered two problems with BOINC CPDN that need to be addressed before I can put CPDN on this group of machines.

One of them, as I mentioned in my title, is sluggishness. Ordinarily, CPDN is quite happy co-existing with everything else and getting out of the way when it needs to. The one exception is when it is checkpointing. When it reaches the end of a \"day\", it checkpoints, and while it is doing that, mouse and keyboard input and screen updates slow right down. The machines I am going to be installing this on are AMD64 3200+ / Windows XP (32-bit) Pro, which should complete a timestep in about 3 seconds; with 48 timesteps in a \"day\", this means a checkpoint about every two and a half minutes. As a checkpoint takes about three seconds, this means that my students will, if I allow CPDN to run, have a three-second freeze in their work every two and a half minutes.

The other issue is the GPF errors. In another message I mentioned that I am getting GPF errors; what I did not mention is that these are coming up on the screen as dialog boxes \"sulphur_cycle.EXE\" / \"This program has caused an error 0xc0000005\". While these boxes are easy enough to close, they really should not be coming up at all -- the other BOINC projects am running, rosetta and SETI, do manage to trap these errors without putting up dialog boxes -- and I am concerned that they will cause my clients to think there is a problem with the computers. I don\'t know if the errors will occur in the new machines -- where I\'m seeing them is on an AMD64/3000+, Windows XP 64-bit, machine with sulphur_cycle 4.22 -- but it would be a problem for the students if it ever did show up, and could be a problem for me as well.

Are there any plans to deal with these two issues?

---

The models actually checkpoint every 3 model days, if that helps. At that time they write a large file which is where the hesitation comes in. The BBC coupled model project, right now, checkpoints every 6 model days, so the hesitation is less frequent, and I assume this will be carried over to the coupled model when it starts here at cpdn. Not sure if that will satisfy your concerns about that problem.

The models actually checkpoint every 3 model days, if that helps. ... The BBC coupled model project, right now, checkpoints every 6 model days, so the hesitation is less frequent... Not sure if that will satisfy your concerns about that problem.

Unfortunately, no. Granted, a freeze every 7.5 minutes is better than a freeze every 2.5 minutes, and presumably a freeze every 15 minutes would be easier still, but I\'m afraid that will not answer sufficiently for my purposes. My experience is that the checkpointing is enough to irritate me very much, and I can\'t expect the students at this school to be able to handle it any better than I can. Unless the checkpointing can be lowered in priority so as to not result in an apparent freeze, I will have to do something less invasive.

---

Hello CWangersky, I have a machine that closely matches those machines you wish to set up in your schools lab, an A64 3200+ with 32-bit windows XP, and have never experienced this sluggishness while it checkpoints.
There is sluggishness at the end of a phase of course, but that is to be expected.

How much ram do you have in your current system? I recommend at least 512 Mb, but 1 GB really helps a lot while running cpdn and other apps at the same time.

I wonder if what you see now could be pertaining to windows xp 64-bit, drivers, or perhaps faulty ram, too little amounts of ram, etc. Cpdn takes quite a bit of ram.

Edit: By the way my cpu is the old model with 1MB of L2 cache, don\'t know how the new 512kb versions react, if there\'s any noticeable difference.

---

As you\'re worried about responsiveness when the user is active you could always set up a separate venue preference with idle detection (or a restricted time period). Make sure you also set keep applications in memory so you don\'t lose work done since the previous checkpoint when work is automatically suspended.

---

"The ultimate test of a moral society is the kind of world that it leaves to its children." - Dietrich Bonhoeffer