Hi,

My system is overcommitted to SZTAKI Desktop Grid and has been running for 111+ hours. I realise they have had some problems with large WUs recently, but my problem is somewhat more troublesome.

Namely, the "To completion" value is INCREASING on SZTAKI with no change in Progress %. Also, when I force QMC and Climateprediction to run they exhibit the same behaviour.

I haven't been able to find any help so far in the forums, so forgive me if this is a known issue I just overlooked. Any help would be appreciated.

Cheers,
wlb

Have you posted this on the SDG forums? It could well be a problem with their science application.

---

Well, that doesn't explain why the same behaviour is being exhibited with QMC and Climateprediction. To me, that may point to the BOINC application itself.

Climate models are 2000+ hours each. This will affect the way BOINC handles other projects, more so if you keep 'fiddling' to force projects to run.

Climate models are 2000+ hours each. This will affect the way BOINC handles other projects, more so if you keep 'fiddling' to force projects to run.

5200+ hours and counting, actually!

I think that I have figured out what is happening, but it is not really a satisfactory solution/answer.

The workunits seem to be dramatically underestimating their CPU requirements, and so as they incrementally update (for SZTAKI this is taking about 12 hours to progress by 1%) they readjust their projected time remaining to a lower value. So in the span of 1 hour it will increase by, say, 5 minutes, and then at the end it will decrease by 5.5 minutes.

I'm happy that it seems to be moving along but I'm concerned that something funny is happening here, because the WUs are taking CONSIDERABLY longer than I am accustomed to.

Any thoughts/ideas/comments/anecdotes would be appreciated. Cheers!

I'm happy that it seems to be moving along but I'm concerned that something funny is happening here, because the WUs are taking CONSIDERABLY longer than I am accustomed to.

Any thoughts/ideas/comments/anecdotes would be appreciated. Cheers!

Does SZTAKI have some very, very, very large WU's?

---

I'm happy that it seems to be moving along but I'm concerned that something funny is happening here, because the WUs are taking CONSIDERABLY longer than I am accustomed to.

Any thoughts/ideas/comments/anecdotes would be appreciated. Cheers!

Does SZTAKI have some very, very, very large WU's?

While I was aware that SZTAKI had long WU issues, that link did have a semi-useful piece of info regarding poor To completion estimates:

To completion estimate issues

However that doesn't really explain why QMC and Climateprediction are also out of whack...it seems improbable that I am currently running the three projects that don't work, seeing as how E@H, S@H and WCG are currently running just fine on my Mac. Perhaps it's the recent client upgrade, or maybe it's just my PC?

Cheers,
wlb

However that doesn't really explain why QMC and Climateprediction are also out of whack...it seems improbable that I am currently running the three projects that don't work, seeing as how E@H, S@H and WCG are currently running just fine on my Mac. Perhaps it's the recent client upgrade, or maybe it's just my PC?

Depends on what client you upgraded from. If it was another 5.x.x series, the DCF shouldn't be affected.

But SDG is missing Adam, for a while now. I hope to lure Nightbird into here, he can explain better about the what and why there. I haven't been there in a long time, only today to check for explanatory threads.

CPDN changed how they calculate their credit per timestep. That shouldn't affect the already running science application, though.

So other things you may want to check is if something else takes up CPU cycles, so your science apps get less. If you have hard drive problems, is your page file still healthy? That sort of things.

---

However that doesn't really explain why QMC and Climateprediction are also out of whack...it seems improbable that I am currently running the three projects that don't work, seeing as how E@H, S@H and WCG are currently running just fine on my Mac. Perhaps it's the recent client upgrade, or maybe it's just my PC?

Depends on what client you upgraded from. If it was another 5.x.x series, the DCF shouldn't be affected.

But SDG is missing Adam, for a while now. I hope to lure Nightbird into here, he can explain better about the what and why there. I haven't been there in a long time, only today to check for explanatory threads.

CPDN changed how they calculate their credit per timestep. That shouldn't affect the already running science application, though.

So other things you may want to check is if something else takes up CPU cycles, so your science apps get less. If you have hard drive problems, is your page file still healthy? That sort of things.

Thanks for your feedback, Ageless Jord! I'm pretty sure that BOINC is getting quite nearly all the CPU time, and I don't think I've changed anything that would cause this error. However some drive maintenance may be long overdue! Thanks for the patience and suggestions.

Cheers,
wlb

...However that doesn't really explain why QMC and Climateprediction are also out of whack...it seems improbable that I am currently running the three projects that don't work, seeing as how E@H, S@H and WCG are currently running just fine on my Mac. Perhaps it's the recent client upgrade, or maybe it's just my PC?
...

For CPDN I'd recommend calculating time to completion by looking at the CPU hours so far compared to the % done - the result should be extremely accurate since the timesteps are so regular.

If you are getting erratic work unit times and also boinc benchmarks, then you may be experiencing thermal throttling, or perhaps AMD cool-&-quiet is turned on.

---

...However that doesn't really explain why QMC and Climateprediction are also out of whack...it seems improbable that I am currently running the three projects that don't work, seeing as how E@H, S@H and WCG are currently running just fine on my Mac. Perhaps it's the recent client upgrade, or maybe it's just my PC?
...

For CPDN I'd recommend calculating time to completion by looking at the CPU hours so far compared to the % done - the result should be extremely accurate since the timesteps are so regular.

If you are getting erratic work unit times and also boinc benchmarks, then you may be experiencing thermal throttling, or perhaps AMD cool-&-quiet is turned on.

Hi MikeMarsUK, thanks for pitching in! While my laptop does have Intel speedstep, it has been plugged into the wall for the whole time. The temp is also about 40 deg C below the factory limits.

The whole thing seems pretty screwy. As of late I have noticed that the SZTAKI run time IS decreasing, albeit very slowly. It updates in 0.1% increments and progresses by about 1% a day, and while the To completion meter is constantly increasing it eventually trims about 30 minutes off per update. The trick was that I had to actually observe the program do an update.

What a mess. Hopefully the estimates will be better in the future (and hopefully it isn't a problem with my PC!)

Thanks for everyone's help! Any more suggestions would be greatly appreciated.
B