No windows work?

I can see linux crunching is being tested now, but was surprised to get "work committed to other platforms" throughout the weekend, and now into Monday. Any time frame for more Windows work?

This is a linux machine:

Of course 5 minutes after I posted that, I just got work :)

And I forgot about the vacation, thanks for the reminder Saenger

Perhaps it really helped ;)

A bit later here:

Is there no work for windows again, or just not for P3s? I keep getting "work for other platforms" message on my P3s. Thx

This will help: http://docking.utep.edu/forum_thread.php?id=8&nowrap=true#852
____________

I'm a volunteer participant; my views are not necessarily those of Docking@Home or its participating institutions.

thx

10/18/2006 10:32:23 AM|Docking@Home|Requesting 8640 seconds of new work
10/18/2006 10:32:28 AM|Docking@Home|Scheduler request succeeded
10/18/2006 10:32:28 AM|Docking@Home|Message from server: No work sent
10/18/2006 10:32:28 AM|Docking@Home|Message from server: (there was work but it was committed to other platforms)

It's a problem with homogeneous redundancy that we are currently working on. It means that there are many results in the shared memory reserved by linux machines and macs. I'm surprised that windows machines don't get work, because it is usually the macs and linux machines that are suffering!

Thanks
Andre

10/18/2006 10:32:23 AM|Docking@Home|Requesting 8640 seconds of new work
10/18/2006 10:32:28 AM|Docking@Home|Scheduler request succeeded
10/18/2006 10:32:28 AM|Docking@Home|Message from server: No work sent
10/18/2006 10:32:28 AM|Docking@Home|Message from server: (there was work but it was committed to other platforms)

Server status shows 851 work units waiting, but none for windows apparently.

Look in this thread , it's not only Windoze. :(

Won't other workunits be unavailable for fixing this issue? I hope some workunits will be added on the database manually...I'll stay tune though.
____________

I'm a volunteer participant; my views are not necessarily those of Docking@Home or its participating institutions.

It has nothing to do with the database or the number of workunits unfortunately (there are more than enough unsent results). The problem is with the shared memory segment of the boinc server and I think the fact that our workunits are fairly short at the moment. I can try making this better by making the workunits longer and by making the shared memory segment larger.

Andre

Won't other workunits be unavailable for fixing this issue? I hope some workunits will be added on the database manually...I'll stay tune though.

Something happened:

It has nothing to do with the database or the number of workunits unfortunately (there are more than enough unsent results). The problem is with the shared memory segment of the boinc server and I think the fact that our workunits are fairly short at the moment. I can try making this better by making the workunits longer and by making the shared memory segment larger.

Andre

I hadn't seen it in awhile until today, but I was still able to download a few WUs before I got the notorious message. I believe Andre had said that thier fix would bring temporary relief, but that the problem would eventually creep back up on us. I think the permanent solution is still in the works.

Atomic

This should be fixed with the solution we got from the World Community Grid people. If it still comes up a lot, please let us know.

Andre

I hadn't seen it in awhile until today, but I was still able to download a few WUs before I got the notorious message. I believe Andre had said that thier fix would bring temporary relief, but that the problem would eventually creep back up on us. I think the permanent solution is still in the works.

Atomic

Well, indeed my windows machine got results, but the number of results is TOO MANY! Actually it has to crunch 7/24 so that every result will be computed... Isn't deadline too strict?

suguruhirahara
____________

I'm a volunteer participant; my views are not necessarily those of Docking@Home or its participating institutions.

My PC also d/l a bunch that if the estimate is accurate has my machine crunching 24/7 to finish the work. It doesn't look like the WUs are any shorter either. (Don't think that I am complaining too much espicially since I didn't have any work for the past few days). Could the deadline be made a little longer in the future? Einstein gives a couple of weeks for a deadline.

Anyway I will do my best to complete these.
____________

My PC also d/l a bunch that if the estimate is accurate has my machine crunching 24/7 to finish the work. It doesn't look like the WUs are any shorter either. (Don't think that I am complaining too much espicially since I didn't have any work for the past few days). Could the deadline be made a little longer in the future? Einstein gives a couple of weeks for a deadline.

Anyway I will do my best to complete these.

I'll check the deadlines. I don't understand yet why some people get so many replicas when they connect to the project, except that the preference for this is maybe too high. How much work does your client ask for? My client asks for 8640 seconds of work (the default) and I get 1 workunit (as it should be). Just be aware that our workunits will always have fairly short deadlines (couple of days) because the workunits themselves are fairly short.

Thanks!
Andre

Well, indeed my windows machine got results, but the number of results is TOO MANY! Actually it has to crunch 7/24 so that every result will be computed... Isn't deadline too strict?

suguruhirahara

Deadline is set to 5 days. That should be more than enough time for a 4 to 15 hours workunit I hope.

Thanks
Andre
____________
D@H the greatest project in the world... a while from now!

Hi Andre,

I'll check the deadlines. I don't understand yet why some people get so many replicas when they connect to the project, except that the preference for this is maybe too high. How much work does your client ask for? My client asks for 8640 seconds of work (the default) and I get 1 workunit (as it should be). Just be aware that our workunits will always have fairly short deadlines (couple of days) because the workunits themselves are fairly short.

I agree and I think this is really an issue that has to be discussed on the boinc email lists. If people are connected to multiple projects and are not connected all the time, the client should be 'smart' enough not to download work to fill all the way up to the deadline, because the chances that all the results will be crunched in that time are not very good...

I've increased the deadline to 10 days; let's see if that makes a difference.

Andre

Hi Andre,

I'll check the deadlines. I don't understand yet why some people get so many replicas when they connect to the project, except that the preference for this is maybe too high. How much work does your client ask for? My client asks for 8640 seconds of work (the default) and I get 1 workunit (as it should be). Just be aware that our workunits will always have fairly short deadlines (couple of days) because the workunits themselves are fairly short.

I set every 7 days (every week) to connect to network from General preferences, and then the client requested 1209600 seconds of new work. As a result the client downloaded 25 results, each of which requires around 7 hours to complete. With dualcore CPU(presler), 12.5*7 = 87.5 hours = 3.64 day is required to complete crunching all of the results.

But, this also means that to complete all of them, because the results need to be crunched in 4 days (due to the length of deadline), simply the machine cannot stop crunching for more than half a day. Umm... There should be more room for setting deadline, shouldn't it? As acmefrog noticed, other projects such as einstein@home or spinhenge@h give several weeks for the deadline.

If many hosts were available for all day long, this wouldn't be a problem. But the fact is that some participants turn off their hosts when they go out / sleep / travel. Using mobile pc is one of the examples. Normally it's not always available.

Actually with a view of these participants the function needs adjustment, because it is "those hosts" that need the function. To tell the truth, hosts which are always available don't absolutely need it, do they? IMO when adjusting a deadline, one should concern the unavailability of these hosts.

Edit: thanks for quick responce to the issue, Andre:D

thanks for reading,
suguruhirahara

Ref: http://boinc.berkeley.edu/sched.php

Need some time for me to read and understand it...

I can certainly understand the need for some breathing room between downloads and deadlines, especially considering some of our beloved BOINC bugs. Also, as Suguru pointed out, many users do not have their systems on all the time, or only have an intermittent internet connection.

I don't think that the deadlines should be TOO long though, as that can lead to other issues, like the "forever and a day" waiting period for validation that so many of our Mac (and some Linux) testers are familiar with. Not only can this be annoying, it can delay the observation, notification, and response to problems that may crop up. Personally, I think 10 days is a little too generous for a 6 hr WU, I feel that 7 days would be more than sufficient to account for the circumstances listed above. I know that a lot of users (particularly avid ones like me who are drawn to new projects!) tend to run multiple projects on the same host, and that is one of the beautiful things about BOINC - but it is a personal choice, and I don't believe that is a good enough reason on its own to push back deadlines for an entire project, especially one that's still in Alpha. At this stage, we need to get results as quickly as possible; and increasing the time that the WUs are "out to pasture" will undoubtledly lead people to have even more of them sitting in their cache.

Just some food for thought!

Atomic

I agree and I think this is really an issue that has to be discussed on the boinc email lists. If people are connected to multiple projects and are not connected all the time, the client should be 'smart' enough not to download work to fill all the way up to the deadline, because the chances that all the results will be crunched in that time are not very good...

I've increased the deadline to 10 days; let's see if that makes a difference.

Andre

That tastes like some very good food miss Atomic :-)

We will set more realistic deadlines per workunit when we have fixed most of our app problems and will start distributing some real science work. The workunit we are distributing right now contains the protein-ligand pair 'trypsin-aminomethylcyclohexane' which we use to test because it isn't too complex so it isn't too much of a deal if the deadline is a bit longer.

Thanks for the food though!

Andre

I can certainly understand the need for some breathing room between downloads and deadlines, especially considering some of our beloved BOINC bugs. Also, as Suguru pointed out, many users do not have their systems on all the time, or only have an intermittent internet connection.

I don't think that the deadlines should be TOO long though, as that can lead to other issues, like the "forever and a day" waiting period for validation that so many of our Mac (and some Linux) testers are familiar with. Not only can this be annoying, it can delay the observation, notification, and response to problems that may crop up. Personally, I think 10 days is a little too generous for a 6 hr WU, I feel that 7 days would be more than sufficient to account for the circumstances listed above. I know that a lot of users (particularly avid ones like me who are drawn to new projects!) tend to run multiple projects on the same host, and that is one of the beautiful things about BOINC - but it is a personal choice, and I don't believe that is a good enough reason on its own to push back deadlines for an entire project, especially one that's still in Alpha. At this stage, we need to get results as quickly as possible; and increasing the time that the WUs are "out to pasture" will undoubtledly lead people to have even more of them sitting in their cache.

Just some food for thought!

Atomic

I agree and I think this is really an issue that has to be discussed on the boinc email lists. If people are connected to multiple projects and are not connected all the time, the client should be 'smart' enough not to download work to fill all the way up to the deadline, because the chances that all the results will be crunched in that time are not very good...

I've increased the deadline to 10 days; let's see if that makes a difference.

Andre

I wasn't sure which thread to put this in since there seem to be several covering almost the same information.

Anyway to help out with the imbalance of too many windows vs linux I have backed off my windows hosts some. I'm not giving up, just relaxing some. I know this won't solve the problem but it may help the situation out some. There are a lot of other projects to krunch for and As I am busy busy myself don't currently have the time to monitor every work unit being processed but I try to read the boards some and keep up and keep an eye on my hosts.