Message boards :
Number crunching :
Server flooding me with tasks
Message board moderation
Author | Message |
---|---|
Khali Send message Joined: 21 Mar 14 Posts: 12 Credit: 2,395,589 RAC: 0 |
I seen there was a new batch of AstroPulse tasks put up on the server so jumped in and allowed new tasks from Seti. I need them to get my 100 hours completed for a WUProp@Home badge. I had tabbed out of the Boinc Client and was browsing the web when I noticed Firefox lagging pretty badly and it kept lagging for about 20 minutes. I went to take a look to see if I could find the problem. What I found was my Boinc client was constantly downloading AstroPulse tasks. I hit the No New Work button and went to check my settings. Both my Boinc Client and my account settings on the Seti web page are set to Minimum Work Buffer 0.1 Max Additional Work Buffer 0.2. I have been using this setting for months and never had any problem. To make sure it wasn't a single time glitch I hit the Allow New Tasks button and I instantly got two more tasks downloading. I now have 85+ AstroPulse tasks waiting to be run on my GTX 780 Ti. No, I'm not complaining. These tasks are semi rare and I am happy to run them. But, you might want to check and see if the server has an issue. I don't want to hog all the tasks and not leave any for any one else. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
Are you reserving a CPU core for every AP task that is run on your video cards (1 core for 1 GPU AP task, 2 cores for 2 AP tasks)? Cheers. |
Khali Send message Joined: 21 Mar 14 Posts: 12 Credit: 2,395,589 RAC: 0 |
Are you reserving a CPU core for every AP task that is run on your video card (1 core for 1 GPU AP task, 2 cores for 2 AP tasks)? I have not reserved any CPU cores. Right now my Boinc client is running AP tasks with 0.561 of a CPU core. I also have a GPU Grid Long Run task running and it is using 0.852 of a CPU core. I also have Poem@Home running on what CPU cores I have left in my i5-3570k, which is taking up 3 CPU cores. My system, i5-3570k, 32GB ram, GTX 680(running GPU Grid), GTX 780 Ti(running Seti AP tasks). Both my GPU's are only running one task at a time. The only odd thing I have noticed is the AP tasks are jumping between using 65% and 85% of the 780 Ti. The only things running on my system are Boinc, Firefox, and StarCraft 1. I got nostalgic the other day and installed StarCraft again. I discovered I can run it and not have to throttle down Boinc at all. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
The only odd thing I have noticed is the AP tasks are jumping between using 65% and 85% of the 780 Ti. My GTX 750Tis, even running 2 MB WUs at a time don't run the GPU at 100%, more around 90-95%. The more recent the hardware, the less optimised the application are, hence running 2 or even 3 WUs at a time is necessary to make full use of the hardware. Grant Darwin NT |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
If ATM Khali is doing just AP task each on his two GTX 780 Ti's then he'll need to reserve a CPU core for each GPU task as I think that Khali is doing other projects on his CPU cores. ;-) [edit] either adjust your use multiprocessors setting to use by percentage of CPU core create an app_info.xml to automatically adjust this setting when you do run these (just do an advance search on "an app_info.xml" back 1yr to find what you need if someone doesn't put 1 here). ;-) Cheers. |
Khali Send message Joined: 21 Mar 14 Posts: 12 Credit: 2,395,589 RAC: 0 |
If ATM Khali is doing just AP task each on his two GTX 780 Ti's then he'll need to reserve a CPU core for each GPU task as I think that Khali is doing other projects on his CPU cores. ;-) I don't understand how this is going to stop the server from sending me more tasks. I tried suspending Poem thus freeing up all of my CPU cores and it made no difference. The problem is either with my client, keeps asking for more work even though I have plenty, or the server is stuck and just keeps sending more tasks my way for whatever reason. Which I still don't understand since I have not changed any settings before this all started. Seti has always ran fine before this. Its almost 5AM here so I am just going to leave it set to not get any more work and let it crunch away at the huge pile of AP tasks I have on hand and head to bed. I will look at this more when I wake up. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
At the rate that those cards are going you should wind up with 200. ;-) Your RAC here will climb. There are optimizations and setting available too, but I'll leave that for someone else to explain. Cheers. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
The problem is either with my client, keeps asking for more work even though I have plenty, or the server is stuck and just keeps sending more tasks my way for whatever reason. Which I still don't understand since I have not changed any settings before this all started. Seti has always ran fine before this. It could be considerable debt has built up towards Seti, and the BOINC manager is now trying to pay it back. Grant Darwin NT |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
I see it got up to 109 tasks before you stopped that by setting NNT. With an APR of 761.66 GFLOPS for your 780's, that's about 1.5 days of estimated work. [1] The servers don't send work unless the client asks for some, a large portion of the scheduler code is not run without a work request. It's highly likely that whatever has gone wrong is client side. With "Minimum Work Buffer 0.1 Max Additional Work Buffer 0.2" the client ought to have stopped asking for more work after getting 21 or 22 AP tasks (10 or 11 for each GPU). But if you look back in the event log to the last time AP tasks were sent, I think you'll find there was a work request. Joe [1] AP tasks have an rsc_fpops_est of 1.821e15, dividing by 761.66e9 gives a runtime estimate of 39.8 minutes per task. Multiplying by 109 gives about 3 days, but two GPUs halve the time. |
Khali Send message Joined: 21 Mar 14 Posts: 12 Credit: 2,395,589 RAC: 0 |
When I woke up today the first thing I did was to shut my system down and do a reboot. That seemed to fix what ever the issue was. Apparently it was on my side after all. I am still not sure why it did what it did. Since this had never happened before I jumped to the conclusion that there was something wrong on the server end. Well we all know what assumptions get you. To clear one thing up. I am only running Seti AP tasks on one GTX 780 Ti. My other GPU, GTX 680, is running GPU Grid and that's all it runs. Every other project has been excluded from the GTX 680. I originally had Seti set as a filler project when there were no Poem GPU tasks to run on the 780 Ti. But when I saw there were some AP tasks I put Poem on stand by and went full tilt into the AP work. Thanks for the suggestions and trying to help me out with this issue! |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
When I woke up today the first thing I did was to shut my system down and do a reboot. That seemed to fix what ever the issue was. Apparently it was on my side after all. I am still not sure why it did what it did. Don't suppose you noticed if the date was wrong at all? That has caused problems in the past with people trying to get work. Grant Darwin NT |
Khali Send message Joined: 21 Mar 14 Posts: 12 Credit: 2,395,589 RAC: 0 |
When I woke up today the first thing I did was to shut my system down and do a reboot. That seemed to fix what ever the issue was. Apparently it was on my side after all. I am still not sure why it did what it did. As far as I can see the date and time is correct on my system. But no, I did not really look at it when I did the reboot. I suspect this is going to be one of those glitches that I will never have a clear answer to why it happened. Unless it happens again and I did further into it. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.