| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Is it time to revisit the 35 WU limit? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 34
|
|
| Author |
|
|
OldChap
Veteran Cruncher UK Joined: Jun 5, 2009 Post Count: 978 Status: Offline Project Badges: 
|
I remember encountering this when the 4 core machine was really popular and 4c/8t were considered great
----------------------------------------Time and technology have moved on so my 16c/32t machines are considered old hat and yet even these cannot "connect to the internet" every 7 days on projects where the runtime is short for each WU. How folks cope with 64t up to 448t machines I have no clue I find rigs need up to 240 wu per thread on my relatively slow machines to last the 10 days seen in settings. The limits are a construct that serve only to annoy the user I think so what use are they? can we do away with them?  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
wasn't it changed to 70? I think I get 70 per core. Once you get up to about 16c, you start to hit the 1000 per machine hard-coded BOINC limit. I hit the 1000 limit on most of my machines except the 4c and 8c. 32c machines hit 1000 and can do those in about 30 hours. At least the short running WUs
----------------------------------------You can create an app_config to limit the number of concurrently executing tasks and then set the n_cpu value in cc_config higher to get around the soft per core limit [Edit 1 times, last edit by Doneske at Nov 14, 2018 1:35:34 PM] |
||
|
|
OldChap
Veteran Cruncher UK Joined: Jun 5, 2009 Post Count: 978 Status: Offline Project Badges:
|
Then maybe it is the 1000ish per core limit that needs addressing too
----------------------------------------Multiply cores/threads by 35 here takes that to 1120 and as I said the most I have is 32t |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1316 Status: Offline Project Badges: 
|
Then maybe it is the 1000ish per core limit that needs addressing too ..... The 1000 tasks limit / project is a BOINC limitation (too many runnable tasks). When you have < 1000 tasks, you get as many tasks as you need to fill your buffer limited to what the project is assigning to your machine per single request. For ZIKA I noticed, I mostly don't get more than 173 in 1 request, so in theory you could get (in my example) a max of 1172 tasks in your buffer incl. running ones).  [Edit 1 times, last edit by Crystal Pellet at Nov 16, 2018 8:35:19 AM] |
||
|
|
VietOZ
Senior Cruncher United States Joined: Apr 8, 2007 Post Count: 205 Status: Offline Project Badges: 
|
For Zika, the max I can get was 1200 tasks on a TR 1950X
----------------------------------------80 pages x 15 tasks per page = 1200 Not always hit that number but I've seen it on different machines as well. |
||
|
|
KLiK
Master Cruncher Croatia Joined: Nov 13, 2006 Post Count: 3108 Status: Offline Project Badges:
|
+1
---------------------------------------- |
||
|
|
hchc
Veteran Cruncher USA Joined: Aug 15, 2006 Post Count: 747 Status: Offline Project Badges: 
|
Is this an issue that should be addressed on the BOINC GitHub? Even on my 4c/4t machines, I'm limited to getting 4 x 70 = 280 work units before I get the message "This machine has reached a limit on tasks in progress." On the current SCC1 short-running work units, my machine churns through its buffer in about 6-8 hours or so. I lose connectivity, which means this machine can be idle for an entire day. I'd much rather be able to buffer 1-2 days or more without some silly limit.
----------------------------------------I'd like to see this 70 task/thread limit removed as well as the 1000 total tasks in progress limit.
|
||
|
|
giba
Veteran Cruncher Brazil Joined: Dec 2, 2004 Post Count: 846 Status: Offline Project Badges: 
|
Well, my point is that any limit have concerns. Why limit if there are crunchers with hardware available to run tasks ? Don't make sense for me !
The central question on WCG ever was contribute to the Science using grid computing as much as possible, so why put limits if someone would like to contribute a little bit more tha others ?      |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Is this an issue that should be addressed on the BOINC GitHub? Even on my 4c/4t machines, I'm limited to getting 4 x 70 = 280 work units before I get the message "This machine has reached a limit on tasks in progress." On the current SCC1 short-running work units, my machine churns through its buffer in about 6-8 hours or so. I lose connectivity, which means this machine can be idle for an entire day. I'd much rather be able to buffer 1-2 days or more without some silly limit. I'd like to see this 70 task/thread limit removed as well as the 1000 total tasks in progress limit. You can double or triple that number with just a few settings and a config code line, but you have to hit the books yourself to figure it out. Not a good idea to make this common knowledge. You'll hit the secondary codeline soon after as a machine is not allowed to have more than 1000 in progress. This defeats the workaround, ending up with even less than the 70 per thread. A 16 core can then have around 60. [Edit 1 times, last edit by Former Member at Sep 19, 2019 6:55:40 AM] |
||
|
|
hchc
Veteran Cruncher USA Joined: Aug 15, 2006 Post Count: 747 Status: Offline Project Badges:
|
I believe I'm being misread. I'm saying that both the per core limit and the 1000 total limit should be eliminated, especially as multi-core CPUs keep coming to market. My only question is if this is something that should be raised as an issue on the BOINC GitHub.
----------------------------------------
[Edit 2 times, last edit by hchc at Sep 19, 2019 9:16:20 AM] |
||
|
|
|