| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Outsmart Ebola Together
Thread: ! no tasks available ! |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 438
|
|
| Author |
|
|
Sandvika
Advanced Cruncher United Kingdom Joined: Apr 27, 2007 Post Count: 112 Status: Offline Project Badges: 
|
well I kind of works for me. I am getting some WU's each day. Unfortunately, I have only been getting them on one of my 3 computers. Each day its a different one. I have a similar issue, in that the WUs are being distributed very unevenly. I looked in the event log for clues as to why one should be running short and found this: | Requesting new tasks for CPU | Scheduler request completed: got 0 new tasks | No tasks sent | Tasks are committed to other platforms This coincided exactly with another device getting a large allocation. I think there may be an overall quota in place for WUs in progress as opposed to the per device settings that I adjusted upwards to see the device through the majority of the time when there is no work available. At least it got a few WUs before hitting this limit so it can get through the night without stopping and hopefully will get a few more in that time, otherwise it will be back to UGM for this device! Good luck    |
||
|
|
Eric_Kaiser
Veteran Cruncher Germany (Hessen) Joined: May 7, 2013 Post Count: 1047 Status: Offline Project Badges: 
|
I observed that I didn't get more than 25 wu in progress per core.
----------------------------------------So that seems to be the quota. On two of my rigs it didn't exceed this quota which means 100 wu. On the third it stays a little below this boundary. 220 was the maximum I observed. Theoretical 275 is naximum if my assumption is correct and because I set wcg to use 11 out of 12 cores. However, I'm getting plenty of wu.  |
||
|
|
KWSN-A Shrubbery
Senior Cruncher Joined: Jan 8, 2006 Post Count: 476 Status: Offline Project Badges:
|
25 per core is a limit imposed by WCG overall. It has nothing to do with the projects and was implemented quite some time back.
----------------------------------------I think it was when HCMD2 was finishing up and things were going wonky (tm).  |
||
|
|
Sandvika
Advanced Cruncher United Kingdom Joined: Apr 27, 2007 Post Count: 112 Status: Offline Project Badges:
|
Hi Eric,
----------------------------------------I googled the message and it suggests that work is prioritised to be sent to certain types of hosts based on their productivity. That's funny because my environment is virtualised using VMWare ESXi (that's a whole new topic for a home computer) and all my WCG virtual machines are clones, so they will behave identically. I've got 4 x 6 cores on WCG. Your discovered limit of 25 WUs in progress per core would yield me upto 25 x 24 = 600 WUs but I just added them up and got 411 = 115 + 123 + 125 + 48 reported on the WCG results status reports. On the actual devices I have 115 + 123 + 125 + 33 WUs. It is the one with the mismatch (48 vs 33) which is not getting the work, so maybe the problem is the mismatch, rather than with the perceived type of host. Maybe it is being blamed for losing or being late with these 15 mismatched WUs? As my environment is virtualised a neat way of dealing with this imbalance would be to over-subscribe the available cores by running it as 4 x 8 cores, and use the "processor usage" preferences in the BOINC manager to use 75% of the processors (ie. 6) by default, increasing if more WUs arrive and decreasing if there's a shortage, but keeping the total running WUs at 24. Unfortunately I need to run multiple instances as the free "home lab "version of ESXi is limited to 8 virtual processors per virtual machine. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
25 per core is a limit imposed by WCG overall. It has nothing to do with the projects and was implemented quite some time back. I think it was when HCMD2 was finishing up and things were going wonky (tm). Yes, and a much higher number for each GPGPU, if there's work for these of course  The effect of 25 is, that the faster devices hardly can buffer a days worth of work per core, the whole 10 day maximum the client allows rather academic. By and large with <dont_use_dcf/> in place this limit has become redundant, it takes the mean runtime for task headers, not the extremes. IOW no more risking buffer overflow [albeit some report they still have this, was it Kersamson?, although the client is not mentioned? Version 6 it could still occur since 6.n and earlier do not listen to <dont_use_dcf>] Would be nice WCG got rid of this frustrating 1-2 work units per request and set BETA test like restriction on OET1. Today got another notice dated the 19th titled 'How your Tablet Can Help Find an Ebola Cure". It can't as yet. Maybe the forum silence is driven by heads down in coding purgatory, for going into X-mas and not having met the annual targets is no incentive.  (Just wonder if uplinger stepped on a duck while there ;P) [Edit 1 times, last edit by Former Member at Dec 20, 2014 1:16:52 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
25 per core is a limit imposed by WCG overall. It has nothing to do with the projects and was implemented quite some time back. I think it was when HCMD2 was finishing up and things were going wonky (tm). Yes, and a much higher number for each GPGPU, if there's work for these of course .... (Just wonder if uplinger stepped on a duck while there ;P) I read a couple of posts earlier that WCG based when or who to send new work to based on productivity! If I can't get any work to be productive with, what's the point? I have chosen two projects to get work for, OET and UGM, and it appears that I am getting only 4 or 5 OET WUs a day. How do I get this increased? If OET is the only project I choose, I run out of work in no time at all! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You read "I googled the message and it suggests that work is prioritised [sic] to be sent to certain types of hosts based on their productivity. " That makes it authoritative of course ;>)
The only host priority that exists is for the _Repair_ work units, which go to hosts that have. 1) an average return rate of under 48 hours, for the work received. 2) are considered reliable for the specific science application, speak having had the last 20 or so valid for for instance FAHV or MCM1. One error for an app resets the counter to zero, another serial 20 valid needed to become reliable again. More in depth, there's a differentiation between zero redundant work, and those always requiring a quorum of 2. Those rules are not well documented, in fact irrelevant to the volunteer cruncher. Repairs are extra copies sent out in place of an Error/Invalid/No Reply and get a deadline that is computed as 35% of the standard project deadline to force a speedier quorum, but there's exceptions to that too. Sometimes extra copies also go out with the original number of days deadline. Don't ask me more as that's all I readily remember. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
After reading many if not all posts on this issue I have come to the point that WCG just needs to dump the forums and start a Q&A section with dashboards......both sides of this issue have to much ego.
|
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
After reading many if not all posts on this issue I have come to the point that WCG just needs to dump the forums and start a Q&A section with dashboards......both sides of this issue have to much ego. Really?
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
  |
||
|
|
branjo
Master Cruncher Slovakia Joined: Jun 29, 2012 Post Count: 1892 Status: Offline Project Badges:
|
Siedentopf wrote: After reading many if not all posts on this issue I have come to the point that WCG just needs to dump the forums and start a Q&A section with dashboards......both sides of this issue have to much ego. Strong ego is important in our lives, w/o it, we are zeroes and we can not move forward. You have probably confused it with "selfishness" and/or "narcissism", aren't you?  Cheers   Crunching@Home since January 13 2000. Shrubbing@Home since January 5 2006  |
||
|
|
|