| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Suggestions / Feedback
Thread: 35 WU Limit too Small |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 18
|
|
| Author |
|
|
Mamajuanauk
Master Cruncher United Kingdom Joined: Dec 15, 2012 Post Count: 1900 Status: Offline Project Badges: 
|
NixChix - Can I suggest you post some of your settings, as it may be an issue with settings restricting how many WU's are cached.
----------------------------------------The reason I ask is I have some 8 core machines (only a couple left) that happily cache 280 WU's. This machine is set to run as follows:
100% of the time Cache is 3.5 days Disk space is 30Gb For example, if you have only set to run 20% of 4 cores, it would lead to 1 processor x 35 WU's On my 64 core machines I regularly see 2240 WU's in the cache...
Mamajuanauk is the Name! Crunching is the Game!
  |
||
|
|
theodolite
Advanced Cruncher Joined: Apr 11, 2014 Post Count: 119 Status: Offline |
Couple of thoughts.
Wouldn't it be great to have 4 x i7 computers and unlimited WU so I could load up with 30,000 of short .25 hour OET WU (you can do the math) so that no one else could get any WU. That would be so righteous.  I'm someone who is happy simply to increase my IP from 12 to 52 pages while supply lasts  and when the longer WU come back, I'll go back to about 12 pages of IP again.  Last I saw, this was a DC project, not focused on a single person's ego.  |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7581 Status: Offline Project Badges: 
|
I assume that oet has same runtimes as fah (~30 minutes/wu. Even though OET is a VINA project the run times are highly variable with reports of anywhere from 10 minutes to several hours. Just depends on the batch. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
NixChix
Veteran Cruncher United States Joined: Apr 29, 2007 Post Count: 1187 Status: Offline Project Badges: 
|
I was getting the full 35-per-core limit on WUs. The problem is that when they take just 15 minutes or less. I am out of town for a wedding at the moment and can't access by BOINCtasks history. The WCG results page already purged the completed WU history, so I can't get actual data. I was seeing 15-20 minutes run-time estimates with a total queue size of 6 hours (24 hrs / 4 cores). The slower crunchers had longer estimates.
----------------------------------------The whole purpose of a queue was to be able to store up a few days of work to cover dry periods. Of course there should be a limit. It should be what you can (and have) crunch in the time allotted for the work. An arbitrary cap of 35 favors slow, intermittent members at the expense of the workhorses. If the job sizes are highly variable and are impossible to size for a specific crunchers's capability, then that just demonstrates why a one-size-fits-all limit of 35 was just plain wrong. Cheers   |
||
|
|
Byteball_730a2960
Senior Cruncher Joined: Oct 29, 2010 Post Count: 318 Status: Offline Project Badges:
|
I don't get it. Is this computer not connected to the internet 24/7?
If so, why would you need to babysit the computer. The only computer I have an issue with is a laptop that runs 23.5/7 but is only connected to the internet twice a week. Otherwise, every other computer is fine as long as my buffer settings are sufficient |
||
|
|
NixChix
Veteran Cruncher United States Joined: Apr 29, 2007 Post Count: 1187 Status: Offline Project Badges:
|
I don't get it. Is this computer not connected to the internet 24/7? Tried to have the lil' crunchers queues partially filled and working hard on OET while still asking for only FAAH work. Fill them with 2 or 3 days of work and then change project selection on the profile. Repeat every 2 or 3 days. It used to be a valid technique when trying to get work for projects when the work flow was sporadic, just like FAAH when the supply of original WUs came to an end and only resends were going out.If so, why would you need to babysit the computer. The only computer I have an issue with is a laptop that runs 23.5/7 but is only connected to the internet twice a week. Otherwise, every other computer is fine as long as my buffer settings are sufficient BTW, the length of the OET WUs has increased somewhat lately. Point is moot since FAAH WUs are now rarer than hen's teeth (239 returned yesterday). The 35 WU limit is still lame. Cheers |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
A side problem is, it's an over-all limit, not a per-science [If only we could set a Max In Progress per science in the app_config.xml with a server side controlled cap **]. 35 for MCM gets you easily overloaded... they take 9 hours on my half [summer] speed 4770, and that means 35 IP = 13 days is possible, were it not the deadline is 7.
** This would allow us to also balance much more what's crunched at a local level without risking idle cores, but mostly this tiring alternation of profile settings would not be needed, buffer lots of one, then switch profile to go netting for something else]. Now have one device that does OET+UGM. Getting many more OET than UGM, so set latter to max compute at 4 of 8. If FIFO determines UGM is older there will always be 4 UGM running, but if few, OET can take it all, or get priority if aged more than UGM. So far see a nice constant 4&4 compute of OET&UGM with a 1.5 day buffer depth. [/Off Trail] |
||
|
|
KLiK
Master Cruncher Croatia Joined: Nov 13, 2006 Post Count: 3108 Status: Offline Project Badges: 
|
Don't know about octo-cores...
----------------------------------------But my quad-cores have all app_config.xml file to run only 1 CEP2 project & 3 of any of the other projects: FA@h, OET, MCM, UGM...& it works just fine! Even though I'm just crunching only UGM & OET currently with optional projects added if there r no jobs on server... maybe same numbers would work on octo-cores? so 6:2...  |
||
|
|
|