| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Outsmart Ebola Together
Thread: Is there a limit on tasks in progress? My cruncher is hungry! |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 20
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Puzzling. Maybe the on_frac / active_frac values are screwed up on the poorly buffering one. Check client_state.xml. For a 24/7 machine [last 10-14 days], the values should be near 0.999000. The other is of course if the actual storage space for BOINC is less than the settings [event would tell how much net can be used]. The lowest of the 3 possible applies, so if e.g. there's 2.2 GB free, but same time it's specified to leave 2.0GB free on the relative partition, the actual storage cannot exceed 0.2GB.
|
||
|
|
Byteball_730a2960
Senior Cruncher Joined: Oct 29, 2010 Post Count: 318 Status: Offline Project Badges: 
|
Is this the one SekeRob?
</host_info> <time_stats> <on_frac>0.996962</on_frac> <connected_frac>0.995987</connected_frac> <cpu_and_network_available_frac>0.995863</cpu_and_network_available_frac> <active_frac>0.999878</active_frac> <gpu_active_frac>0.000623</gpu_active_frac> <client_start_time>1429655374.740461</client_start_time> <previous_uptime>58145.141539</previous_uptime> <last_update>1429713518.876942</last_update> |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Very yes on 0.999878 :D
But, there's too <gpu_active_frac>0.000623</gpu_active_frac> BOINC is supposed to maintain buffers separately for CPU and GPU, but JIC, would stop the client, edit that one to 0.999000 and start again. |
||
|
|
theodolite
Advanced Cruncher Joined: Apr 11, 2014 Post Count: 119 Status: Offline |
I have a single core machine running windows 7 with a one day buffer and OET exclusively.
----------------------------------------I have never had more than 2WU at any for the last 8weeks. Go figure [Edit 1 times, last edit by theodolite at Apr 22, 2015 11:39:13 PM] |
||
|
|
Byteball_730a2960
Senior Cruncher Joined: Oct 29, 2010 Post Count: 318 Status: Offline Project Badges:
|
SekeRob, I should edit <gpu_active_frac></gpu_active_frac> to
<gpu_active_frac>0.999000</gpu_active_frac> ? Theodolite, it seems that I am not the only one with this issue. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
Interesting ! Since the OET1 WUs require more time (about 1 hour or longer), the buffers of the both hosts fill up again in a very similar way.
----------------------------------------Maybe the distribution of multiple short WUs (between 10 and 20 minutes) over several days makes the buffer management more challenging for fast systems ... ? Thank you again for your inputs. Cheers, Yves |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7846 Status: Offline Project Badges:
|
Interesting ! Since the OET1 WUs require more time (about 1 hour or longer), the buffers of the both hosts fill up again in a very similar way. Maybe the distribution of multiple short WUs (between 10 and 20 minutes) over several days makes the buffer management more challenging for fast systems ... ? Thank you again for your inputs. Cheers, Yves The great disparity in the length of the OET1 units wreaks havoc with the buffers. I leave mine set at 1 day. When I get those onslaughts of short units I end up with enough units to last 3 or 4 days as the units get longer. It eventually works itself out, but it boomerangs around a bit. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have a single core machine running windows 7 with a one day buffer and OET exclusively. I have never had more than 2WU at any for the last 8weeks. Go figure What CPU is you 'single core' exactly, and what -actual- runtimes [elapsed]? Just copy/paste your result list from here http://tinyurl.com/3448g5 in a reply. Never had, never have trouble getting OET when set to exclusive. Either they fill the buffer properly, or they are limited to 35 per core if they run really really short, but those bouts never take too long [does not persist]. If you want them, you can get them, as the below last 7 day project stats show with a high on April 17 of 365K results validated: 04/23/2015 31:313:03:48:18 51,458,331 215,853 04/22/2015 29:167:20:12:12 47,033,351 277,790 04/21/2015 29:212:18:06:50 46,722,905 335,978 04/20/2015 28:193:18:14:37 45,482,240 239,119 04/19/2015 29:249:15:51:41 46,723,301 378,854 04/18/2015 29:029:02:16:15 45,980,733 249,349 04/17/2015 29:119:23:14:24 47,447,798 364,873 |
||
|
|
theodolite
Advanced Cruncher Joined: Apr 11, 2014 Post Count: 119 Status: Offline |
The system is in a remote location. I am reasonably certain it is on 24x7, but I have no control over that. Set to use 100% of processors, 90% CPU time. Lifetime stats on this unit are:
xxxxx 04/24/2015 12:11:00 2:314:04:17:38 2,004,090 4,770 Given that the install date is 09/28/2010 16:35:02, maybe the issue is that it is being turned off overnight/weekend and I didnt realize that. OET1_ 0000708_ xEBGP-FA_ rig_ 74529_ 0-- xxxxx In Progress 4/24/15 12:11:01 5/4/15 12:11:01 0.00 / 0.00 0.0 / 0.0 OET1_ 0000707_ xEBGP-FA_ rig_ 44697_ 0-- xxxxx In Progress 4/24/15 07:31:12 5/4/15 07:31:12 0.00 / 0.00 0.0 / 0.0 OET1_ 0000707_ xEBGP-FA_ rig_ 18857_ 0-- xxxxx Valid 4/24/15 00:29:04 4/24/15 12:11:01 6.61 / 6.72 110.1 / 110.1 OET1_ 0000707_ xEBGP-FA_ rig_ 25827_ 0-- xxxxx Valid 4/23/15 21:30:11 4/24/15 07:31:12 3.99 / 4.05 64.8 / 64.8 OET1_ 0000696_ xZAGP-FW_ rig_ 86573_ 1-- xxxxx Valid 4/23/15 16:39:57 4/24/15 00:29:04 0.71 / 0.72 10.0 / 16.2 OET1_ 0000706_ xEBGP-FA_ rig_ 74413_ 0-- xxxxx Valid 4/23/15 06:49:48 4/23/15 21:30:11 5.48 / 11.44 184.1 / 184.1 OET1_ 0000704_ xZAGP-S_ rig_ 96431_ 0-- xxxxx Valid 4/23/15 03:46:56 4/23/15 16:39:57 2.40 / 2.46 37.3 / 37.3 OET1_ 0000696_ xZAGP-FW_ rig_ 60646_ 1-- xxxxx Valid 4/21/15 07:26:41 4/21/15 09:32:44 1.07 / 1.08 15.8 / 26.2 |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Absolutely. The amount of buffering is prorated to the actual average up-time per unit of time [24 hours] in the values the client maintains as posted by vcd683s above. If the on_frac is for instance 0.500000 [12 hours] and the active_frac of that is 0.500000, half of 12 hours the client computes, the amount of buffered work is adapted to that. Maximum maintained with 24/7 over last 10 days, else the value is slowly decreased, based on actual uptime.
----------------------------------------That said, not sure if partial computing percent goes into the equation, but would be surprised it would not, so if the "Use at most nn% of CPU time" is set to less than 100% [default profile is 60%], then that too has buffered work total impact. Kind of [late] Spock logic.  [Edit 1 times, last edit by Former Member at Apr 24, 2015 4:37:13 PM] |
||
|
|
|