| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: GO Fight Against Malaria
Thread: 10 days of new work remaining for GFAM |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 83
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
BOINC 'computes' the buffer -per- processor thread, meaning if you have a quad and set the cache to 10 days, you really get 40 processor days. That's assuming that your device processes tasks near the mean runtime the servers are computing from current actual returns.
There is the variability of run times and the varying performance of devices. As what I was bemoaning up in this thread, my Linux box does them 60+% faster than the average, which means that if tasks arrive, they show 1.6 the estimated runtime factor then they'll actually take to complete. When I want to buffer 2 days on that device, I really have to set 2*1.6 i.e. about 3.3 days, to get 2 days -per- processor. But if I want to have 10 days [maximum], there's no hope of getting that since the 10 / 1.6 = 6.25 days. The server wont have the device get more since the deadline is 10 days. Anything more and it will tell you the device is on for X% and computing Y% of that. Additional work not completing in time. That's a [not flawless] safety against hosts going wild. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I use to buffer work when I was running 1-3 threads about when I came to WCG from Grid.org.. now I got 20 threads working on projects I don't buffer since pretty much one completes as another one comes in.. :D most I will see a window from me at full power is 1-2 hours the most between workunits. depending on the project I am doing at the time..
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The point of the buffering in the end game is to overcome any work shortage to reach a personal goal, when knowing that supply will dry up imminently, a safeguard so to speak. In my case, still not there, but getting
4417 World Community Grid 6/3/2013 1:15:38 PM Tasks won't finish in time: BOINC runs 98.3% of the time; computation is enabled 99.9% of that! That's life and we're live... doing the hives thing or as they say in one language 'sweating carrots' :D |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
@ Gopal Sharma See these FAQ for guidance: BOINC: Deadlines for Standard & Rush Tasks and What Clients Get These When Overdue Tasks Are Really Not Worth Completing! It is not recommended to exceed 7 days of work buffer [WCG recommend maximum 4] if e.g. running the 10 day deadline projects, alone or in a mix, simply because of potential outages, particularly if work is only returned periodically. It would be helpful to know what version of BOINC you are running? Dear jonnieb-uk Thanks for sharing the links. They were helpful. . I have reset buffer and Aborted 87 workunits. So like you say, buffer > 4 not recommended, then probably folks talking about it are doing it inspite of this not being a good choice. Cant figure out how it helps them. By the way, my machine runs nearly 24x7 (except few reboots) and is also always connected to internet. So network outage, machine downtime were not considerations for me. I have just one Lenovo T420 laptop and am able to return 8-10 units in every 24 hours cycle. so for 10 days maybe 80-100 workunits is what I can do. Thats my rule of thumb estimate. Seems like I was fed 200 units instead of that !!. The Version of BOINC I am on is Version 6.10.58. wxWidgets Version is 2.8.10. Thanks again for sharing the infromation. Gopal Sharma |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
BOINC 'computes' the buffer -per- processor thread, meaning if you have a quad and set the cache to 10 days, you really get 40 processor days. That's assuming that your device processes tasks near the mean runtime the servers are computing from current actual returns. There is the variability of run times and the varying performance of devices. As what I was bemoaning up in this thread, my Linux box does them 60+% faster than the average, which means that if tasks arrive, they show 1.6 the estimated runtime factor then they'll actually take to complete. When I want to buffer 2 days on that device, I really have to set 2*1.6 i.e. about 3.3 days, to get 2 days -per- processor. But if I want to have 10 days [maximum], there's no hope of getting that since the 10 / 1.6 = 6.25 days. The server wont have the device get more since the deadline is 10 days. Anything more and it will tell you the device is on for X% and computing Y% of that. Additional work not completing in time. That's a [not flawless] safety against hosts going wild. I will echo your observations on BOINC grossly overestimating the Time need to compute. In my case GFAM tasks were estimated at 2.5 times ie 25 hours instead of 10 that it always took. However, not sure what really prompted the over feed of workunits to me. I understand the logic for you have proposed for 4 core*10 days = 40 day work. It just seems in my case I was sent some 200 workunits whereas on average I manage to do 80-100 workunits in 10 days. Not sure what happened there. Anyways I just aborted 87 units. Hope that helps some of others looking for them if needed. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Are we running on fumes already? [you know the feeling, on that motorway, with the needle firmly indicating red zone and knowing the next pump-station is 50 miles away]
(still not getting any... 98.9% on, 99.9% of that running ;) |
||
|
|
astrolabe.
Senior Cruncher Joined: May 9, 2011 Post Count: 496 Status: Offline |
I have 41 in my buffer so I hope I have enough to get to 9 years. The idea of 9 years + 100 days evaporated when the estimate shrank from 10 days to 4 days in 1 day.
|
||
|
|
Jack007
Master Cruncher CANADA Joined: Feb 25, 2005 Post Count: 1604 Status: Offline Project Badges: 
|
3 hours from this post (before)
----------------------------------------I received 67 wu, and 76 wu on 2 machines.  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I got a bunch of units last night as well and they are not due unil the 12th. Should be enough for a bronze
---------------------------------------- [Edit 1 times, last edit by Former Member at Jun 3, 2013 5:55:48 PM] |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
17,430 results left to send which is 3-4 hours of steady work remaining.
|
||
|
|
|