| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Beta Testing
Forum: Beta Test Support Forum
Thread: Requiring 100% CPU to get beta work |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 26
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
This is advanced stuff. If you are genuinely interested, then I (again) strongly encourage you to look it up yourself. And, again, I encourage you not to answer unless you have something useful to say. "look it up" is not meaningful without giving something specific about what to look up and where. And, again, it shouldn't be necessary for users to research something which should be simple. These units run for exactly the same amount of CPU time on every machine (give or take a few mins). Their real world run time is thus equal to that amount of CPU time divided by the proportion of time that they will be devoting to WCG. If the estimate that the scheduler uses is something other than that, the only way to know what that might be would be for someone here to tell us. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You have your answer. I'm recommending a lock for this thread. You not only won't answer, but you want to lock the thread so that nobody else will be able to answer it either? Wow. Do you suppose you could be any less helpful at all? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
i think that we should let our computers do all the thinking and not worry ourselfs about points and badges and meaningless stuff, just don't look at it all working and it won't get on your nerves
|
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
An 8-hour rice unit takes how long to run? 8 hours. A 1-hour beta takes how long to run? 1 hour! And the DCF matters how exactly? It's very unusual that a wu takes exactly the same amount of cpu-time, regardless of which computer it's running on. Therefore, there's nothing that is an "8 hour-wu" as far as BOINC's Scheduling-server goes, instead it's a wu with rsc_fpops_est of 3.1e13. The Scheduling-server uses rsc_fpops_est together with client-supplied info about p_fpops, on_frac, active_frac, cpu_efficiency and duration_correction_factor to calculate EWD (estimated run-time). If your "connect about every N days" > delay_bound (time to deadline) or if EWD + estimated_delay > delay_bound you'll get the "won't finish in time"-message. estimated_delay is how much work you've already got... My recollection is a little fuzzy here, but if not mistaken this is the expected run-time of the cached work per cpu...  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
bionicle3112 wrote:
i think that we should let our computers do all the thinking and not worry ourselfs about points and badges and meaningless stuff, just don't look at it all working and it won't get on your nerves Quite true. I generally just leave boinc doing its thing. Occasionally I take an interest in what it's doing out of curiosity and comment on things that appear odd or interesting. It's no big deal. The only thing that gets on my nerves is those who have nothing pleasant or useful to say, but want to drown out discussion. Ingleside wrote: The Scheduling-server uses rsc_fpops_est together with client-supplied info about p_fpops, on_frac, active_frac, cpu_efficiency and duration_correction_factor to calculate EWD (estimated run-time). If your "connect about every N days" > delay_bound (time to deadline) or if EWD + estimated_delay > delay_bound you'll get the "won't finish in time"-message. Interesting. If estimated_delay is included, that's going to make things very random. It'll depend on how far through its current work a machine is, even though the beta units will have tight deadlines and automatically be done in preference anyhow. (unless you have existing WUs that are close to deadline too) If EWD is calculated as you say, even for these WUs, then it's amazing that my old P3/866 managed to pick up several of those betas. The EWD for that machine should have been huge. (Larger than the machine which didn't get any.) DCF might be the issue, but that should just reflect the difference between one machine running 90% of the time. This will be +/- some random factor, as DCF actually fluctuates a great deal. In other Boinc systems, having only a single type of project, DCF perhaps works a little better (maybe?), but here that single variable is spread over a number of projects (what we might call "projects" -- Boinc treats "WCG" as the "project") which behave totally differently. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
bionicle3112 wrote: i think that we should let our computers do all the thinking and not worry ourselfs about points and badges and meaningless stuff, just don't look at it all working and it won't get on your nerves Quite true. I generally just leave boinc doing its thing. Occasionally I take an interest in what it's doing out of curiosity and comment on things that appear odd or interesting. It's no big deal. The only thing that gets on my nerves is those who have nothing pleasant or useful to say, but want to drown out discussion. Ingleside wrote: The Scheduling-server uses rsc_fpops_est together with client-supplied info about p_fpops, on_frac, active_frac, cpu_efficiency and duration_correction_factor to calculate EWD (estimated run-time). If your "connect about every N days" > delay_bound (time to deadline) or if EWD + estimated_delay > delay_bound you'll get the "won't finish in time"-message. Interesting. If estimated_delay is included, that's going to make things very random. It'll depend on how far through its current work a machine is, even though the beta units will have tight deadlines and automatically be done in preference anyhow. (unless you have existing WUs that are close to deadline too) If EWD is calculated as you say, even for these WUs, then it's amazing that my old P3/866 managed to pick up several of those betas. The EWD for that machine should have been huge. (Larger than the machine which didn't get any.) DCF might be the issue, but that should just reflect the difference between one machine running 90% of the time. This will be +/- some random factor, as DCF actually fluctuates a great deal. In other Boinc systems, having only a single type of project, DCF perhaps works a little better (maybe?), but here that single variable is spread over a number of projects (what we might call "projects" -- Boinc treats "WCG" as the "project") which behave totally differently. Hi Kremmen, emails and blog post quite often get misunderstood, purly because we can't see the persons face or here how they are saying it, I am sure this is what happens here sometimes, keep the blood pressure down, the spirits high and the wu's and cpu time massive. LOL Simon [Edit 1 times, last edit by Former Member at Jun 15, 2008 9:15:51 AM] |
||
|
|
|