| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Numbers on active machines? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 12
|
|
| Author |
|
|
zombie67 [MM]
Senior Cruncher USA Joined: May 26, 2006 Post Count: 228 Status: Offline Project Badges: 
|
I'm curious. How many active machines are currently crunching for WCG, and the split between UD and BOINC?
---------------------------------------- |
||
|
|
retsof
Former Community Advisor USA Joined: Jul 31, 2005 Post Count: 6824 Status: Offline Project Badges: 
|
It's a difficult statistic to get, since you can generate as many machines as you want and some machines are part timers. I have over 100 devices myself, but only 4 are currently active. Other computers have gone away and other configurations are unused.
----------------------------------------Some third party sites have attempted to estimate it by using "24 hour" machines, but even that was not very accurate.
SUPPORT ADVISOR
----------------------------------------Work+GPU i7 8700 12threads School i7 4770 8threads Default+GPU Ryzen 7 3700X 16threads Ryzen 7 3800X 16 threads Ryzen 9 3900X 24threads Home i7 3540M 4threads50% [Edit 1 times, last edit by retsof at Apr 21, 2007 3:41:11 AM] |
||
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges: 
|
Wouldn't WCG have to know?
----------------------------------------Couldn't they just see how many discreet devices have returned results within some give time period - say 30 days? Of course some devices would fall through the cracks, but it would still have to be a pretty good estimate. I've had to register the same device name multiple times, but in those cases you would just pick up the last verified result. If the same device gets registered under different names within the same 30 day period, that would inflate the estimate, but what percentage of the time is that likely to happen? Plus, even in that case, it's going to be registered under the same user name (most likely), so if you see 4 devices that have reported consistently for the past couple of weeks and 10 more that haven't reported for longer than that, you can probably safely exclude the 10 - or at least assign them a weighted value of less than one. If you then run this for multiple overlapping time periods, you can reduce the error by even more. So worst case scenario is that we get an estimate with a + or - error margin of maybe a few percentage points. Even with a margin of 10% or 20% I think it would be still be something worth knowing.   |
||
|
|
zombie67 [MM]
Senior Cruncher USA Joined: May 26, 2006 Post Count: 228 Status: Offline Project Badges:
|
It's a difficult statistic to get, since you can generate as many machines as you want and some machines are part timers. I have over 100 devices myself, but only 4 are currently active. Other computers have gone away and other configurations are unused. Yeah, but that's just a flaw (being unable to merge machines). It's simple enough to look at only the numbers within the last X days. Some third party sites have attempted to estimate it by using "24 hour" machines, but even that was not very accurate. Heh. and how would anyone know that they were not very accurate without accurate stats?  Seriously though, I'm sure that the project folks have numbers at hand. They would be required information for discussions when prospective customers approach WCG for work. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Yesterday the equivalent of 60,329 full time devices = Days of CPU time from the stats :O
----------------------------------------The sum u can do is going to BOINC stats, see the active devices reported and extrapolate to the total WCG points. I'll check back later for a honors giving, for who got closest (IMHO)
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges:
|
Yesterday the equivalent of 60,329 full time devices = Days of CPU time from the stats :O The sum u can do is going to BOINC stats, see the active devices reported and extrapolate to the total WCG points. I'll check back later for a honors giving, for who got closest (IMHO) Being one to sit in the back of the class, I'm probably a little slow. Which BOINC stat should we add to the 60,329? Active hosts (114,687)? This doesn't look right since total and active hosts is the same but total users is more than double the number of active users (37,963 vs. 17,630). I'm assuming "user" in this context means unique login and not unique machine name. What am I missing? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
This URL ( http://vspx27.stanford.edu/DCcomparison.html ) divides the total time run in a day in hours by 24 to derive the equivalent number of active computers running 24/7.
|
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Being one to sit in the back of the class, I'm probably a little slow. Which BOINC stat should we add to the 60,329? Active hosts (114,687)? This doesn't look right since total and active hosts is the same but total users is more than double the number of active users (37,963 vs. 17,630). I'm assuming "user" in this context means unique login and not unique machine name. What am I missing? Think the word 'sum' led to misinterpretation. A sum was meant to be a formula :) Looks like something broke at BOINCstats.... in past active devices was for all projects a representation of 'active' devices.... last I saw it, it was around 29,000 for WCG. With about 3,700,000 credits, 25879000 WCG points were made (47.9% of Friday total). Translating becomes more iffy by the day, as the UD agent is single core, where BOINC can use as many cores as allowed in the device profile with a trend persisting towards BOINCification. Anyway, the future statistics pages are intended to include active members and active devices (those returning a result on a calendar day e.g.). When they come..... don't know. Given that some UD clients crunch a week and longer on a single result, any number is just a statistical number anyhow.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges:
|
This URL ( http://vspx27.stanford.edu/DCcomparison.html ) divides the total time run in a day in hours by 24 to derive the equivalent number of active computers running 24/7. If I'm following all of this, the number at that link, 60,329, is just for UD clients. Sek I believe is saying to add in the active cpu's from BOINC clients that crunch for WCG. That's were it gets tricky since you have some cpu's taking several days for a work unit. Therefore they're in fact active, but not listed as such except on the day that they return the wu - I think. Anyway, I guess the short answer then is still "we don't know" but the number will bear some resemblence to whatever you get when you add the 1. UD agent hours divided by 24 PLUS 2. number of active hosts for WCG/BOINC. How does that sound? |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
No, it's a number for all. It's 'too simple' to be obvious, but take the WCG CPU years as reported for Friday - 165:104 CPU years -, do 165 X 365 + 104 and you get there. The Grid.org device number is higher though their CPU years is much lower as they report real active devices (those returning a result). Thus if you correlate them with their 108 CPU years of Friday, you get to about 92,262 devices for WCG. Difference is, Grid.org is pure wall-clock, whereas 40%+of WCG is true CPU time.
----------------------------------------BTW, the algorithm used by Stanford can be found in the link to the Perl script on their page.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
|