| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Why are the stats wrong? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 24
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello all,
I have started switching all my machines that were rolling UD Agent to the BOINC. Ever since I did that, I have noticed that the calculation time is drastically down in my devices page, giving me ridiculous results. I have a total of 20 machines that have been running 24h/day for 2 weeks, and in the total time running for some reason it gives me 3 days, 4 days. Those machines are dedicated to running BOINC, just as they were dedicated to running UD. BOINC is running as a service. When they were running UD, the results were on the dot, I would cumulate around 20 days of calculation a day...now it's not working anymore Would anyone know why the calculation time has gone down so low? |
|||
|
||||
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Hi,
----------------------------------------Saw 2 RedDust crunchers at BOINCstats, but presume this is you: http://boincstats.com/stats/user_graph.php?pr=wcg&id=121862 No idea why you took a ditch. Have you visited the My Grid > Result Status page to see what the status of your work is. It allows you to filter by device and type of work and validation status. When converting from UD to BOINC, wallclock time does not go anymore, rather real CPU time is counted for validated work, which can delay a week or more in some cases. See the Start Here forum Migration FAQ for many a question. ttyl
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
What is CPU time VS Wall Clock time?
I looked at my results and only have 4 results awaiting validation, and lots of work in progress (as expected). The processors are all working full time on the projects, all of them are calculating 24h a day. I'm afraid I don't understand what CPU time is...if a CPU is working all day, isn't the CPU time equal to 24 hours? Or is it related to the speed of the CPU? Most of the computers running have P4 CPUs. Thanks. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
The old agent would count 24 hours a day if the PC runs 24 hours a day, but with BOINC it only counts the real CPU portion it gets. On a very busy computer you easily run 80% CPU Efficiency.
----------------------------------------Now did you elect to change the My Grid > Device Manager > Device Profile as the 'Standard' which uses 60% or the 'Maximum output' which uses 100%? When starting BOINC there is the initial validation delay but once you have build a steady stream you will see your production creeping up. You will though rarely see a constant as the validation of each work unit can take different times depending on how quick the others in the quorum process their copy. This is something you'd never see with UD where points were given before validation and regardless if valid or not. ttyl
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Feb 11, 2008 9:45:55 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello again,
The "Default" profile, on which all my units are set, is set to "Maximum output". As I mentioned in my previous post, I only have 4 results pending validation right now. I'm really wondering what's going on... |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
I run Maximum Output as the Science runs at lowest priority and only tweak individual clients where 'interference' is perceived.
----------------------------------------Think it needs a few more days, in mean time check if Maximum Output was set to Default and is attached to the profile that is used by the device. If you post a copy of the Global_Prefs.XML of one of the client and also, if any, global_prefs_override.xml, we can tell you more. With 4 in PV and no errors or invalid listed and maximum output and 24/7 your numbers, certainly in the run time should be higher on 20 devices.
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 2 times, last edit by Sekerob at Feb 11, 2008 10:26:58 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi, RedDust.
I see your BOINC statistics have skyrocketed since the beginning of February. They are just beginning to level off now. I think the reason you may be simply looking at some misleading numbers. Where are you getting this figure of "CPU hours per device per day"? As far as I know, WCG don't provide this number directly. http://boincstats.com/stats/user_graph.php?pr=wcg&id=121862 |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Tend to have a quick peak at the My Grid > My Statistics > Devices Statistics to see the date/time of last valid result returned for each and drill on to see the daily hours history per device. That quickly identifies which have a problem. For a real quick single window, current CPU efficiency and colour coded picture of all devices in a LAN the BOINCview 3rd party tool is incomparable. Keeps track of completed work long after the Result Status page has removed it (4 days after validation).
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges: 
|
Hi RedDust!
----------------------------------------I see other points to consider. From your stats I have the impression that a good share of your crunching farm consists of older machines. And older machines were not only slower but also they had less RAM. If many of them don't have enough RAM to avoid swapping their CPUs are waiting for their disks (slower too, probably) and while waiting they do not increment the CPU clock. Another possibility is that the screensaver options are active. On a slow machine that could account for much CPU time, even if there is no switched on screen to show them (as I presume). Whether the above guesses are valid or not I think the most practical way to investigate this problem is to start the TaskManager on one of the most typical ones, sort the list of processes (second tab) on the processor time and see what the process percentage is for the Science application. If it is far from 100 % see if the difference goes to the "System Idle Process" line or if any other background task is taking too much of your CPU time. Cheers. Jean. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello all,
All machines running (except 1) have at least 512MB RAM in them along with a P4 processor (minimum 1.7ghz up to 3.4 ghz + a few dual cores). Screen saver is off on all the machines. When I look at the stats per device, it shows me really low numbers still. I took a couple of machines randomly and checked the processes. Boinc is taking around 80% of the processor on both (idle is around 20%). What doesn't make sense is that for a given day, I'll have around 10 hours of calculated time on the device even though the machine is on 24 hours. Now that doesn't come up to 80% of 24 hours. Not even 60% actually. Those machines, as mentioned are completely dedicated to working on the Boinc process. They have Windows and Boinc installed, that's it. No other apps. |
||
|
|
|