Run time hours lost
log in |
Message boards : Number crunching : Run time hours lost
Author | Message |
---|---|
I have been noticing lately that the time I run for WCG is not being accurately reported here. Today shows I only logged a little over 3k hours here while the my contribution chart on the WCG site shows more than 4k hours of run time which is what it should be. Why are the same amount of hours not being recorded here? | |
ID: 4273 · Rating: 0 · rate: / Reply Quote | |
I have not seen my hours update for over 24 hours. | |
ID: 4274 · Rating: 0 · rate: / Reply Quote | |
Yes, it doesn't count in "running time", but in "last day" there is a minus. | |
ID: 4276 · Rating: 0 · rate: / Reply Quote | |
It's still not showing the correct amount of run time. I have 200+ cores running @ WCG. That should give me 4800+ hours of daily run time. It's always 300-400 hours short on that here. Am I missing something? | |
ID: 4285 · Rating: 0 · rate: / Reply Quote | |
It's still not showing the correct amount of run time. I have 200+ cores running @ WCG. That should give me 4800+ hours of daily run time. It's always 300-400 hours short on that here. Am I missing something? Probably not. First I find that OSX and Linux hardly ever log 24 hrs/core. On some apps I might get as low as 20hrs/day which I'm not losing sleep over. You might have some time lost between returning your WUProp and starting the next one. Look at your computer with the highest loss and see if there's a delay in there. In all, I'd say a 10% loss, while not perfect, isn't a massive loss. If I'm running multiple GPU tasks to increase efficiency, I rarely get 100% of the hours. I have no idea why. Some projects cough up almost 100% some as low as 75%. But as WCG has no GPU work, this can't be it. I don't know if you're running VMs to get 200+ cores or if you have 200+ bare metal cores. | |
ID: 4287 · Rating: 0 · rate: / Reply Quote | |
It's still not showing the correct amount of run time. I have 200+ cores running @ WCG. That should give me 4800+ hours of daily run time. It's always 300-400 hours short on that here. Am I missing something? 10% loss seems to be about right but that sure adds up over the years. It's really no biggy. Just a curiosity. Only the 2 4 core android devices I run seem to show the full run time. (192 hours per day) I'm not running VMs. I guess it would be more accurate to say I have 200+ threads (106 Intel Xeon cores)running Linux. Thanks for the explanation. | |
ID: 4288 · Rating: 0 · rate: / Reply Quote | |
1. Check hosts -> activity -> each info block from-to | |
ID: 4289 · Rating: 0 · rate: / Reply Quote | |
When running some ATLAS tasks recently (which run in their own VM and used up lots/all the resources [RAM & Disk]) some of my times were out a lot. 2. Add <report_results_immediately>1</report_results_immediately> to cc_config This can be added to the project (by admin), to report tasks on completion immediately. It's used by GPUGrid. cc_config may only function on some BM versions. | |
ID: 4291 · Rating: 0 · rate: / Reply Quote | |
It's getting worse at WCG. 18-20% lost run time. | |
ID: 4316 · Rating: 0 · rate: / Reply Quote | |
Last update for WCG. I have been monitoring this for the past month. | |
ID: 4536 · Rating: 0 · rate: / Reply Quote | |
Message boards :
Number crunching :
Run time hours lost