User Tools

Site Tools


cluster:167

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
cluster:167 [2018/08/01 08:14]
hmeij07 [Jul 2018]
cluster:167 [2018/08/01 10:11]
hmeij07 [July 2018]
Line 28: Line 28:
 | Job Hours2 % | 39  |  8:1  |  5 | more realistic...hp12 rarely used | | Job Hours2 % | 39  |  8:1  |  5 | more realistic...hp12 rarely used |
  
-The logs showing gpu %util confirm the extremely low GPU usage. When concatenating the four gpu %util values into a string, since 01Jan2017, the string '​0000'​ has occurred 10 million times out of 16 million observations. (GPUs are polled every 10 mins). Sad. The surprising strong GPU job count is due to the Amber group launching lots of small GPU jobs.+The logs showing gpu %util confirm the extremely low GPU usage. When concatenating the four gpu %util values into a string, since 01Jan2017, the string '​0000'​ has occurred 10 million times out of 16 million observations. (GPUs are polled every 10 mins). The surprising strong GPU job count is due to the Amber group launching lots of small GPU jobs.
  
 So were these 25 days in June 2018 an oddity? ​ So were these 25 days in June 2018 an oddity? ​
Line 53: Line 53:
   * The assumption that ''​hp12''​ was barely used in July 2017 might not be correct   * The assumption that ''​hp12''​ was barely used in July 2017 might not be correct
  
-Based on Jul17 we process about 60-70 times more CPU Job Hours than GPU Job Hours, that seems consistent with Jul18. The metric of Job Hours consumed versus Available Hours in %, the picture is probably more like Jul17...30-40% of CPU cycles are consumed and 25% of GPU cycles. ​+Based on Jul17 we process about 60-70 times more CPU Job Hours than GPU Job Hours, that seems consistent with Jun18. The metric of Job Hours consumed versus Available Hours in %, the picture is probably more like Jul17...30-40% of CPU cycles are consumed and 25% of GPU cycles. ​
  
 If we take total hours consumed from Usage Report (the 313,303 hours for Jul17) we consumed about 45% of available hours (without hp12 in the mix). If we take total hours consumed from Usage Report (the 313,303 hours for Jul17) we consumed about 45% of available hours (without hp12 in the mix).
Line 59: Line 59:
 We shall wait for Jul18 metrics. We shall wait for Jul18 metrics.
  
-==== Jul 2018 ====+==== July 2018 ====
  
 ^ Metric ^ CPU ^ Ratio ^ GPU ^ Notes July 2017 ^ ^ Metric ^ CPU ^ Ratio ^ GPU ^ Notes July 2017 ^
Line 68: Line 68:
 | Job Count | 12,​798 ​ |  18:1  |  722 | processed jobs irregardless of exit status | | Job Count | 12,​798 ​ |  18:1  |  722 | processed jobs irregardless of exit status |
 | Avail Hours | 898,​752 ​ |  50:1  |  17,856 | total cpu cores, total gpus | | Avail Hours | 898,​752 ​ |  50:1  |  17,856 | total cpu cores, total gpus |
-| Job Hours |  ​260,997  ​|  ​69:1  |  ​3,805 | cumulative hours of consumed usage | +| Job Hours |  ​322,207  ​|  ​1732:1  |  ​186 | cumulative hours of consumed usage | 
-| Job Hours % | 30  ​|  ​1:1  |  ​26 | as a percentage of available ​+| Job Hours % | 36  ​|  ​36:1  |  ​| as a percentage of available |
-| Avail Hours2 | 696,​384 ​ |  47:1  |  14,880 | total for cpu cores minus hp12's 256 cores, total for gpus | +
-| Job Hours2 % | 37  |  1.5:1  |  26 | more realistic...hp12 rarely used |+
  
  
 +Based on the utilization string '​0000',​ meaning all gpus are idle on a single node as polled every 10 mins during July 2018, the GTX gpus were 65% completely idle and the K20s were 61% completely idle.
  
 **[[cluster:​0|Back]]** **[[cluster:​0|Back]]**
cluster/167.txt · Last modified: 2018/08/01 10:11 by hmeij07