Differences

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

--- cluster:167 [2018/06/27 17:18]
hmeij07 [CPU vs GPU]
+++ cluster:167 [2018/06/28 11:46]
hmeij07 [CPU vs GPU]
@@ Line 10: / Line 10: @@
   * That period covers 600 hours of time
   * Assume 99% utilization of cpu core or gpu device
-  * Available time is measured per cpu core but by gpu device
+  * Available time is measured per physical cpu core but by gpu device
   * There is no good/bad metric
   * Never collated such data before
@@ Line 16: / Line 16: @@
-^ Metric ^ CPU ^ Ratio ^ GPU ^ Notes ^
+^ Metric ^ CPU ^ Ratio ^ GPU ^ Notes June 2018 ^
 | Device Count | 72  |  3:1  |  24 | cpu all intel, gpu all nvidia |
 | Core Count | 1,192 |  1:54  |  64,300 | physical only  |
 | Memory | 7,408  |  51:1  |  144 | GB |
 | Teraflops | 38  |  1.5:1  |  25 | double precision, floating point, theoretical |
-| Avail Hours | 715,200  |  50:1  |  14,400 | total cpu cores, total gpus |
 | Job Count | 2,834  |  3:1  |  1,045 | scheduled jobs irregardless of exit status |
-| Job Hours | 221,136  |  77:1  |  2,872 | cummulative hours of consumed usage |
+| Avail Hours | 715,200  |  50:1  |  14,400 | total cpu cores, total gpus |
-| Avail Hrs Util% | 31  |  6:1  |  5 | weeping... |
+| Job Hours | 221,136  |  77:1  |  2,872 | cumulative hours of consumed usage |
+| Job Hours % | 31  |  6:1  |  5 | as a percentage |
 | Avail Hours2 | 561,600  |  39:1  |  14,400 | total cpu cores - hp12's 256 cores, total gpus |
-| Avail Hrs2 Util% | 39  |  8:1  |  5 | more realistic... |
+| Job Hours % | 39  |  8:1  |  5 | more realistic...hp12 rarely used in June18|
+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.
+So were these 25 days in June 2018 an oddity? March is Honors' Theses time so lets look at Jul17 so we can compare that to Jul18 in august.
+^  Total Monthly CPU+GPU Hours  ^^^^^^^^^^^
+^Ju17^Aug17^Sep17^Oct17^Nov17^Dec17^Jan18^Feb18^Mar18^Apr18^May18^
+|313,303|273,051|128,390|111,224|280,101|51,727|306,453|222,585|437,959|262,227|294,724|
-The logs showing gpu %util confirm the extremely low GPU usage. When concatenating the four gpu %util values int 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 launches lots of small GPU jobs.
+^ Metric ^ CPU ^ Ratio ^ GPU ^ Notes July 2017 ^
+| Device Count | 72  |  4:1  |  20 | cpu all intel, gpu all nvidia |
+| Core Count | 1,192 |  1:42  |  50,000 | physical only  |
+| Memory | 7,408  |  74:1  |  100 | GB |
+| Teraflops | 38  |  1.7:1  |  23 | double precision, floating point, theoretical |
+| Job Count |   |  :  |   | scheduled jobs irregardless of exit status |
+| Avail Hours | 886,848  |  60:1  |  14,880 | total cpu cores, total gpus |
+| Job Hours |  260,997  |  69:1  | 3,805  | cumulative hours of consumed usage |
+| Job Hours % | 30  |  1:1  |  26 | as a percentage |
+| Avail Hours2 | 696,384  |  47:1  |  14,880 | total cpu cores - hp12's 256 cores, total gpus |
+| Job Hours % |  37 |  1.5:1  |  26 | more realistic...hp12 rarely used in June18|
+  * Some noise in this data with the inability to match start and end of job (~15% of records)
+  * The assumption that ''hp12'' was barely used might not be correct
+  * 939591860
 **[[cluster:0|Back]]**

DokuWiki

User Tools

Site Tools

Differences

Page Tools