Off-by-1000 error in worker dashboard/billing?

I have a trivial worker installed. As per the worker dashboard:

  • It received 113 request in the last 24 hours
  • Median cpu time was 1 millisecond, 99.9th percentile was 1.8 millisecond.
  • Total Execution duration was reported as 11.8 GB-seconds.

Now I understand workers can use 128MB of RAM and are billed as using 128MB of ram for the entire execution duration.

So now the numbers don’t add up:

  • If I go by the median execution time reported, counting 113 requests, I get 113 requests * 1millisecond/request * 128MB = 113*128 MB-milliseconds = 14464 MB-millisec = 14.125 GB-millisec = 0.014125 GB-seconds.
  • If i go by the total execution time reported, then I get 11.8 GB-seconds/113 request = 0.1044 GB-seconds per request = 106.930 MB-seconds per request = 106930 MB-milliseconds per request. 128MB is the billed RAM, so this amounts to 106930/128 = 835 milliseconds per request. But as per dashboard even the 99.9th percentile execution time is 1.8 milliseconds.

I think the dashboard has a bug, possibly merely a display bug, not a billing bug: I think the total execution time value being reported is correct, but the units being reported should be GB-milliseconds, not GB-seconds. Probably just a typo. Everything adds up then.


Please be sure to open a ticket with our Support team so this can be further investigated.

1 Like