Questions on ops/sec and usec/ops in rocksdb perf benchmark results

[Alec Hothan]

Hi there,

Looking at some of the results table in https://github.com/facebook/rocksdb/wiki/Performance-Benchmarks

Test1 table looks good and consistent however

Test 2 table seems to show inconsistent data, for example:

Version Opts ops/sec mb/sec usec/op p50 p75 p99 p99.9 p99.99

6.10.4 None 138496 35.1 462.1 617.1 764.3 1240 1295 2484

with ops/sec = 138496, the average usec per op would be

1,000,000 / 138496 = 7.22 usec (assuming it is average ops/sec)

while the usec/op column shows 462 usec which is 2164 op/sec (assuming it is average usec/op)

Subsequent tables are also exposing the same discrepancy. Am I interpreting the columns incorrectly?

Thanks

 

[Mark Rambacher]

There are 64 threads in these tests. 

1,000,000 / 138496 = 7.22 usec

7.22 usec * 64 threads = 462.1 usec

So the ops/sec is a per-thread result whereas the usec/op is a per-process one.

There are likely a few other math/rounding issues as the thread values are summed up into a single result (rounding of microseconds to seconds, etc).

Does that answer your question?

/ Mark

--
You received this message because you are subscribed to the Google Groups "rocksdb" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rocksdb+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/rocksdb/1ce94c00-96f5-4ad1-9dbe-77915153ee3en%40googlegroups.com.

[Alec Hothan]

Thanks for the quick reply Mark, that explains the data reported.

  Alec