How is the performance? about journal HOT 3 CLOSED

We have done some preliminary benchmarking. Journal is really just a very thin wrapper around SLF4J.

Any difference is strictly because of the asynchronous nature of Journal. Journal has a dedicated thread consuming the log queue, so if all your CPUs are saturated this will actually be slower, not faster. Using SLF4J with AsyncAppender, you should be able to get exactly the same performance profile.

from journal.

In addition, the macros ensure that if you have a particular log level turned off, the logger calls will be completely removed from the compiled code. So e.g. your debug log statements will have no overhead in production.

from journal.

Published the latency benchmark in 7b8d45c

Running it on my laptop gives the following timings:

Journal:
  min: 14 ns
  max: 87724 ns
  avg: 49.300918 ns
SLF4J:
  min: 1243 ns
  max: 32737417 ns
  avg: 4592.998692 ns

So Journal reduces the latency of logging by roughly two orders of magnitude.

from journal.