----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/64217/#review192386 -----------------------------------------------------------
Thanks! We've done the cross-process benchmarking technique a few times already to avoid the local message passing, but it's rather complicated. The question has come up before of whether we can add an option to disable local message passing in libprocess in order to simplify these benchmarks. Have you looked into this? My intuition is that it shouldn't be too difficult of a change to make and I'd be happy to help you with it. - Benjamin Mahler On Nov. 30, 2017, 6:03 p.m., Ilya Pronin wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/64217/ > ----------------------------------------------------------- > > (Updated Nov. 30, 2017, 6:03 p.m.) > > > Review request for mesos and Benjamin Mahler. > > > Bugs: MESOS-8120 > https://issues.apache.org/jira/browse/MESOS-8120 > > > Repository: mesos > > > Description > ------- > > These benchmarks test the performance of the old libprocess message passing > based scheduler API and the new v1 scheduler HTTP API. Unlike other Mesos > benchmarks, these are not built in 'mesos-tests' binary to avoid libprocess > message passing shortcut that kicks in when both peers are in the same > process. > > > Diffs > ----- > > src/Makefile.am 4a3b728dc2cbfaf4b10068562778a87a6b331238 > src/examples/CMakeLists.txt d4f1af4f072efdc68fa0b722f42b1d8aa1779b6e > src/examples/test_scheduler_driver_bench.cpp PRE-CREATION > src/examples/test_scheduler_v1_bench.cpp PRE-CREATION > > > Diff: https://reviews.apache.org/r/64217/diff/1/ > > > Testing > ------- > > Verified manually by launching a master and runninng benchmarks. > > > Thanks, > > Ilya Pronin > >
