The benchmark results might not directly apply to your situation. Since the benchmark server only sends data back and doesn't perform anything meaningful, unlike real workload servers, the actual QPS you'll see will likely be much lower and the difference between C++ and Java/Go becomes insignificant. The raw numbers you see from the benchmark should be correct.
On Thu, Oct 3, 2024 at 3:43 PM 'Amirsaman Memaripour' via grpc.io < [email protected]> wrote: > Thanks again. Just to verify, we expect to see the C++ implementation to > offer half of the throughput of the Go and Java implementations, providing > a peak of 260K QPS on an 8 core server, correct? > > On Thursday, October 3, 2024 at 6:36:46 PM UTC-4 [email protected] wrote: > >> Those numbers look correct to me. >> >> FYI, We're actively working on C++ performance enhancements (namely >> EventEngine and Promise migration), and we anticipate seeing improvements >> once those are implemented. I'm hoping that those changes are complete some >> time next year. >> >> >> On Thursday, October 3, 2024 at 3:21:34 PM UTC-7 Amirsaman Memaripour >> wrote: >> >>> Thanks for sharing the link. I'm wondering if the issue referenced in this >>> PR <https://github.com/grpc/grpc/issues/14899> is addressed, and if the >>> numbers reported for C++ are correct? In other words, should I expect ~260K >>> QPS for running the streaming, secure `ping` benchmark against an 8 core >>> server? If so, why is the C++ implementation not as efficient as the Go and >>> Java? >>> >>> [image: Screenshot 2024-10-03 at 6.17.42 PM.png] >>> >>> On Thursday, October 3, 2024 at 5:24:02 PM UTC-4 [email protected] >>> wrote: >>> >>>> https://grpc.io/docs/guides/benchmarking/ is what we have for this >>>> topic. You can get continuous benchmark data from >>>> https://grafana-dot-grpc-testing.appspot.com/?orgId=1 >>>> >>>> On Tuesday, October 1, 2024 at 1:44:25 PM UTC-7 Amirsaman Memaripour >>>> wrote: >>>> >>>>> Pinging in case this didn't show up in your radar :) >>>>> >>>>> On Tuesday, September 24, 2024 at 1:11:21 PM UTC-4 Amirsaman >>>>> Memaripour wrote: >>>>> >>>>>> Hi folks, >>>>>> >>>>>> Are there any published latency numbers for the baseline overhead of >>>>>> a C++ ping server, using gRPC's CQs? I'm primarily interested in the per >>>>>> request CPU overhead of gRPC's RPC stack on the server-side, and if there >>>>>> are studies on tuning the number of polling threads and CQs to optimize >>>>>> that cost and maximize throughput on a few CPU cores. >>>>>> >>>>> -- > You received this message because you are subscribed to a topic in the > Google Groups "grpc.io" group. > To unsubscribe from this topic, visit > https://groups.google.com/d/topic/grpc-io/N1q33b5qEP8/unsubscribe. > To unsubscribe from this group and all its topics, send an email to > [email protected]. > To view this discussion on the web visit > https://groups.google.com/d/msgid/grpc-io/d6763f5e-b05c-4633-82c7-ce08823be005n%40googlegroups.com > <https://groups.google.com/d/msgid/grpc-io/d6763f5e-b05c-4633-82c7-ce08823be005n%40googlegroups.com?utm_medium=email&utm_source=footer> > . > -- Regards, Esun. -- You received this message because you are subscribed to the Google Groups "grpc.io" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/grpc-io/CAFp8MWxV-0ypCOgJLQLAWr0dT-d%2BKXac4cm-Mi52ApGBC_VPfA%40mail.gmail.com.
