Performance compared to zeromq

[ActivePeter]

I am currently evaluating the performance of various inter-process communication libraries and plan to compare your library with ZeroMQ. To better understand the performance characteristics of your library, could you please provide the following performance data:

• Key performance metrics such as throughput, latency, and resource consumption.
• Any comparison data with ZeroMQ under similar test conditions, particularly in high concurrency and large-scale communication scenarios.

Thank you very much for your help. I look forward to your response!

[elfenpiff]

@ActivePeter We have a benchmark suite which you can use, see: https://github.com/eclipse-iceoryx/iceoryx2/tree/main/benchmarks

Like ZeroMQ, iceoryx2 has multiple messaging patterns like (event, publish-subscribe, or request-response) so you can compare them individually since they all have their own benchmark. The setup of the benchmark is described in the documentation there.

Key performance metrics such as throughput, latency, and resource consumption.

1. For the latency we have a graph in the main README.md and also a comparison with other posix mechanisms.
2. Throughput does not make sense for zero-copy solution since it is infinite. Once the data is produced (no matter how large it is) the transmission of the data is always reduced to delivering a pointer to a shared memory location. Therefore, the throughput is arbitrary large. This is something where a zero-copy solution shines most in comparison to a network protocol.
3. Resource consumption is minimal.

• We do not start threads.
• The binary should be around ~2 - 3MB but we did not yet start with the optimization there
• Shared Memory Segments (/dev/shm/*) depend on the payload size used by the user.

Any comparison data with ZeroMQ under similar test conditions, particularly in high concurrency and large-scale communication scenarios.

We have not created any comparison benchmarks between iceoryx and any other framework. Feel free to create one; we would be interested in the results.

If you want to start writing those benchmarks, please feel free to reach out to us. We are happy to support you there if some API calls or the documentation are unclear.

[elBoberido]

@ActivePeter see also this discussion for benchmarks with low level IPC mechanisms
#435