Create serializer benchmarks and integrate to CI

[klimashkin]

Hi! Would it be possible to retire msgpack5?
According to my tests on different payloads msgpack5 is more than 3x times slower than the official @msgpack/msgpack, with the identical output.

Here some articles, related to the latter:
https://appspector.com/blog/how-to-improve-messagepack-javascript-parsing-speed-by-2-6-times
https://appspector.com/blog/how-to-use-streaming-data-parsing-in-javascript

[oberstet]

According to my tests on different payloads msgpack5 is more than 3x times slower than the official @msgpack/msgpack, with the identical output.

we don't have tested different cbor libraries on diffferent run-times (browser, node, ..) using different payloads yet

we have tested crossbar (and autobahn|python) in such systematic way, here are 2 links to give you a clue:

• overview https://crossbario.com/docs/benchmarks/serialization/index.html
• example result https://crossbario.com/docs/benchmarks/serialization/vmprof_pypy_cbor_normal_medium.html

until we have similar comprehensive test results, claims to "X being faster than Y" are not convincing and it doesn't make sense to change stuff in blind

IMO, we need to systematically produce results at least for:

• different serializers (libraries)
• chrome, firefox and node
• different WAMP messages, with different payload sizes and payload complexity

[klimashkin]

Thanks for sharing those benchmarks, didn't see them before!

In my case I was just naively testing simple buffer(stringify), stringify+gzip, messagepack and cbor on the latest NodeJS using the npm libraris. Encode/decode on each approach was run 1000 times in a loop.
CBOR was always slightly bigger than the MessagePack (Length=size)
Sharing the results here as it might be helpful for somebody

5.5Kb JSON

160Kb JSON

1.5Mb JSON