The following caveats do apply to this code base:

• Only stock configs were used for Kafka and Redis (apart fom a few network tweaks). It's by no means production-ready. Given the fact that the Redis benchmark only uses a simplified stream class and the Python language bindings are nowhere complete, do not try this at home! :-)
• I used Confluent's Python client as this apparently is the fastest one on the market if current lore is anything to by.
• Andy's Python client has been pursuaded to understand Streams. As Salvatore's code is not GA yet (and I do expect a few changes on the server side), only a subset of the functionality has been implemented to make the benchmark work. Especially the Consumer Groups functionality is missing from the client implementation. Once the server-side code has been stabilized, I will complete the client side implementation alongside the test harnishes needed to Q/A this codebase and submit a corresponding PR.
• The polishing of the command line argument code is left as an excercise for the interested reader.
• This is for fun only, and not for profit!