InMemory Stream Issues, would you replace with Kafka or use Orleans Persistent streams

[sajjadhs]

Hi, In Memory streams are not recommended for production use, but I'm using it as I needed to remove any latencies, and my system is tolerant against recovering lost data.
But I see many issues with Orleans stream when running multiple instances, and for the times when some message sizes are big or the number of messages is too high, in such scenarios I get grain locating issues, or missing some hops to silos, or getting issue with an object not found in the cache, etc.

So I decided to fully replace in-memory streams, considering above mentioned issues, would you replace Orleans streams with AWS-hosted or self-managed Kafka, or would go with persistent stream options with Orleans?

For me replacing with some persistent stream options for Orleans looking to be simpler but since it will be out of my ecosystem, in which is all resources are on AWS, going to Microsoft Azure will add extra hops to my works.