Hi, I haven't used Onyx, but it's come onto my radar because I'm looking for a distributed pub-sub solution which supports load-balanced collections of subscribers (to scale event processing). Latency is a significant factor for my use-cases, as is ordering, and at-least-once delivery guarantees. My question to the channel is: should I start to go deep on Onyx and spike something out or is it simply not a good fit? I've taken a look at MQTT brokers but I cannot find a robust clustering solution that can take the abuse I'm throwing at it with my resilience tests; SNS+SQS exhibits some bad worst case latencies which rule that out in my mind.

I'll be wanting to scale publications out to 10Ks / second

and this is for server-to-server in AWS VPC

@codeasone It’s a good fit for that, yeah. Onyx supports in-order processing for certain topologies and has at-least-once/exactly-once (for aggregations) delivery guarantees. When tuned properly, it hits single digit millisecond end-to-end latency in AWS.

Thanks @michaeldrogalis I'll take a closer look then 🙂

@codeasone Sure. You can hit us up at <mailto:[email protected]|[email protected]> if you have long, detailed questions. Otherwise lots of help to be had right here. 🙂