Hi Martin,
Thanks for your question.
Shane made some great points. I’d like to re-iterate that the performance comparison aspects of that article need to be taken with a grain of salt. There are lies, damn lies, and benchmarks.
NSQ performance is solid, see
http://nsq.io/overview/performance.html. As stated on that page, single node performance isn’t very interesting for a platform like NSQ because it’s designed to be deployed in a distributed manner, with many nodes participating in the delivery of messages to consumers.
It’s been on my TODO list to publish benchmarks performed on a larger cluster. In my previous throughput testing I’ve been able to easy saturate gigabit links on multiple producing/consuming machines, but this deserves a more thorough examination.
The real power of NSQ, that I’m sure the community can attest to, is how easy it is to build real systems on top of. I sincerely doubt it will be your bottleneck but I can damn near guarantee you’ll enjoy working with it.
Hope this helps.
Regards,
Matt