[shanemhansen]

I'm not sold on Samza, but I can tell you that creating isolated services that create their datastore from a stream of events is a really useful pattern in some use cases (ad-tech).

I've made use of NSQ to stream user update events (products viewed, orders placed) to servers sitting at the network edge which cache the info in leveldb. Our request latency was something like 10 microseconds over go's json/rpc. We weren't even able to come close to that in the other nosql database servers we tried, even with aggressive caching turned on.

[anonymousDan]

What don't you like about Samza out of interest? Something fundamental with their model or more implementation related?

[shanemhansen]

I've seen organizations have lots of trouble operationally with kafka (which samza uses). I've seen NSQ be extremely reliable operationally.

However they offer very different guarantees so it's an apples to oranges comparison. NSQ isn't really designed to provide a replayable history, although you can fake it by registering a consumer which does nothing but log to file (nsq_to_file) and that works pretty well.

(disclaimer: the nsq mailing list has lots of chatter these days, nsq may be growing features I'm not aware of)

[jonathanoliver]

What troubles did you typically see with Kafka? Was it Zookeeper related? (Also, good to talk to you at the Gopher Meetup on Tuesday)