[BMarkmann]

I understand the underlying architecture is fundamentally different, but the end result seems kind of akin to some of the "complex event processing" (CEP) tools out there like Esper (ie. feeding incoming data into established queries instead of executing queries against at-rest data). Would this fit similar sorts of use cases / fit into the CEP market?

[nehanarkhede]

Yes KSQL is ideal for use cases similar to those that CEP was initially targeted for and more - from real-time anomaly detection, monitoring, analytics to application development and Streaming ETL. As you alluded to, the big difference is that KSQL is designed as a distributed Streaming SQL engine that can run at Kafka scale.

[edan0]

If you are interested in performing ultra low latency (<1ms) CEP via SQL, take a close look at SQLstream Blaze (http://www.sqlstream.com) and it's full implementation of Allen's Interval Algebra (https://en.wikipedia.org/wiki/Allen%27s_interval_algebra) via SQL temporal predicates: http://sqlstream.com/docs/sqlrf_planned_feature_temporal_pre.... This stuff runs at 1,000,000 events per second per core and can scale out in conjunction with Kafka across any number of servers.