reading your description it seems like each data point you get is a self contained item, is that right? or are you recalculating some data for the entire block with each new message?
On Sun, Mar 19, 2017 at 9:27 PM, Nate McCall <n...@thelastpickle.com> wrote: > I think you would be better served by using a streaming system like Apache > Flink (http://flink.apache.org) and checkpointing occasionally to > Cassandra. > > This is a significant increase in complexity, but you are describing a > real-time streaming use case with the need for watermarking time windows > and Flink has that all built in. > > >