For straightforward tumbling windows, the regular DSL windowing performs noticeably better than a custom process function because it takes advantage of an internal API to avoid some serialization overhead.
There's a simple example of a ProcessWindowFunction in [1], and an example of using a KeyedProcessFunction to do windowing in [2]. Best, David [1] https://ci.apache.org/projects/flink/flink-docs-release-1.12/learn-flink/streaming_analytics.html#window-functions [2] https://ci.apache.org/projects/flink/flink-docs-release-1.12/learn-flink/event_driven.html#example On Mon, Jan 4, 2021 at 1:05 AM Navneeth Krishnan <reachnavnee...@gmail.com> wrote: > Hello All, > > First of all Happy New Year!! Thanks for the excellent community support. > > I have a job which requires a 2 seconds tumbling time window per key, For > each user we wait for 2 seconds to collect enough data and proceed to > further processing. My question is should I use the regular DSL windowing > or write a custom process function which does the windowing. I have heard > that the DSL window has more overhead versus the custom window function. > > What do you guys suggest and can someone provide an example of custom > window function per key. Also given the window time is very less (2 secs) > would there be more overhead in firing so many timers for each key? > > Thanks! > > Regards, > Navneeth >
