@klmn: Thanks, glad to know that someone else is using the same architecture and processing more data than us - it can get lonely being on the cutting edge! 😃

Are onyx’s windowing aggregation functions like :onyx.windowing.aggregation/sum just for convince? Or do they offer some performance benfiet? I assume I can perform an arbitrary calculation on a window on a trigger. For example sum or standard deviation. If the benfiet is performance, because the sum is updated in real time, then could you role your own auto-updating function like standard deviation?

You'll need to create your own aggregation following the pattern in onyx.windowing.aggregation. It is there that you provide the logic for updating a window, and merging discrete window states.

You can perform arbitrary computation on the trigger, but you want to avoid using stuff like conj and then triggering on all the segments because it'll mean checkpointing the full segments, rather than the computation results / differences

Onyx 0.9.9 is out. Just a small bug fix this time around.