General Guideline on Sink Count in relation to Performance

[zacharyhansen]

HI all - amazing product you have built!

I had a general question that I am looking for direction on - what is the performance / scaling impact of adding M sinks to a sequin deployment?

As in should I attempt to limit the number of sinks I have in the application and instead do filtering on the webhook / consumer side (ie scale consumer servers) or does sequin scale pretty well with adding many sinks where I can really push most of this filter logic to the filter definition of the sinks.

In my product, Im using sequin as the trigger for many async processes in my application. For example one bucket would be triggering user defined workflows on create, update, delete of records, another would be system workflows on similar update create and delete and will likely have more buckets of these different async types. I wanted to see if you recommend a best practice for whether you believe I would create many sinks with the filters set via sequin or if I should attempt to limit the number of syncs and move filters to the consumer application.

Likely theres no silver bullet type of answer here - but mainly looking to get an idea of what an acceptable number of syncs are based on memory / cpu assuming constant WAL messages or how this sink number ends up effecting scaling of a single deployment.

Thinks in advance!