4 ms·

In the companies I've worked at, those kinds of workloads have been done in Spark or in Beam (GCP Dataflow, etc.)