found this breakdown on how to stop throwing money at bigger clusters just to deal w/ lagging data. instead of just scaling out, it uses
netflix maestro and
apache iceberg to tackle the root cause of rising costs and stale batches.
it's way better than the usual "just add more nodes" strategy . anyone else moving away from
traditional batch processing for this?
article:
https://dzone.com/articles/netflix-maestro-apache-iceberg