SeccoSQL

Experimental distributed SQL engine that separates communication from computation.

SeccoSQL is an experimental distributed SQL engine on Spark designed for complex SQL and graph-style queries.

Its defining idea is to explicitly decouple relational operators into pure communication and pure computation operators, which exposes a much larger plan space than conventional distributed SQL engines.

That separation makes it possible to reorder operators at a finer granularity and reduce communication cost more aggressively on distributed workloads.

Links: Code ยท SIGMOD 2022 paper