The latter is correct. Spark is only used for input/output, and CPLEX doesn't provide a direct integration with Spark. So it wouldn't distribute the solve across the cluster. On the other hand, even if CPLEX doesn't know anything about Spark, you can certainly build a platform with which it is possible to distribute
multiple solves across a cluster, each each solve using a single node.
Note that DOcplexcloud is being replaced with
Decision Optimization for Watson Studio. You can find more information in the
DO for WS section of
https://medium.com/@AlainChabrier/decision-optimization-education-a50cada93856.------------------------------
Xavier Nodet
Program Manager, Development
CPLEX Optimization Studio
------------------------------