Low latency Python execution

By default, RapidMiner starts a new Python interpreter for every Python operator embedded in a RapidMiner process. For most use cases this behavior is desirable as it guarantees complete script isolation and the 100-1000 ms of overhead for initializing the Python interpreter are usually negligible.

When deploying processes as web services where the expected response time is in the 50-100 ms range, the above overhead will be too large. To handle this use-case, we offer an alternative, remote Python execution engine named RapidMiner StandPy. StandPy is shipped and deployed as an optional component of RapidMiner AI Hub. Please visit the RapidMiner StandPy deployment page on how to deploy, operate and use this component.