Training and Serving Machine Learning Models at Scale

In recent years, Web services are becoming more and intelligent (e.g., in understanding user preferences) thanks to the integration of components that rely on Machine Learning (ML). Before users can interact (inference phase) with an ML-based service (ML-Service), underlying ML model must learn (training from existing data, a process requires long-lasting batch computations. The management these two, diverse phases is complex meeting time quality requirements hardly be done manual approaches. This paper highlights some major issues managing ML-services both training inference modes presents initial solutions able meet set minimum inputs. A preliminary evaluation demonstrates our allow systems become efficient predictable respect their response accuracy.