Scheduling elastic machine learning process through containers

Abstract

The following objectives were established for this thesis: • Modification of the resource management scheme within an application/framework to allow dynamic selection of the parallelism of operations (elasticity). • Design and implementation of an internal controller for each container to manage dynamically allocated computational resources, and a communication mechanism between the system and the application confined in the container. • Design and implementation of a scheduler for containers running elastic applications using orchestration techniques to efficiently manage the computational resources of the system.