Design and Implementation of Task Scheduling Planner

Proper resource allocation provides efficient matchmaking of resources and tasks and allows researchers to reduce the costs of computations. This paper therefore is both a comparative study of primitive and adaptive scheduling policies, and implementation of the second one. The main goal of the project is to implement the moldabble scheduling approach with quota constraints for jobs (tasks) and resources, employing machine learning methods in a distributed computing system - Skygrid. The project represents gRPC protocol description, Go implementation of FIFO scheduling technique with quota constraints and the system with tasks which were specified to run on the unfixed number of CPUs. The comparative analysis explored the role of moldable scheduling in a distributed computing system approach. General applicability is defined by scientists needs for prototyping new experiments at CERN. However, because of technical features, the project is reproducible and could be fitted to any distributed computer system for scheduling and resource management purposes. Other applications are also discussed.