Micro-Scheduling for Dependable Resources Allocation

In this work, we introduce a general approach for slot selection and co-allocation algorithms for parallel jobs in distributed computing with non-dedicated and heterogeneous resources. Parallel job scheduling provides many opportunities for the resources allocation and usage efficiency optimization. Firstly, there are many options to select the appropriate set of resources based on primary target criteria in a knapsack-like problem. The secondary optimization, or , is possible when selecting over a variety of suitable resources providing the same primary target criteria values. Micro-scheduling step usually relies on the resources meta-features, secondary parameters and their actual utilization. Such two-level optimization may be used to obtain heuristic solutions for many scheduling problems. In this paper we present micro-scheduling applications for the dependable and coordinated resources co-allocation, resources usage efficiency optimization, preference-based and fair scheduling implementations.