A Taxonomy of Schedulers -- Operating Systems, Clusters and Big Data Frameworks

TL;DR

This paper provides a comprehensive taxonomy of workload schedulers across operating systems, clusters, and big data frameworks, focusing on their architecture, design factors, and scalability improvements, with special emphasis on Google's Borg.

Contribution

It introduces a hierarchical taxonomy of schedulers based on key design factors influencing throughput and scalability, highlighting incremental improvements and detailed analysis of Google's Borg.

Findings

Hierarchical taxonomy of schedulers based on architecture and design

Key design factors impacting throughput and scalability

In-depth analysis of Google's Borg system

Abstract

This review analyzes deployed and actively used workload schedulers' solutions and presents a taxonomy in which those systems are divided into several hierarchical groups based on their architecture and design. While other taxonomies do exist, this review has focused on the key design factors that affect the throughput and scalability of a given solution, as well as the incremental improvements which bettered such an architecture. This review gives special attention to Google's Borg, which is one of the most advanced and published systems of this kind.