Literature Survey: Statistical Characteristics of Google Cluster Trace

Abstract

Usually, data sets used in computer network research are not widely shared because of the risk of security breach. However, with the release of Google Cluster trace (GCT), academics are able to validate their work with a common set of data. Furthermore, it helps researchers to understand the real nature of huge workload in cloud data centres. This paper aims to collect and summarize the characteristics of GCT that has been discovered since its inception in 2011. With the summary, researchers can reduce literature research time and focus on exploring new angles and perspectives of study. This summary can also be used to validate findings, and probably detect any anomalies that might exist. In this paper, features that have been investigated are consolidated, especially in tabulated forms. The focus is on data composition characteristics, namely the characteristics of jobs, tasks, machines and pattern in the cluster.