Network Optimization Through Traffic Patterns and Efficient Bandwidth Consumption in Large DCN

Abstract

This paper presents large-scale analysis of traffic patterns in eighteen data centres (DCs). Data centre network (DCN) design emerges with improved architecture, as very little real information is available on existing DC traffic and characteristics. We analyzed traffic patterns of different organizations operating diverse applications. Data gathered for this analysis is over a time span of one year to observe the traffic patterns and present future predictability in network behavior. The paper presents traffic patterns and application behaviours of different applications for the observed time frames. Time frame as long-range dependence (LRD) that is maximum time span of one complete year and minimum of second time frame in small time scale (STC). The goal is to propose a traffic pattern oriented efficient bandwidth utilization and load balance by minimizing energy consumption. Our proposed framework is based on precise analysis of traffic parameters for large DCN. To improve bandwidth consumption and efficient link utilization of a network by defining a baseline threshold value that manages load balance for DC internal Traffic Engineering (TE).