A Novel Approach to Adaptive Flow Scheduling for Energy Efficient Data Center Network

Abstract

Virtualization and cloud computing have changed the computing paradigm and offers everything as service. The backbone of cloud computing is the massive data center servers. The major concern with the cloud data center is the massive amount of energy they consume. The data center network contributes to major share in energy consumption. So approaches to design energy efficient data center networking techniques are gaining importance. This work aims to conserve energy in a Spine leaf data center network. The idea of the proposed Adaptive flow scheduling technique (AFS) is to constantly monitor the traffic characteristics of the data center and accordingly activate or deactivate the spine switches. An adaptive threshold is used to identify the minimum utilized switches and a deadline based threshold is used to identify the switches that have flows with minimum completion time. The identified switches are put to sleep mode. The proposed work is evaluated using the Cloudsim toolkit with traffic patterns that model the real data center traffic. The simulation results effectively demonstrate that the proposed system consumes 15% less energy as against the baseline data centers with static set of active switches and state of art technique.