Energy Savings When Migrating Workloads to the Cloud

TL;DR

This paper analyzes how migrating workloads to the cloud, especially using optimal Lift-and-Shift strategies and auto-scaling, can significantly reduce energy consumption based on real-world data from thousands of machines.

Contribution

It demonstrates the potential energy savings of cloud migration with optimal instance selection and auto-scaling, supported by extensive real-world data analysis.

Findings

Optimal cloud instance selection yields significant energy savings.

Auto-scaling further reduces energy consumption without application refactoring.

Real data from 40,000 machines across 300 data centers supports conclusions.

Abstract

In the cloud environment, data centers are efficiently manipulated by cloud service providers (CSPs) in terms of energy consumption. Consequently, migrating workloads to clouds can result in lower energy consumption. This paper demonstrates that the Lift-and-Shift migration with optimal selections of cloud instances can provide significant energy savings, and explains how much and where the energy savings are obtained from. Additionally, the analysis on the variation of energy consumption is given when Auto-Scaling is deployed showing that further energy savings are expected even without refactoring applications. All the conclusions and analyses are based on the real data collected by Cloudamize Inc. from May 2016 to August 2016 over 40,000 machines across approximately 300 data centers.