Project: Sustainable Cloud Data Centers

Introduction

The focus of this project is on data center sustainability, to cut operational cost and increase renewable integration for data centers. The computing capacity and scale of data centers are increasing to meet the soaring demand for IT applications and services. Mega data center can host thousands of servers and require up to tens of megawatts of electricity. The high power consumption causes two serious consequences. First, generating and delivering this power to data centers result in large electricity bills. Data center operators may face millions of dollars annually charged by the electrical grid. Second, the enormous energy consumption can lead to negative environmental impacts. Data centers are still heavily dependent on the brown energy drawn from the current electrical grid which produces much of its power by burning carbon intensive fossil fuels. For example, the energy consumption of Google in the year 2010 is equal to that of about two hundred thousand households in US and about two million households in China.

Sub-project 1: Geographical Load Balancing

Background

This project leverages Geographical Load Balancing (GLB) to reduce operational cost and mitigate emissions of geo-distributed data centers. GLB is defined as spatial workload placement among geographically distributed data centers. This method explores the spatial flexibilities of workloads and the spatial diversities of parameters. User requests can be routed to the different geo-distributed data centers. The selected data center processes the request using available resource as required and then returns the result to the user. The routing decision depends on several inter-related parameters including the electricity price, grid emission factor, renewable power generation, etc. For example, the energy cost can be reduced by sending requests to data centers where the electricity price is cheaper (as illustrated by the figure). Emission reduction can be achieved by routing requests to data centers with lower emission factors or with sufficient renewable energy. There are also some constraints to meet including geographical proximity, data center/server capacity, service level agreements, etc. For instance, sending requests to data centers far away from clients results in significant access latency. Overloading a data center with too many requests may incur serious service delay perceived by clients.

Case Study

1. GreenStreaming. This case study uses video streaming systems as an example to demonstrate the effectiveness of GLB framework to reduce energy cost. We have built a small prototype testbed and made a demo video. The demo shows (1) energy cost reduction while maintaining the video quality; (2) preference to assigning requests to data centers with lower electricity price; (3) the effect of bandwidth on the video quality.

We have made a series of demos as follows. These demos show how to leverage GLB framework to reduce the electricity cost for data centers in the context of video streaming systems.

Demo 1: This demo comes with illustrative video players and may provide a intuitive understanding.

Demo 2: This demo comes with more sophisticate charts and may provide a deeper understanding.

Publications

1. Fanxin Kong, Xingjian Lu, and Xue Liu, "Distributed Data Center Bandwidth Allocation for Cloud-Based Streaming", in IEEE Transactions on Sustainable Computing (TSUSC), 2017.

2. Xue Liu and Fanxin Kong, "Datacenter Power Management in Smart Grids", in Foundations and Trends in Electronic Design Automation, now publishers Inc., 2015. (Monograph)

3. Fanxin Kong, Xingjian Lu, Mingyuan Xia, Xue Liu and Haibing Guan, "Distributed Optimal Datacenter Bandwidth Allocation for Dynamic Adaptive Video Streaming", in ACM Multimedia (MM), 2015. (Full paper)

4. Xingjian Lu, Fanxin Kong, Jianwei Yin, Xue Liu, Huiqun Yu and Guisheng Fan, "Geographical Job Scheduling in Data Centers with Heterogeneous Demands and Servers", in IEEE International Conference on Cloud Computing (CLOUD), 2015.

5. Ashikur Rahman, Xue Liu and Fanxin Kong, "A Survey on Geographic Load Balancing based Data Center Power Management in the Smart Grid Environment", in IEEE Communications Surveys and Tutorials (COMST), 2014.

6. Zhonghao Sun, Fanxin Kong, Xue Liu, Xingshe Zhou and Xi Chen, "Intelligent Joint Spatio-temporal Management of Electric Vehicle Charging and Data Center Power Consumption", in 5th International Green Computing Conference (IGCC), 2014.

Sub-project 2: Data Center Renewable Integration

Background

The growing awareness about global climate change has boosted the need to mitigate greenhouse gas emissions and spurred efforts to accelerate the penetration of renewable energy sources (e.g. wind and solar power). This project focuses on integrating renewable for data centers. There are some challenges. First, a fundamental difficulty here is that renewable energy sources are usually of high variability. Data centers must absorb this variability through employing many additional operations (e.g., operating reserves, energy storage), which will largely raise the cost of using renewable energy. To reduce such cost, we explore the temporal and spatial flexibility of data center workload to align with the renewable variability. Second, modern data center operators are incorporating both green energy sources and brown energy sources into data centers' power supply. Challenge arises due to distinct characteristics of energy sources used for different goals. We study how to select optimal energy sources and plan their capacity for data centers to meet cost, emission and service availability requirement.

Case Study

1. GreenPlanning. This case study focuses on the construction or investment phase and optimizes energy source selection and capacity planning for green data centers. We present a framework to strike a judicious balance among multiple energy sources, the electrical grid and energy storage devices for a data center in terms of cost, emission, and service availability.

Publications

1. Fanxin Kong and Xue Liu, "GreenPlanning: Optimal Energy Source Selection and Capacity Planning for Green Datacenters", accepted by the by the ACM/IEEE 7th International Conference on Cyber-Physical Systems (ICCPS), 2016.

2. Fanxin Kong and Xue Liu, "A Survey on Green-Energy-Aware Power Management for Datacenters", in ACM Computing Surveys (CSUR), 2015.

3. Fanxin Kong, Xue Liu and Lei Rao, "Optimal Energy Source Selection and Capacity Planning for Green Datacenters", in Proceedings of the ACM International Conference on Measurement and Modeling of Computer Systems (SIGMETRICS), 2014. (Extended abstract)

4. Chuansheng Dong, Fanxin Kong, Xue Liu and Haibo Zeng, "Green Power Analysis for Geographical Load Balancing Based Datacenters", in 4th International Green Computing Conference (IGCC), 2013.

Devices Used & Testbed Built for the Sustainable Data Centers Project

GLB Testbed (e.g., GreenStreaming Case Study)