Towards cost-aware Ca- pacity provisioning and load balancing in fault-tolerant dis- tributed data centers
No Thumbnail Available
Date
2017
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Many critical e-commerce and nancial services are deployed on geo-distributed data centers (GDCs) for scalability and availability. Recent market surveys show that failures are common in the data centers and this results in a huge nancial loss. Designing data centers for high availability includes spare capacity provisioning across the data centers. The work in this thesis addresses the problems of cost- aware capacity provisioning and load balancing in fault-tolerant GDCs (to mask the failures at a site). We propose optimization models for cost-e ective planning and operation of the GDCs and propose algorithms for solving these. First, we propose an optimization model to distribute the servers across the GDC such that, the total cost of ownership (TCO) for an operator is minimized. The model identi es the optimal server distribution and optimal request routing policy to exploit the spatio- temporal variation in the electricity prices and user demand for minimizing the TCO. Next, we extend the optimization model for capacity planning in GDCs collocated with renewable energy sources. Using this model, the operators can reduce their carbon footprint by maximizing the green energy usage, while minimizing the TCO. We also extend this model to consider GDCs powered by both brown and green energy sources. In such a case, we use an objective to minimize the total cost while ensuring that a certain percentage of green energy is always used.
Description
Supervisor: T. Venkatesh
Keywords
COMPUTER SCIENCE AND ENGINEERING