Please use this identifier to cite or link to this item:
Title: Energy Efficient workload Distribution in Geographically Distributed Data Centers
Authors: Khalil, Muhammad Imran Khan
Keywords: Physical Sciences
Computer & IT
Issue Date: 2020
Publisher: University of Engineering & Technology Peshawar
Abstract: Power management in geographically distributed data centers is a contemporary and challenging issue for Cloud Service Providers (CSPs). In order to satisfy user demands, these data centers (DCs) consume a large amount of energy, resulting in high electricity costs and adverse environmental impact. To minimize the energy cost, current studies indicate the workload allocation strategies among geographical distributed DCs mostly focusing on the use of renewable energy and cheaper rates of electricity. In this work, we investigate the problem of energy cost minimization for geographically distributed data centers with the guaranteed quality of service (i.e., service delay) under time-varying system dynamics. The work is based on three different aspects.First, we summarize various optimization-based workload distribution strategies and optimization techniques proposed in recent research works based on commonly used optimization factors such as workload type, load balancer, availability of renewable energy, energy storage and data center server’s specification in geographically distributed data centers. The survey presents a systemized and a novel taxonomy of workload distribution in data centers. Secondly, we propose a green geographical load balancing (GreenGLB) online algorithm based on the greedy algorithm design technique for the interactive and indivisible workload distribution. An indivisible workload is a sequential task, which cannot be further divided and must be assigned to a single data center. The basic idea of our algorithm is to assign the incoming workload at each time considering the current offered prices of electricity, the renewable energy levels, and respecting the given set of constraints. The experimental results based on the real-world traces illustrate the effectiveness of GreenGLB over the existing workload distribution techniques and attain a significant reduction in the energy cost of the geo-distributed DCs. The final aspect of the study deals with the energy cost minimization problem of geo-distributed DCs considering call option in the electricity derivative market under time-varying system dynamics. A call option is an agreement between CSPs and electricity suppliers, which gives the rights (not obligation) to option holders to purchase a specified amount of electricity over a specified time at a predefined fixed price. To minimize the energy cost, we propose an online algorithm for interactive workload distribution called option pricing based geographical load balancing (OptionGLB). The basic idea behind OptionGLB is to allocate the workload at each discrete timeslot to a DC based on maximum battery power, where the option strike price is less than the spot price in a real market, and/or cheapest prices of brown energy in real-time market. The experimental evaluation based on the real-world data indicates the efficacy of OptionGLB over current workload allocation strategies.
Gov't Doc #: 20198
Appears in Collections:PhD Thesis of All Public / Private Sector Universities / DAIs.

Files in This Item:
File Description SizeFormat 
Muhammad Imran Khan Khalil Computer Sci 2020 UET peshwar prr.docxphd.Thesis2.53 MBMicrosoft Word XMLView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.