Simulation framework for carbon-aware workload shifting in the cloud
Loading...
URL
Journal Title
Journal ISSN
Volume Title
School of Science |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
2024-10-28
Department
Major/Subject
Security and Cloud Computing
Mcode
Degree programme
Master's Programme in Security and Cloud Computing
Language
en
Pages
58
Series
Abstract
Carbon-aware workload shifting algorithms are a method to utilize the global scale and flexibility of computing resources offered by cloud computing concepts to minimize the greenhouse gases emitted from the computation tasks. Specifically, the same computational work can produce more or less emissions depending on the geographic location or time when the work is done. Some workload shifting algorithms have already been implemented by major cloud providers like Google, but they are continually iterated upon and improved. This results in a myriad of workload shifting algorithms with different performances and specialties evaluated in custom simulators with different network layouts. In our thesis, we design and develop a simulation framework which can test and compare carbon-aware workload shifting algorithms to aid their development and to determine what scenarios a given algorithm may perform well in. The performance of the algorithms are measured via the amount of carbon dioxide equivalent greenhouse gases produced during a simulation. Some tools already exist to compare algorithms, but they often are highly specialized or put a significant emphasis on the specific resources available in simulated data centers. A core aspect of the framework is the interchangeable carbon intensity data and workloads. The interchangeability of these components save time during the development of workload shifting algorithms by enabling experiments using carbon intensity data from different sources and different workloads. There are common formats for the carbon intensity data and the workloads, a data frame column structure for the former, and a custom API data frame extension for the latter. The common formats ease the compatibility between a given algorithm and the workloads or carbon intensity data not custom designed specifically for the given algorithm. Additionally, this enables two or more algorithms to utilize the same workload without the workload needing to be custom implemented for each algorithm, which in turn supports comparing the algorithms during development.Description
Supervisor
Premsankar, GopikaThesis advisor
Fodor, ViktoriaDán, György
Keywords
Carbon-aware workload shifting algorithms, cloud computing, green computing, data centers, simulation framework, workload shifting