Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.632949
Title: Holistic cloud computing environmental quantification and behavioural analysis
Author: Garraghan, Peter Michael
ISNI:       0000 0004 5364 4504
Awarding Body: University of Leeds
Current Institution: University of Leeds
Date of Award: 2014
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
Cloud computing has been characterized to be large-scale multi-tenant systems that are able to dynamically scale-up and scale-down computational resources to consumers with diverse Quality-of-Service requirements. In recent years, a number of dependability and resource management approaches have been proposed for Cloud computing datacenters. However, there is still a lack of real-world Cloud datasets that analyse and extensively model Cloud computing characteristics and quantify their effect on system dimensions such as resource utilization, user behavioural patterns and failure characteristics. This results in two research problems: First, without the holistic analysis of real-world systems Cloud characteristics, their dimensions cannot be quantified resulting in inaccurate research assumptions of Cloud system behaviour. Second, simulated parameters used in state-of-the-art Cloud mechanisms currently rely on theoretical values which do not accurately represent real Cloud systems, as important parameters such as failure times and energy-waste have not been quantified using empirical data. This presents a large gap in terms of practicality and effectiveness between developing and evaluating mechanisms within simulated and real Cloud systems. This thesis presents a comprehensive method and empirical analysis of large-scale production Cloud computing environments in order to quantify system characteristics in terms of consumer submission and resource request patterns, workload behaviour, server utilization and failures. Furthermore, this work identifies areas of operational inefficiency within the system, as well as quantifies the amount of energy waste created due to failures. We discover that 4-10% of all server computation is wasted due to Termination Events, and that failures contribute to approximately 11% of the total datacenter energy waste. These analyses of empirical data enables researchers and Cloud providers an enhanced understanding of real Cloud behaviour and supports system assumptions and provides parameters that can be used to develop and validate the effectiveness of future energy-efficient and dependability mechanisms.
Supervisor: Xu, Jie ; Townend, P. M. Sponsor: EPSRC ; National Natural Science Foundation of China ; National Basic Research Program of China ; STRAPP project (Trusted Digital Spaces through Timely Reliable and Personalized Provenance)
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.632949  DOI: Not available
Share: