Combinatorial optimisation for sustainable cloud computing

dc.check.embargoformatEmbargo not applicable (If you have not submitted an e-thesis or do not want to request an embargo)en
dc.check.infoNot applicableen
dc.check.opt-outNot applicableen
dc.check.reasonNot applicableen
dc.check.typeNo Embargo Required
dc.contributor.advisorO'Sullivan, Barryen
dc.contributor.advisorMehta, Deepaken
dc.contributor.authorDe Cauwer, Milan
dc.contributor.funderScience Foundation Irelanden
dc.description.abstractEnabled by both software and hardware advances, cloud computing has emerged as an efficient way to leverage economies of scale for building large computational infrastructures over a global network. While the cost of computation has dropped significantly for end users, the infrastructure supporting cloud computing systems has considerable economic and ecological costs. A key challenge for sustainable cloud computing systems in the near future is to maintain control over these costs. Amid the complexity of cloud computing systems, a cost analysis reveals a complex relationship between the infrastructure supporting actual computation on a physical level and how these physical assets are utilised. The central question tackled in this dissertation is how to best utilise these assets through efficient workload management policies. In recent years, workload consolidation has emerged as an effective approach to increase the efficiency of cloud systems. We propose to address aspects of this challenge by leveraging techniques from the realm of mathematical modeling and combinatorial optimisation. We introduce a novel combinatorial optimisation problem suitable for modeling core consolidation problems arising in workload management in data centres. This problem extends on the well-known bin packing problem. We develop competing models and optimisation techniques to solve this offline packing problem with state-of-the-art solvers. We then cast this newly defined combinatorial optimisation problem in an semi-online setting for which we propose an efficient assignment policy that is able to produce solutions for the semi-online problem in a competitive computational time. Stochastic aspects, which are often faced by cloud providers, are introduced in a richer model. We then show how predictive methods can help decision makers dealing with uncertainty in such dynamic and heterogeneous systems. We explore a similar but relaxed problem falling within the scope of proactive consolidation. This is a relaxed consolidation problem in which one decides which, when and where workload should be migrated to retain minimum energy cost. Finally, we discuss ongoing efforts to model and characterise the combinatorial hardness of bin packing instances, which in turn will be useful to study the various packing problems found in cloud computing environments.en
dc.description.statusNot peer revieweden
dc.description.versionAccepted Version
dc.identifier.citationDe Cauwer, M. 2018. Combinatorial optimisation for sustainable cloud computing. PhD Thesis, University College Cork.en
dc.publisherUniversity College Corken
dc.relation.projectinfo:eu-repo/grantAgreement/SFI/SFI Principal Investigator Programme (PI)/10/IN.1/I3032/IE/New Paradigms in Constraint Programming: Applications in Data Centres/en
dc.relation.projectinfo:eu-repo/grantAgreement/SFI/SFI Research Centres/12/RC/2289/IE/INSIGHT - Irelands Big Data and Analytics Research Centre/en
dc.rights© 2018, Milan De Cauwer.en
dc.subjectCombinatorial optimisationen
dc.subjectCloud computingen
dc.subjectWorkload consolidationen
dc.subjectBin packingen
dc.titleCombinatorial optimisation for sustainable cloud computingen
dc.typeDoctoral thesisen
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
2.7 MB
Adobe Portable Document Format
Full Text E-thesis
License bundle
Now showing 1 - 1 of 1
Thumbnail Image
5.62 KB
Item-specific license agreed upon to submission