Authors: Anders N. Gullhav; Bjørn Nygreen
Addresses: Department of Industrial Economics and Technology Management, Norwegian University of Science and Technology (NTNU), NO-7491, Trondheim, Norway ' Department of Industrial Economics and Technology Management, Norwegian University of Science and Technology (NTNU), NO-7491, Trondheim, Norway
Abstract: A provider of cloud software services is faced with different decisions related to the operation of his services. Firstly, he needs to configure the services such that the quality of service is in accordance with the specified requirements, and secondly, the services need to be deployed in the provider's private cloud in a cost and energy-efficient manner. We argue that these two sets of decisions need to be taken simultaneously in order to make an optimal decision, and we present mixed integer linear programming models optimising these decisions. One of our models also allows the provider to deploy services in a public cloud. We are testing a direct formulation against a reformulation utilising pre-generated node patterns, and observe that the reformulation produces solutions of better quality within a fixed runtime. Furthermore, the effects of bursting services into a public cloud are investigated.
Keywords: cloud computing; private cloud; hybrid cloud; service deployment; software-as-a-service; SaaS; multi-tier services; quality of service; QoS; service replication; fault tolerance; energy efficiency; mixed integer linear programming; MILP; optimisation models; pre-generated columns; cloud services; cloud data centres; reformulation; node patterns.
International Journal of Cloud Computing, 2015 Vol.4 No.2, pp.130 - 149
Available online: 06 May 2015Full-text access for editors Access for subscribers Purchase this article Comment on this article