Resource provisioning optimisation for cloud computing systems serving multi-class requests

Abstract

With many organisations switching to cloud-based systems, the variety of domains using cloud services is increasing steadily. Often, companies own private cloud computing systems to cater to their own needs and reduce dependency on third-party services. These systems serve a multitude of requests of different classes that often require different types of resources. A queuing network-based cost model has been employed to quantify the economics of the system, and an optimisation problem has been set up. In the cost function, higher-order terms have been used which accounts for diminishing marginal productivity. This law is used in microeconomics to study the behaviour of cost for individual service-producing entities, i.e., firms. In this study, the cost of the system has been optimised by calculating the optimal quantity of resources provisioned to the cloud system and the behaviour of the cost of maintaining a profitable cloud-computing system has been discussed.