Resource Allocation for Maximum Performance and Minimum Cost for 3-tier SaaS Application in Azure

Abstract

Windows Azure is one of the most common commercial clouds, which offers a scalable and elastic platform to host applications. The cloud customers can choose from various number of different type instances according to their needs in order to avoid application bottleneck. Scaling the resources can be done on two ways, i.e., either to increase the instance with additional resources (CPU, RAM, HDD), or to add an additional instances, each with the same resources as the first one. Since the customer should pay different price for the scaling, we are interested which scaling provides better performance, and if we can generalize it. The goal of this paper is to determine which resource organization produces better performance and cost, as well as better price-performance ratio, i.e., if it is better to host the web application in a large number of ”smaller” application instances or small number of ”bigger” instances. The results show that the latter provides much better performance for less price.