Empirical prediction models for adaptive resource provisioning in the cloud.
Abstract
Cloud computing allows dynamic resource scaling for enterprise online transaction systems, one of the key characteristics that differentiates the cloud from the traditional computing paradigm. However, initializing a new virtual instance in a cloud is not instantaneous; cloud hosting platforms introduce several minutes delay in the hardware resource allocation. In this paper, we develop prediction-based resource measurement and provisioning strategies using Neural Network and Linear Regression to satisfy upcoming resource demands. Experimental results demonstrate that the proposed technique offers more adaptive resource management for applications hosted in the cloud environment, an important mechanism to achieve ondemand resource allocation in the cloud. © 2011 Elsevier B.V. All rights reserved. 1. Introduction Cloud computing is one of the most popular buzzwords in today's enterprise. An intrinsic feature of the cloud that differentiates it from traditional hosting services is its seemingly infinite amount of resource capacity (e.g. CPU, memory, Network I/O, disk etc.) offered at a competitive rate. It provides opportunities for start-up companies to host their applications in the cloud; thus, eliminating the overhead of procuring traditional infrastructure resources which typically takes several months. However, a near-infinite resource pool for scale and flexibility is not the only potential of the cloud. Cloud hosting providers offer this near-infinite resource on demand using different pricing models, e.g. pay-per-use model for workloads with unforeseeable characteristics, reserved instance pricing model with long-term commitment of availability and spot instance model for workloads with flexible completion time. Therefore, application providers are able to choose an appropriate pricing model based on the anticipated workload characteristics and provision the resources accordingly in the cloud. The pay-as-you-go model and dynamic resource provisioning features of the cloud reduce the overhead associated with static provisioning, which is not considered as ∗ Corresponding author at: National ICT Australia, Sydney, Australia. Tel.: +61 293762194. E-mail ad