PROJECT TITLE :

Predictive Auto-scaling of Multi-tier Applications Using Performance Varying Cloud Resources

ABSTRACT:

The performance of the same kind of cloud resources, such as virtual machines (VMs), can change over time for a number of reasons, the most significant of which are hardware heterogeneity, resource contention among co-located VMs, and the overhead of virtualization. The performance variation can be significant, which introduces challenges to the process of learning workload-specific resource provisioning policies in order to automatically scale cloud-hosted applications in order to maintain the desired response time. The fact that bottlenecks can occur simultaneously on multiple tiers makes it even more difficult to auto-scale multi-tier applications while using a minimum of resources. In this paper, we address the problem of using performance varying VMs for gracefully auto-scaling a multi-tier application using minimal resources to handle dynamically increasing workloads and satisfy the response time requirements. Specifically, we look at how this problem can be solved by using virtual machines with different levels of performance. The system that has been proposed makes use of a supervised learning method in order to determine the appropriate resources provisioning for multi-tier applications. This determination is made on the basis of a prediction of the application response time and the rate at which requests are received. The supervised learning method is responsible for learning a state transition configuration map that encodes a resource allocation states that is invariant to the performance variations of the VMs that are running underneath. Utilizing resources with variable performance is made easier with the help of this configuration map. When compared to more traditional predictive auto-scaling methods, our experimental evaluation using a real-world multi-tier web application hosted on a public cloud demonstrates an improved application performance with significantly fewer resources being utilized.