Talk

PC2 Kolloquium: Automated provisioning of virtual machines in a cloud environment

Event dates:	13 July 2011 15:00
Organizer:	Paderborn Center for Parallel Computing
Location:	F0.346 (PC2 User Center)

Masterarbeitsvortrag von Herrn Alexander Krieger

<style> </style> One of the most challenging problems for data centers is the cost efficient utilization of hardware and software resources. Large data centers have a considerable interest in autonomic resource management solutions which reduce the degree of human intervention in provisioning processes. One of the challenges facing the SaaS providers is to ensure predefined performance goals (Service Level Objectives) in a stochastic environment. Various attributes like application parameters and the sharing of resources affect the overall performance behavior of a service in an unpredictable way. The purpose of this thesis is to describe and examine a machine learning approach based on Support Vector Machines (SVM) that can deal with unpredictable workload fluctuations in a horizontal scaling scenario and ensure a predefined performance goal (SLO) is met.