Self-Configuration Framework for Networked Systems and Applications

Abstract

The increased complexity, heterogeneity and the dynamism of networked systems and applications make current configuration and management tools to be ineffective. A new paradigm to dynamically configure and manage large-scale complex and heterogeneous networked systems is critically needed. In this dissertation, we present a self configuration paradigm based on the principles of autonomic computing that can handle efficiently complexity, dynamism and uncertainty in configuring networked systems and their applications. Our approach is based on making any resource/application to operate as an Autonomic Component (that means, it can be self-configured, self-healed, self-optimized and self-protected) by using two software modules: Component Management Interface (CMI) to specify the configuration and operational policies associated with each component and Component Runtime Manager (CRM) that manages the component configurations and operations using the policies defined in CMI. We use several configuration metrics (adaptability, complexity, latency, scalability, overhead, and effectiveness) to evaluate the effectiveness of our self-configuration approach when compared to other configuration techniques. We have used our approach to dynamically configure four systems: Automatic IT system management, Dynamic security configuration of networked systems, Self-management of data backup and disaster recovery system and Automatic security patches download and installation on a large scale test bed. Our experimental results showed that by applying our self-configuration approach, the initial configuration time, the initial configuration complexity and the dynamic configuration complexity can be reduced significantly. For example, the configuration time for security patches download and installation on nine machines is reduced to 4399 seconds from 27193 seconds. Furthermore our system provides most adaptability (e.g., 100% for Snort rule set configuration) comparing to hard coded approach (e.g., 22% for Snort rule set configuration) and can improve the performance of managed system greatly. For example, in data backup and recovery system, our approach can reduce the total cost by 54.1% when network bandwidth decreases. In addition, our framework is scalable and imposes very small overhead (less than 1%) on the managed system.