Autonomic model for self healing and self protection in grid computing using multi agents

Abstract

Grid Computing has evolved as the new level of distributed system by combining large number of dynamic heterogeneous resources to work as single unit. This helps in utilizing enormous power of large number of computational resources to solve highly computational specific problems. However, the complex nature of grid computing, also brings with it the increase in rate of failures and security attacks which are diffcult to handle manually. In this thesis, an automated model for self-healing and self-protection of grid environment using multi-agents is presented to deal with such failures and security attacks. The major contribution of the thesis is a model called SHAPE. SHAPE stands for self healing and protection environment. Self-healing and self-protection properties of autonomic computing are used to provides a holistic approach for the design and development of SHAPE. This helps grid environment to adapt itself to meet the requirements of fault tolerance and security from attacks without manual intervention. SHAPE is novel idea that provides many features: (a) It is the first initiative that provides the capabilities for both Self-Healing and Self-Protection in one system. (b) In terms of fault handling, it uses both active and proactive approaches to provide the functionalities to deal with hardware, software and network failures for distributed systems. (c) For self-protecting system from attacks, it has intelligence to keep updating the network profile of intrusion detection system (IDS) to provide protection from attacks like Distributed denial of Service (DDoS), Remote to Local (R2L), User to Root (U2R), and Probing attacks. (d) From agent-based component design principle, SHAPE is highly scalable, robust and reliable model. Architecture of SHAPE is based on multi-agent component architec- ture which consists of two broad categories namely self-healing and self-protection. Under self healing, di erent algorithms are proposed for monitoring and handling hardware, network and software failures.