Analysis of a Highly Available Fault Tolerant Cluster Architecture

Overview

Large and small enterprises are relying more on the highly available systems for mission-critical applications like media distribution, data warehouses, e-commerce, business intelligence systems, and scientific computing. Automated decision-making and control systems have requirements for high-availability and fault-tolerant computing environments. In addition, computational grids are used for modeling and simulating complex scientific and engineering problems. Clustering technology shows greater promise to solving high availability issues. Cluster technology is used to provide fault tolerance and availability of computer systems running mission critical applications. Clusters have technical benefits that outweigh their comparable higher maintenance cost. Cluster provides high availability that corporate mission critical applications depend on. A cluster is a type of parallel or distributed processing system, which consists of a collection of interconnected computers working together as a single integrated computing resource. The cluster nodes can work collectively as an integrated computing resource, or they can operate as individual computers. Cluster middleware is responsible for offering an illusion of a unified system image and availability using a collection of independent but interconnected computers. Grid integrates networking, communications, computation, and information to provide a virtual platform for computation and information. This research proves that clustering improves availability, redundancy, and takes away the inflexibility of conventional systems. Two distinctive cluster architectures were tested and showed improvement in the high availability cluster framework. Network failover was verified using IP multipathing that was able to achieve complete fault isolation and network recovery. The research also showed measurable improvements in network failure and recovery, node failure and recovery, and storage path failure and recovery. The primary contribution of this research was shown in the significant improvements of high-availability in a clustered system environment. The second contribution of this research is that hardware component relocation within a PCI bus and fine tuning access to the PCI bus showed immediate improvement on high availability. Finally, by designing a redirect router, cluster resource failover issues due to network router overload was eliminated, improving high availability.