High Availability Cluster Failover Mechanism Using Artificial Neural Networks

Cloud computing is popular for its utility based computing model for business. As we consume compute as a utility, There comes a question about Availability and reliability of the service. Business requires high available computation/application infrastructure which provides nearly continuous application availability. Many High Availability applications in market are configured into a cluster, which typically involves at least two systems or more. The cluster monitors the critical resources for changes that may indicate a failure and trigger fall over operation to one of the cluster system based on policies (explained in detail in further sections). When a high available cluster is deployed having large number of systems as a part of cluster, it becomes cumbersome for a cluster to monitor the health and to detect the best available node for the resource to fall over. In Cloud environment where cluster systems can be used for other applications during idle time, these policies may not serve the purpose for fall over as the required CPU/memory/network for the highly available application may fall short because of the applications already running on the fall over node. So in this paper we present the way to detect the best system on which application has to fall over using artificial neural networks. Based on the parameters such as available CPU, Memory, Network load. Artificial neural network monitors the nodes and chooses the best node on which application to fallover.