Performance Measurement and Analysis of High-Availability Clusters

Abstract

High-availability clusters are groups of servers that provide a reliable framework for applications to achieve a minimum downtime and quick recovery time without any human intervention and yet are completely opaque to the users. Almost all industries are continuously pursuing the goal to minimize their critical application downtimes by using various techniques such as fault tolerance, redundancy, mirroring and clustering. During downtime, applications become unavailable to end users, which can lead to financial, reputation and regulatory business impacts. High-availability clusters provide a mechanism to migrate complete applications or services from one server to another seamlessly without any human intervention at the time of failure of any critical component or the complete server. Hence, an application would start on a healthy server, without end users realizing this failover. In this work, key features and aspects of two cluster products, 'Symantec Veritas Cluster Suite' and 'Red Hat Cluster' were compared against each other based on various parameters. A simulated environment was created to perform a comprehensive analysis of performances of both products. In this work, measurement of average failover time was taken and compared as the key reliability and serviceability attribute. Thus, based on this experimental work, it is concluded that in a controlled test environment running a simple web-server application, Red Hat cluster gives a better failover performance as compared to Veritas Cluster Suite. However, in a large-scale environment, if we consider factors like operating system compatibility, supported applications, compatibility with volume managers, hardware compatibility, reliability, fault tolerance, predictive failure, self-healing etc., then the Veritas Cluster Suite comes out as the winner.