Proposed Framework for Fault Prediction and Monitoring in Cloud Environment

Abstract

In cloud computing, users do not have any information about the physical resources like server location, server capacity etc. whenever any fault or failure occurs in any physical resource, vendor has to deal with the problem to provide uninterrupted service to end user. Reliability, availability in Cloud computing are critical requirements to ensure correct and continuous system operation also in the presence of failures. Therefore there is need to design a Fault tolerance framework using fault prediction and monitoring tools. The solution provided in this thesis make use of a software tool known as HAProxy, Hadoop and Nagios that offers high availability and can handle failures in the framework of virtual machines. Using these tools, a Fault prediction and monitoring Framework has been proposed and a prototype is implemented in Linux by using two web server applications that may comprise of faults. HAProxy balances load between these servers and migrates request between server on failure. Also data is stored in Hadoop that provides data replication at multiple nodes. Nagios provides a Fault Prediction mechanism to predict fault in Hadoop before complete cluster goes down. The complete framework provides autonomic and transparent fault tolerance capability to cloud applications. The experimental results show that framework comprising of above mentioned tools can make applications to recover from these faults in a few milliseconds thus increasing system availability and reliability.