AFTM - Agent Based Fault Tolerance Manager in Cloud Environment

Abstract

As the number of cloud users are increasing with times, the probability of failures also increases that takes place in any cloud virtual machine. Failures can occur at any point of time in service delivery. There are numerous techniques for reacting proactively towards these failures. In this framework, a service provider is allocated to the user on the basis of ranking of the service provider. This ranking is done by considering parameters such as trust values (calculated by feedback mechanism), checkpointing overheads, availability and throughput. Checkpoints are beneficial in triggering savepoint so that minimal loss of data takes place if any failure occurs. This paper has also compared the proposed framework with OCI framework which is based on triggering checkpoints on constant rates. Results have proven that AFTM has 33% to 50% better efficiency results as compared to optimal checkpoints interval (OCI) framework. The results shown in paper demonstrates how better the checkpointing overheads, availability and throughput are handled by using AFTM framework. Also, the overheads were reduced to 50% as compared to OCI framework