A Method based on Multi-agent Systems and Passive Replication Technique for Predicting Failures in Cloud Computing

Abstract

Background: Cloud computing refers to the computing capacities of remote computers, where the user has considerable computing power without having to own power units. The probability of failures, which can occur during execution, increases in the number of nodes. Since failures cannot be completely avoided, one solution is to use failure tolerance mechanisms. Predicting failures has become a major task for engineers and software developers, as failure increases resource usage costs. Objectives and Methods: This paper presents a hybrid method of predicting failures in a cloudcomputing environment based on the passive replication technique and multi-agent systems. It detects failures, improves the average response time and minimizes lost time. This method makes it possible to efficiently and transparently guarantee the continuity of cloud computing services in the presence of failures. Results: The results show that the proposed method performs well in the presence of failures, improves the response time and minimizes the additional costs caused by the failures. Conclusion: This paper proposes a hybrid method of predicting failures in cloud-computing based on the passive replication technique and multi-agent systems to detect failures and minimize lost time. The replication technique works by duplicating some system components, which are deployed simultaneously across different resources. This technique aims to make the system robust, increase availability and guarantee the execution of jobs. In addition, it is suitable for long-running tasks.