Fault tolerance approach for crash failures in cloud environment

Abstract

Fault tolerance is an important issue in the Cloud Computing which is mainly concerned with providing the guarantee of availability and reliability of a system. It also checks a system that must work its functions in a proper manner and data must not lose even some crash / some failure of components of the system occurs. Therefore, Fault tolerance of Cloud Computing provides an environment in which a system can work in a corrective manner using the fault handle techniques. The main benefit of implementing fault tolerance of Cloud Computing is to provide the failure recovery of the components, the cheapest cost of handling fault and also metrics for improving performance of a system. In Cloud Computing, resources are available as services like Software as service (SAAS), Platform as resources (PAAS) and Infrastructure as a service (IAAS) on private and public networks. Cloud Computing has the ability to use applications on the Internet that store and protect data while providing a service. It can be used as storage Cloud to hold application, business, and personal data. In this thesis, various fault tolerance techniques have been explored to handle the crash failure of Name Node, further, the analysis of cost effective tools like HA proxy server and NFS of Linux. NFS is the repository for Name Node System data in the Cluster. The framework for the fault tolerance approach has been purposed that will handle crash failures of Name Node in the Cluster. In this framework, VM Workstation has created a cluster of Linux nodes. In this cluster, a Node HAProxy server has been installed to monitors all servers which have been providing the services on Hadoop. Various Hadoop Name servers have been created for handling the crash of Name node. These nodes have acted as Backup Name Nodes Server for fault tolerance. There has been one Linux server which has provided the services of NFS share for common repository Hadoop server File System metadata. In the thesis, use of data scheduling, job migration and data replication technique have given the solution. HAProxy has been used for handling migration of jobs of Hadoop servers in the case of Name Node Failure. HAProxy has been migrated jobs from one Name Node to another Name Node in the case of Name Node Failure. NFS Server has been provided the common repository for Name Node File System. Using this proposed Framework, services of Hadoop Server has been resumed with any loss of data. The experimental results show that the use of load balancing and distributed files system techniques handles the framework of the Name Node crash for this purpose