The Research And Implementation Of A Client-oriented Hot Standby

Currently, with the increasingly rapid developments of computers and networks, the number of the company's core businesses becomes much bigger than before. Many critical business requirements for system availability are increasing. How to ensure the system run as usual and to ensure the application not to be affected has been more and more concerned when the servers subject to some failures.The existing methods of achieving high system availability are cold backup technology, hot standby technology, cluster technology and load balancing technology. Though, these technologies have their own advantages, there are still some inadequacies. With the continuous development of virtualization technology and the continuous use of virtual environment, more and more enterprises run their critical businesses on the Virtualized servers, and how to make hot standby on Virtualized servers and to ensure the application on Virtualized servers not to be affected when Virtualized servers have some failures are more concerned by people. There are some existing hot backup modes of virtual machines such as the Remus of Xen virtual machine, but this system may get a wrong decision when the heartbeat does not work normally, and causes malfunctions of the system. In this article we propose a client-oriented Hot Standby, based on the traditional hot standby mode and the Remus of Xen virtual machine.This mechanism introduces the third-party client, retaining the advantages of the original heartbeat mechanism of Remus. When the heartbeat is working between the main and the standby virtual machine, the standby server and the client communicate with each other, the backup server can get the main virtual machine state from the client server. If heartbeat does not work, the standby server can obtain the main virtual machine state and determines whether the backup virtual machine based on the application getting from the main virtual machine. This process can reduce the wrong decision in hot standby system and ensure that the system work normally.Then we implement the mechanism in the Linux operating system. By testing the system, we verify that the system can significantly reduce the wrong decision in hot standby system.