Research And Implementation Of Management Software Based On Multi-Point-Mutual-Redundancy For High Available Server

Computer server in the key areas such as telecommunications, banking, securities, must maintain the status of continuous and stable operation. And computer server needs provide uninterrupted service for users. If the server goes down it will lead to serious consequences. To solve the above problems, we can adopt the way of improve server availability to achieve continuous service. High availability server system comprising a plurality of compute nodes. There are many backup resources for compute nodes in the system. While computing node fails, the other nodes will take over the service running on the fail node. This paper research the design of management software for high available serve, and propose the availability management based on multi-point mutual redundancy. Under this management mechanism, the software realize the management of computing nodes and system services, ensuring the availability of service. This management mechanism manage the computing nodes with a way of logical service group. Compared with the existing mechanism, this mechanism avoids excessive redundant of backup resources and improves the utilization of computing resources. This mechanism are more suitable to the case that computing resources a re not sufficient.This paper first study the two-node high-availability solution and the N+1 high-availability solution. And it analyzes the existing high-availability server ’s hardware structure. Then it propose the availability management mechanism based on multi-point mutual redundancy. It also analysis the availability of this mechanism compare with existing management solutions. This paper design and implement the high availability management software according to the server’s hardware structure. The software manage the high-availability system based on multi-point mutual redundancy. Computing nodes of server system are divided into several logical services group, and each service group can provide services independently. The software realizes the switching of service node with the virtual IP technology, ensuring that the user access to the service through a unique IP. So the system can block to users that the node logic structure inside the system. In order to monitor the heartbeat of computing nodes, this paper design and implement the heartbeat detection model based on voting. It also proposed the heartbeat detection solution in two-node system to prevent network segmentation. The software can handle four failures: service resource failure, heartbeat failure, failure of shared storage systems and with out-band failure. The paper design the details of the fault-handling mechanisms, and adopt the strategy that failures are handled globally by decision node. Then it describe the detailed definitions of system failure, and design different fault-handling strategies for each failure, realizing the efficient handling of computing nodes’ failures.