Research And Implementation Of Management Software For High Available Server

With the advent of the development of mobile digital era and the big data era,computers and networks have become our inseparable part of daily life.High-performance server achieved by multi-machine collaboration has now beenwidely applied to various fields, at the same time the impact of such a server failurecaused the emergence of increasingly serious. Therefore, the study of highavailability server technology, ensuring a stable server to provide users withcontinuous service becomes increasingly important. High availability can beachieved from a hardware and software perspective. Highly reliable hardwaredesign using the new technology makes the hardware itself has a lower failurerate.At the same time, the pursuit of improving the performance of the hardwareitself will lead to higher failure rate, which brought great difficulties to thehardware design.What is more, it improves the cost of high availability hardwaredesign. So people tend to use management software to achieve high availability onmulti-server platform.In this paper, research about the existing high-availability technology is basedon the fault tolerance and load balancing.The author has designed the managementsoftware and implemented on the Kylin system, choosing the right technologywhich matched the characteristics of the actual server architecture.The author hasdesigned two different heartbeat diagnostic and fault handling mechanism for theLoad balancing layers and the real nodes layers. On the load balancing layer,ensured the rapid diagnosis of node failure and node switching and designed themodule on the reference of VRRP(network redundancy router protocols). Themethod maintained a single external IP which can shield the internal structure of thesystem; On the real working node layer, have used accelerated heartbeat mechanismto decrease the failure of false detection rate, while achieving real node failure faultisolation. The paper focused on the deficiencies of the existing load balancingprogram, used a new way via virtual switch ports to achieve cross-LAN NAT modewhich can cope with the back-end servers distributed in different LANs better. Inthe design, the paper useed a modular design to assign functions to differentmodules, and provided a Web page combined with command line interface for thesystem administrator to monitor the state.