Design And Implementation Of Load Balancer High Availability Solution Based On Consistent Hash Ring

With the rapid development of technology and the emergence of different production pattern and living manner, all kinds of services are showing a explosive growth in a crazy way. In the day when the performance of a single machine is relatively stable,it's a common way to form a cluster using several single-machines. As the entrance of the whole system,load balance services can be used to guarantee some high availability and security.Meanwhile, load balance services become the bottleneck of the application cluster.This thesis focused on solving the high availability problem of load balancers for common service application, and some work has been accomplished as follows:1. When going through the architecture and running procedure of the traditional load balancer high availability, two drawbacks has been discovered: firstly, traditional method uses active/passive plan based on Virtual Router Redundancy Protocol (known as VRRP),because of whose feature the virtual IP transportation between master node and backup node can be very time-consuming and make the whole system unresponsive during this interval.Secondly, there is only one machine running, as master node, and those backup machines are in the idle state at the same time, which leads to a huge waste of physical resources. There still is room for advance.2. Trying to solve those two problems of the traditional load balancer high availability solution, this thesis analyses disadvantage of different solutions, and bring the advanced solution of this thesis. The advanced thesis set up a registry node collection the running information of load balance cluster regularly, and develop a client kit to retrieve the cluster information from registry node for initializing a consistent hash ring based on various machine configuration dynamically. Every request will get a appropriate IP address from the generator of the consistent hash ring, and when the request URL is assembled, the request will be send to the corresponding machine. Furthermore, timeout mechanism is employed to identify bad connections, assuring every client request will be handled timely. On the theoretical framework, this advanced solution can solve the problems of traditional solution.3. Installed on OpenStack cloud platform, the advanced solution for load balance high availability was implemented to test high availability and load balance of the load balance cluster based on different parameters and strategies. Result shows that the advance solution reduces the unresponsive time during the downtime period, and requests can be allocated according to the machine abilities, which means load balance of load balancers has been achieved. Those experimental results show that the advanced high availability solution of load balancer is practicable.