An implementation of LVS-based highly-available Radius server

The goal of Linux Virtual Server (LVS) is to provide a basic framework to build highly available and scalable network services such as Radius authentication and accounting in our implementation. Scalability is achieved by transparently inserting or removing a node in the cluster, and high availability is supported by detecting node or daemon failures and reconfiguring the system appropriately. This article, rather than proposing a new cluster technique, illustrates the basic concept of LVS, an implementation of LVS-based highly-available Radius server, and the performance testing and analysis of the whole system by a testing software named testrad developed by us. Through the benchmarks, we can conclude that the performance of the system can linearly increase along with the increasing number of the real servers before the director gets fully-loaded. Thus, it can provide high availability and scalability for network services.