Implementation and efficiency analysis of composite DNS-metric for dynamic server selection

Abstract

User requirements for a high availability and fast response time of network services require placing more than one server for accessing particular network service, often using multiple communication links and locations. Dynamic Server Selection (DSS) is a new DNS method for the optimal server selection of a multiple available network service that allows dynamic selection of a server on the client side based on the information of the server load and its network topological distance from the client. The server selection is based on the calculations of a composite DNS-metric in which servers, whose IP addresses are sent in a DNS response, are ranked from the optimal to the least suitable. Calculation parameters are server response time, which the client measures for each server independently, and the server load, which is specified by the server administrator and is forwarded to the client together with the rules for calculating the composite DNS-metric. The DSS method has the lowest overall network service response time in comparison with the other four observed methods (Geographical, Hops, Random and RTT) which, in measurements done in a real time environment for two servers accessible by two independent internet links each, have a longer response time from 8.5 to 26.8% compared to DSS. Results of the proposed analytical model for calculating the efficiency index of the DSS method are compared with results of the practical measurements confirming the relevance of the analytical model. In measurements, the DSS method achieved a high average efficiency index of 1.23.