Predicting DNS Server Load Distribution

The Domain Name System (DNS) resolution is usually served by multiple geographically distant servers. In planning and optimizing DNS server number, placement, and capacity, it is important to predict server load distribution given some knowledge about the network locations and query rates of active caching resolvers. This poster proposes an analytical model for predicting the DNS server load distribution. The model identifies the dependencies among server load distribution, server selection behavior, DNS response RTT, request processing delay and propagation delay. The preliminary simulation results show that DNS server load distribution can be predicted by solving the model using the iterative approach.