Abstract

Domain name system (DNS) resolution service is usually provisioned by multiple authoritative servers for performance and robustness. Estimating the query load distribution among multiple authoritative servers is one of the key issues arising with DNS server load balancing and optimization. We propose an analytical model of Round-Trip-Time (RTT)-sensitive server selections consisting of cache servers, authoritative servers and clients, which makes it possible to infer DNS server load accurately. A DNS server fingerprint approach is then proposed to identify RTT-sensitive server selections from BIND’s. Finally, we present a server load estimation method based on server selection classification. Under BIND server selection algorithm, the solution of the server selection model is obtained using iteration method, which is validated by the simulation results.

Key words: Round-Trip-Time, domain name system (DNS) server fingerprint, server selection, load distribution estimation.