Benchmarking DNS64 implementations: Theory and practice

RFC 8219 defined benchmarking methodology for IPv6 transition technologies, including DNS64, which is used together with NAT64 to enable IPv6-only clients to communicate with IPv4-only servers. This paper investigates the performances of the most important DNS64 implementations, BIND, PowerDNS, and Unbound as a function of the number of CPU cores using the compulsory tests of RFC 8219. High differences are pointed out both in their single core DNS64 performances and in their scale-up from 1 to 16 CPU cores: whereas Unbound shows the highest single core performance, PowerDNS scales up the best. A serious issue is reported regarding BIND: its DNS64 performance does not increase from 4 to 16 CPU cores at all. A measurement complementary to RFC 8219 is introduced which proves to be useful in the investigation of the issues identified during testing. For the optional tests of RFC 8219, the requirements for the tester are clarified, which enabled us to carry out the tests at the given rates. Finally, a complementary performance metric, computing power relative DNS64 performance is introduced, which may provide further information to network operators to support their selection of the DNS64 implementation, which suits the best for their needs.