Roman Yasinovskyy, Alexander L. Wijesinha, Ramesh Karne To study the effect of NAT traversal, we also use Teredo. The IPsec scenarios that are evaluated include no-security (in which traffic bypasses IPsec), network-to-network (in which traffic is tunneled between IPsec gateways), and client-to-network (in which traffic is tunneled between the client and IPsec gateway). We use the popular Openswan implementation of IPsec and focus on ESP with the authentication option. The measures used for evaluating VoIP performance are delta (packet inter-arrival time), jitter, packet loss, throughput, and Mean Opinion Score (MOS). We also determine the time for the IPsec key exchange and call set up using SIP. Our results indicate that VoIP performance with IPsec in IPv4-IPv6 transition networks is not significantly different from that in today’s IPv4 networks. In particular, we find that 1) performance with IPv4, IPv6, or 6to4 is similar; 2) the overhead due to NAT traversal with Teredo is comparable to that when using NAT with 6to4 on the edge device; 3) performance degrades significantly when the amount of background traffic exceeds network capacity regardless of whether IPv4, IPv6, 6to4, or Teredo is used.
VoIP performance with IPsec in IPv4-IPv6 transition networks