Empirical Performance of IPv6 vs. IPv4

 under a Dual-Stack Environment

 

 

 


 Home

 Measurement Setup

 Dual-Stack List

 Connectivity

 Hop Count

 RTT

 Throughput

 OS Dependence

 IPv6 Address

 Provisioning

 IPv6 Tunnel

 Performance

 Scripts

 References

 Internal Access

 

Measurement Setup

The measurement study reported in this work was conducted between late 2006 and the first half of 2007. Our measurement setup consists of seven IPv6/IPv4 dual-stack machines as shown in Fig. 1. Our gateway is a dual-boot machine, running Windows XP and Fedora Core 5. Our client machines are named ”BSD1-BSD4” and ”Fedora2-Fedora3”, with the BSD client machines running FreeBSD 6.1 and the Fedora client machines running Fedora Core 5. The network interfaces of each machine are associated with one public IPv4 address and one IPv6 address. Our testbed is connected by a 100Mbps link to the public IPv4 Internet via the Hongkong Advanced Research Network (HARNET). For IPv6, a 45Mbps link to Internet2 is also provided by HARNET, which is shared with 7 other universities in Hong Kong. Our measurements are designed to investigate the performance differences perceived by dual-stack hosts as they choose between IPv6 and IPv4. Each measurement test is conducted between one of the dual-stack machines in our testbed and another publicly available dual-stack host site in the world. We have compiled a list of 10,534 sites with IPv6 and IPv4 capabilities from the following sources: the list maintained in [8], and the ones by major ISPs and DNS servers including WIDE, IIJ, Consulintel, and ISC [5]. We then used ping and ping6 to determine the sites’ reachability. As shown in Table I, amongst the 10,534 sites claimed to have dual-stack capabilities, only 2,014 sites are found to be reachable via both IPv6 and IPv4. These 2,014 sites thus form our list of dual-stack sites used in our measurement tests. In our study, every measurement test is repeated 10 times during different times of the day and different days of the week, and the reported results are averaged over these runs.

Fig. 1. Setup of the CUHK IPv6/IPv4 Testbed

 

Table I. Dual-Stack Test Results

 

Pingable by IPv6

Yes

No

Pingable by IPv4

Yes

2014

7054

No

208

1103