B. Huffaker, M. Fomenkov, D. Moore (CAIDA, SDSC, UCSD) and E. Nemeth (U. Colorado)

This paper discusses an analysis of the Asia-Pacific network connectivity that focuses on performance, AS and country peering, and third party transit.

As the Internet grows relentlessly, so do the difficulties in monitoring and understanding its complexity. In pursuit of greater insight into the shape of the infrastructure, CAIDA is carrying out active measurements that capture and track a cross-section of the global Internet topology. The skitter tool developed by CAIDA measures the forward IP path and round-trip-time (RTT) from a source to many destinations. It functions similarly to a lighweight traceroute using ICMP echo request probes - rather than traceroute's default UDP probes, and accurate kernel timestamps. An essential design goal of skitter is to do its pervasive measurements while incurring minimal load on the network infrastructure and on the final destination hosts. In line with this goal, skitter probe packets are small (52 bytes), and we restrict the frequency of probing.

The ten skitter hosts used in this analysis are located in San Diego, San Jose, and Washington, DC, two in Tokyo, and one each in London, Toronto, Taejon (Korea), Singapore and Hamilton (New Zealand). Each source queries about 29,000 destinations all over the world including about 2000 Asia-Pacific destinations. The skitter hosts gather path and RTT data continuously, probing each destination on approximately 1.5 to 2 hour intervals. Losses of connectivity are also recorded.

Our data yield insights into the Internet topology at multiple logical and geopolitical scales: from individual destinations, to entire path, to Autonomous Systems (AS's) traversed, to countries visited. Specific combinations of scale and measurement type answer different questions about the network infrastructure in this part of the world:

• By studying distributions of path lengths and RTTs at different geographical locations we obtain quantitative metrics of connectivity. That, in turn, allows us to identify typical and atypical behavior of traffic throughout the region studied.
• By examining best-case/typical/worst-case RTTs and loss rates between countries we can assess general Internet performance and, possibly, suggest the areas where infrastructure improvements are needed.
• By looking at the topology at the AS level, we can infer some peering relationships and shed light on policy based routing.
• By analyzing our test traffic between two countries (or even from a country to itself), we can determine if a significant portion of it is ever carried by other intermediate countries. Awareness of this situation is important to policy makers, businesses, and individuals, who care about influences those third party countries may have.