...
We have at least one monitoring node in South Korea, Japan, China and Taiwan. With the help of those, we gathered the traceroute results for the neighbours as well as the world neighbouring countries as shown in the figures below.
...
The traceroute results (annotated, raw) originating from China show that China is directly connected to its neighbours i.e. Taiwan (#5, 61 ms), South Korea (#27, 73 ms) and Japan (#8, 100 ms) and the traffic generally takes the direct routes. However in some cases, the results show that traffic destined for selected nodes in South Korea is either routed via Japan (#7, 114 ms) or Japan and US (#16, 209 ms) or Hong Kong and United Kingdom (#54). Similarly, for selected destinations in Japan the traffic destined for Japan is routed via Hong Kong (#24, 206 ms) or Taiwan (#25, >90 ms). Also, traffic to for selected nodes in Taiwan is sometimes directed via South Korea (#27, 245 ms). On the other hand when the destination is within China, the traffic stays within China which is what one would expect. Random tests to other countries in the Asia showed traffic flowing as expected (i.e. as per the submarine links). However the results below were interesting. The RTT values encouraged us to conclude that traffic was indeed traversing these links.
...
The results (annotated, raw) from South Korea show that Korea is directly connected to Japan (#1, 36 ms), Taiwan (#6, 214 ms), China (#11, 110 ms) and the US and the traffic generally takes the direct routes. However some results show that traffic destined for selected nodes in Japan is directed via the US (#23, 182 ms). Similarly, traffic destined for particular nodes in Taiwan is sometimes routed via the US and Australia (#2, >272 ms) and the traffic destined for specific nodes in China is sometimes routed via Hong Kong (#80, 97 ms). Lastly the traffic destined within Korea is observed to stay within Korea as one would expect. Random tests to countries in the Asia showed traffic flowing as expected (i.e. as per the submarine links). However the results below were interesting. The RTT values encouraged us to conclude that traffic was indeed traversing these links.
...
The traceroute results (annotated, raw) originating from Taiwan show that Taiwan is directly connected to its neighbours Japan (#1, 34 ms), China (#17, 61 ms), Korea (#7, 45ms), Hong Kong (#63, >30 ms) as well as the United States and the traffic generally takes the direct routes. However in some cases the results show that traffic destined for selected nodes in Japan is routed via the US (#23, >140ms) or Hong Kong (#74, 73ms). Also, the traffic destined for particular nodes in South Korea is sometimes routed via Japan (#18, >65 ms). As one would expect, the traffic destined for a node within Taiwan stays within Taiwan.
Interestingly though, the traceroute results (annotated, raw) , show that Japan is directly connected to its neighbours i.e. Taiwan (#4, 34ms), Korea (#13, 40ms), China (#14, 130 ms) and nearly all the traffic to destinations in neighbouring countries is routed via the direct linkslinks. In very few cases the traffic destined for selected nodes in Korea is routed via the US (#28, 252 ms). Similarly traffic destined for selected nodes in China is routed via the US (#16, 219 ms). Random tests to countries in the Asia showed traffic flowing as expected (i.e. as per the submarine links). However the results below were interesting. The RTT values encouraged us to conclude that traffic was indeed traversing these links. It is worth noting that traffic destined for Vietnam from both China and Korea is routed via the US, when Japan is directly connected.
Japan to Thailand: Japan -> United States -> Canada -> Thailand
Japan to Vietnam: Japan -> Vietnam
Japan to Singapore: Japan -> Singapore
Japan to Malaysia: Japan -> United States -> Malaysia
The anomalous behaviour of traffic to selected nodes/networks may be attributed to the BGP policies enforced by ISPs.
Comparison with Human Development Indices
...
- Except for Mongolia - which shows average results - all East Asian countries show good (or acceptable) Internet performance. The trends show that these results are continuing to improve.
- Results as seen internationally are strongly influenced by the condition of the network within the country.
- This was confirmed by the step changes seen in performance in 2005/2006 when an Internet Exchange Router was setup connecting the ISPs of Mongolia.
- Nearly all results are skewed by the performance observed for the last few hops.
- Results as seen internationally are strongly influenced by the condition of the network within the country.
- Improvements (/changes) in the national backbones - as in the case of Mongolia in 2005 - result in step changes in Internet performance.
- Internet performance is closely tied with Human Development Indices - HDI.
- This is evident from the strong correlation of these parameters with Internet performance metrics as observed by PingER. In the case of East Asia, Japan, South Korea and Taiwan show good Internet performance results while maintaining high values for HDI. On the other hand, Mongolia shows average (and in some cases below average) Internet performance statistics while maintaining average HDI values.
- Generally the traffic is routed via optimal links. However, traffic for some destinations are routed indirectly even though shorter paths exist (which are used for traffic in general). This may be attributed to the BGP policies defined by the ISPs.
- Increase in PingER coverage results in accurate results. This is observed by the increased correlation of PingER results with Human Development Indices over the years 2006 and 2007.