I believe that the best method to understand how the backbone network we run is working and what the service quority we are providing to our customers is investigating in the real traffic and analyzing it. One concern is intarnational link usage analysis. Because it is essential for us to have international links to outside of Japan to provide connectivity to the world, measuring traffic flow, volume, and performance by source/destination country will give us an idea as to where we should have a link with how much bandwidth, in order to minimize the link costs and maximize overall performance. It is also good to know traffic flow within our domestic backbone to determine our backbone topology. Another concern is traffic characteristics analysis per application services. To know what percentage of our backbone capacity is utilized by what services in usual basis, we can tell what extent we can committ our service quority for each services to our users. Fault analysis would be another concern. The ability to trace back to the past in terms of what type of traffic we had on the backbone during the faulty stage which has already gone by some primitive and non-essential action, just rebooting a router for example, is very important to isolate the real cause of the fault and fix it later. That is also important to analyze the relation between some symptoms and faults, which can lead to realtime fault detection and diagnostics.