A few years ago the MBone was a new and small enough system that it that maintaining detailed maps of its topology did not require a sophisticated infrastructure. The early map generating efforts, which culminated in a four-page PostScript printout requiring hand-tiling, were abandoned in mid-1993 [1]. Since then, the topology of the MBone has only been available through lists of tunnels and its detailed geography has not been drawn.
We build on the work of researchers at the University of Cambridge, who maintain a publicly available list of the topology of the MBone [4]. This data is generated by the mwatch program developed at University College London which lists the IP addresses (and often hostnames) of multicast routers at tunnel endpoints. We have augmented the Cambridge list with information about MBone routers in some of the firewalled private networks.
Hidden within this large data set are highly suboptimal tunnel placements, misconfigurations, and outdated parts of the MBone topology that should be removed. However, without effective visualization tools, it is almost impossible to sort through this set and derive conclusions about how to improve the current configuration. As of June 1996, the MBone topology data set consisted of over 75 pages of textual data as shown in Table 1.
Table 1: example MBone topology data from University of Cambridge
Since the end goal is to analyze and improve the distribution efficiency of the MBone overlay on the Internet, the most effective method would be to draw the tunnels as segments corresponding to each unicast hop the tunnel traverses. However, determining the unicast path would require dealing with a much larger volume of connectivity information as well as determining the current routing policies of the major providers, a prohibitive task at this point.
We therefore made the simplifying assumption that the logical unicast topology was closely related to the physical geography, and use a geographic-based visualization to draw conclusions about the overall structure. Figure 1 shows a 3D interactive map of the MBone tunnel structure.
There has been previous work in geographical network visualization, most prominently by Eick et al. at Bell Labs. This group has presented the 2D interactive SeeNet system for realtime visualization of telephone networks and offline visualization of Internet traffic [2]. The more recent SeeNet3D system visualizes worldwide Internet traffic using a visual metaphor of arcs on the globe [6]. Our current work uses the same visual metaphor, but in a different context. The influential work of Cox and Patterson at NCSA was a videotape rather than interactive software.
This paper is a case study: our primary goal is incremental progress toward solving a real world problem using appropriate visualization techniques. Our contributions include the construction of an accurate database of locations indexed by IP address and domain name, and an extension of the ``arcs on globe'' representation to include hypertext links in the 3D representation of the arcs.