Geo-Spatial

This visualization is nicknamed the Battleship view after the popular board game of the same name. The background is a cloudless composite photograph of the Earth. The plane in the foreground is a visualization of campus geography based on GIS data. Network connections are drawn as arcs between their endpoints. Some arcs connect two on-campus hosts and others connect an on-campus host to a location in the rest of the world. On-campus locations are are determined user our Topology Mapper which locates the edge port serving each host on the network, and in turn, the physical location of those ports. Off-campus locations are determined using a commercial service that maps IP addresses to latitude and longitude.

Space Shield

The Space Shield visualization moves from geographic location to network location. On-campus hosts are located on the center plane using polar coordinates that represent their IP addresses. External hosts are located on a hemispherical dome. From the perspective of an internal host, this hemisphere is like the night sky --- a constellation of points. This external hemisphere can be changed between different views based on the external hosts' IP addresses, their geographic locations, or their location in a BubbleTree layout (see below).

BubbleTree

The BubbleTree uses a force-directed layout algorithm developed in conjunction with the University of New Mexico and based on the Lennard-Jones potential used in molecular modeling. It is efficient to compute and provides an automatic layout for hierarchical data such as DNS names. Each top-level domain is one of the larger circles. Within that domain and circle are a set of other circles representing sub-domains such as acm.org and ieee.org in the .org circle. The layout is recursive and supports arbitrary hierarchy depths. Some country-code domains (like .uk and .jp) exhibit an internal structure that is a microcosm of the Internets domain structure for commercial, educational, and governmental domains.

Hive Time Series

Hive is a system for statistical feature extraction from network flow data. Hive data is high-dimensional. This visualization projects data-points onto three dimensions. We animate movement of points over time and provide trails showing the recent path of each point.

The user has on-the-fly control of which dimensions are used and when dimensions are changed, the visualization does a smooth morphing from one projection to another. We also have a "Random Walk" mode that automatically cycles through different projections.

Hyper-Box Viewer

The HyperBox viewer projects high-dimensional bounding volumes (Hyper Boxes) onto a 3-D coordinate space.

Parallel Coordinates

Parallel Coordinates are a well-known technique for displaying high-dimensional data points. A traditional, Cartesian scatter-plot of high-dimensional points only shows two or three dimensions at a time. A parallel coordinate view represents each data point as a line that traverses a series of parallel axes. Each line crosses each axis at the location determined by the point's value in that dimension. With this view you can see all dimensions at once and examine the correlation between adjacent dimensions.

Our parallel coordinate visualization also displays hyper-boxes as semi-transparent regions in this space.