Graphviz / GMap
Graphviz is an open-source network visualization system. It optimizes layouts for readability and to reveal structure, with the goal of approaching the quality of hand-made diagrams. Although Graphviz can run as a stand-alone system, its programs and libraries were designed to be incorporated in vertical applications for networking, security, software engineering and bioinformatics. It has an extensive set of features and drivers for formatting and displaying real-world diagrams. See graphviz.org for information and downloads.
Recent work is aimed at visualizing larger, denser networks. Many real-world data sets have tens or hundreds of thousands, even millions of objects, presenting new technical challenges in visualization. There are many applications that need the ability to render and interact with networks of this size. Another problem is to cope with the apparently high intrinsic dimensionality of many of these networks. Scalable layout (sfdp) and rendering (smyna) algorithms are two new additions to Graphviz.
Another recent project, GMap, was inspired by a cartographic metaphor for visualizing clusters in networks. We often noticed people trying to find structure in complex networks. For example in a network of TV/video recommendations, they may try to identify regions containing new programs, sports, children's entertainment, documentaries, and action/adventure programs. Why not draw these in an appealing form that simulates a "map" of this landscape? Visit the GMap gallery for examples and demos.
MINGLE, a multilevel agglomerative edge bundling algorithm, attempts to reduce clutters in visualizing large scale networks by bundling similar edges together. In doing so it aims to reduce total screen ink.
Visualizer
Many services run on complex configurations of servers, routers and networks. VizGems provides end-to-end views for near-realtime monitoring of service reliability and performance. It provides data integration in the human interface, combining streams that describe physical assets, logical customer configurations, statistics, and events such as alarms and other conditions. It runs on a set of general-purpose tools for reliable, scalable data acquisition, storage, processing, querying and visualization.
Some of its key attributes are:
- Shows customer impacts of events.
- Visual querying and drill-down.
- Runs in standard web clients.
- Views are generated automatically from data and metadata.
- Integrates realtime and historical data.
- Scales to over 400 million events or transactions in a single view.
- AT&T vcodex compression can store years of raw data, for tasks like life cycle analysis.
VizGems was recently adapted to run on portable devices (like smartphones and the iPad), and a speech-enabled interface ("Show critical events for routers in the Singapore data center") is running as a full-scale prototype.
VizGems manages services, such as hosting, content distribution, managed routers and VoIP for some of AT&T's largest enterprise customers. A marketing presentation explains some of its features and applications.
3Dlab
Can we capture live video of objects and scenes from multiple cameras, and process it to acquire 3D models and images? The 3Lab provides an environment for studying this problem and others that require video processing and high-performance computing. Possible applications are to capture and display 3D scenes remotely, insert objects into the virtual world, or have gaze-corrected video conferences. This is joint work with Vinay Vaishampayan, Amy Reibman, and Chao Tian. Future directions include integration with augmented reality services.
ZLab
Storage, broadband media and computing technology makes it practical to acquire and operate on a vast corpus of thousands of hours of movies and television. The challenge is to create algorithms and programs to search this data (and associated meta-data) to discover interesting patterns and relationships. This project is aimed at understanding the possibilities for future video services.
LiveRAC, Haystack, Ptolemy
These tools are aimed at visualization of time series and maps, at large scale. LiveRAC and Haystack are tools for exploring large sets of time series. LiveRAC, created by visiting researcher Peter MacLachlan, is based on Tamara Munzner's According Drawing framework and provides semantic zooming in a reorderable grid of time series. The grid is stretchable, with guaranteed visibility of critical data items. The display of the time series in a cell adapts to the available area, from color blocks, to sparklines, to detailed charts. A CHI 2008 paper described a Multidimensional In-depth Long-term Case study (MILC) carried out with the cooperation of AT&T Internet hosting Life Cycle Engineers. Haystack is a successor, based on a click-to-zoom metaphor. It has been adapted to trials with the AT&T Darkstar project that monitors over 30,000 core and edge routers. Ptolemy is a complementary service to integrate and display network and other geo-coded events on maps, and is based on open source technologies such as OpenLayers and NASA WorldWind.
