Author Information: Graham J. Wills Research, Bell Laboratories (Lucent Technologies) Room 1U334, 1000 E. Warrenville Rd. Naperville IL 60566 Tel: (630) 979 7338 Fax: (630) 713 4982 Stephen G. Eick Technical Manager, Bell Laboratories (Lucent Technologies) Room 1G351, 1000 E. Warrenville Rd. Naperville IL 60566 Tel: (630) 713 5169 Fax: (630) 713 4982 Next Generation Internet Visualization ====================================== The primary characteristics of today's internet are its huge size, explosive growth, and overwhelming complexity, all of which create a formidable barrier to utilizing the Internet at its the full potential. For the service provider it is hard to understand how the network is being used as evidenced by the recent overload caused by flat-rate access tariffs. For the application developer, limited information is available about how best to position new offers. For the user, the challenge is often simply to locate needed information and services. As in any scientific endeavor, the necessary first step is toward addressing this complexity is to better understand it. Information Visualization is one of the few scientific discipline to have this as its core goal (see GE95 for an introduction to Information Visualization). Within Bell Laboratories, the Information Visualization group has been researching this problem for a number of years, developing methodology and tools for the analysis of large databases, with network information a core focus (EiWi93,BEW95, EiWi95, EiFy96). Through our association with a large telecommunications company, we have had the opportunity to research problems involving large, real-life, business-critical network data sets. One goal, for example, of the network research is to be capable of visualizing the telecommunications calling pattern of each individual in the United States within one interactive environment. Our current progress allows the real-time exploration of up to a few million network entities (people, machines, IP addresses) with a standard workstation. Research Methodology ==================== Our research goals are synergistic with the needs of the NGI. The tools and methods we produce must be: * Understandable by anyone. No statistical, mathematical or computational background should be necessary. * Interactive. The ability to manipulate the data in real time, with changes and feedback occurring within 100 milliseconds is vital for the human cognitive system perceive smooth changes. * Scalable. The Internet will not get any smaller, so all methods must not degrade when realistic-sized data is encountered. * Useful on real problems. Our aim is to have impact on real-world problems. There are three major components to our research: * principles, guidelines and visual metaphors for effectively representing information; * software infrastructure encapsulating our fundamental results; and a * suite of novel applications. One of our key research challenges is inventing visual representations for information which often has no shape or form. Our objective is invent displays that help users navigate though and incredibly complex information spaces. Two particularly novel representations enable us to show tens to hundreds of thousands of lines of text on a single screen (SeeSoft) or a similar number of relationships (NicheWorks). The second tangible output of our research is Bell Labs VZ library. VZ is a novel (U.S. Patent 5,564,048) object-oriented library designed specifically to facilitate visualization. Its unique system of differentiating between visual representations and interactive manipulations forms a foundation for building insight-producing information visualizations. Some VZ features include: * Encapsulation of display mechanisms, color-coding and linking between views so that changing one view initiates instant updates in other views. * Views and techniques for focusing on data subsets and drilling down through large data sources to examine smaller subsets. * Generic mechanisms for handling user interaction which create a rich, but consistent set of possible manipulations. * Data storage and manipulation methods necessary for strongly inter-connected data sources. As well as these basic components, we have crafted a family of complete views, each of which focus on one aspect of the data and are built to work together in an interactive environment. A partial list of such views includes: * Statistical views: e.g. histograms, bar charts, scatter plots * table views * network-oriented views * text views * geographical views The third result of our research is a set of approximately two dozen applications that could form a foundation for NGI Visualization and support many core activities such as performance measurement or caching. Some well-known examples include: * NicheWorks: provides an interactive environment for exploring relationships in huge graphs with up to a few million nodes. We have used NicheWorks to visualize Web page associations, local calling communities, and for network fraud detection. * SeeNet: layers communication patterns over a map of the US. This tool allows geographical patterns and communications traffic to be explored jointly. Further, this tool allows time-based network information to be incorporated into the analysis. * 3D Globe: Uses advanced 3D visualization techniques, such as day-night lighting, transparency control and filtering/focusing mechanism to overcome the display clutter that plagues network node and link displays. Summary ======= Visualization is an enabling technology. Its purpose is to allow people to understand a complex problem, to investigate methods, patterns and anomalies, and to pose queries. The first step in solving a problem is to understand it. To understand something, you need to see it. Information Visualization provides this investigative facility. Since our research at Bell Labs has had large network data as a core focus for several years we think that our tools may help be a launching point for NGI. We have created not only fundamental building blocks and methodology, but several systems specifically aimed at network visualization. These systems and techniques are scalable, multi-platform and have been used in business to investigate fraud, to explore customer calling patterns domestically and internationally and, most recently, to analyze patterns in the world-wide web. It is our strong opinion that Information Visualization is and will be a crucial technology in the future. As networked information sources proliferate and communication speeds increase, tools for managing and analyzing the data flow will be critical for any project involving distributed participants, such as remote learning, monitoring and data sharing. Equally important will be the ability to locate and investigate misuse of facilities; hacking, data theft and national security fall in this area. We look forward to continuing our research into these important areas. References ========== BEW95 Richard A. Becker, Stephen G. Eick, and Allan R. Wilks. Visualizing network data. IEEE Transactions on Visualization and Computer Graphics, 1(1):16-28, March 1995. GE95 Nahum Gershon and Stephen G. Eick. Visualization's new tack: Making sense of information. IEEE Spectrum, pages 38-56, November 1995. EiWi93 Stephen G. Eick and Graham J. Wills. Navigating Large Networks with Hierarchies. In Visualization '93 Conference Proceedings, pages 204-210, 25-29 October 1993. San Jose, California. EiWi95 Stephen G. Eick and Graham J. Wills. High Interaction Graphics. European Journal of Operational Research:445-459, 1995. EiFy96 Stephen G. Eick and Daniel E. Fyock. Visualizing corporate data. AT&T Technical Journal, 75(1):74-86, January/February 1996. Wi96 Graham J. Wills. Selection: 524, 288 Ways to say 'This is Interesting'. In Information Visualization '96 Conference Proceedings, pages 54-61, 1996. San Francisco, California. -- Graham Wills Data Visualization, Bell Labs gwills@research.bell-labs.com Silk for Calde!