Large Scale Communication Networks
IPAM Spring 2002
March 11 - June 14, 2002
Events: IPAM Math Dept CS Dept EE Dept
Organizing Committee: David Donoho (Stanford), John Doyle (CalTech), Deborah Estrin (UCLA), Kristina Lerman (USC/ISI), Walter Willinger (AT&T Labs-Research) The long-term program will involve a community of senior and junior researchers. In the course of the program, activities will involve:
The intent is for long-term participants to have an opportunity to learn about large-scale communications networks from the perspectives of many different fields--mathematics, control theory, computer science, electrical engineering--and to meet a diverse group of people and have an opportunity to form new collaborations. In addition to these activities, there will be opening tutorials, three workshops, and a culminating workshop at Lake Arrowhead. Full and partial support for long-term participants is available, and those interested are encouraged to fill out an online application. Support for individual workshops is also available, and may be applied for through the online application for each workshop. We are especially interested in applicants who are interested in becoming core participants and participating in the entire program (March 11 - June 14, 2002), but give due consideration to applications for shorter periods. Funding for participants is available at all academic levels, though recent PhD’s, graduate students, and researchers in the early stages of their career are especially encouraged to apply. Encouraging the careers of women and minority mathematicians and scientists is an important component of IPAM's mission and we welcome their applications. Special Events
Program:The program will take place from March 11 to June 14, 2002 at the Institute for Pure and Applied Mathematics (IPAM), a new NSF funded international mathematical sciences research institute located on the UCLA campus.
The Internet has become a gold mine for new, exciting and challenging mathematical problems, where scale, complexity, and dynamic play key roles. The goal of this program on the interface between the mathematical/physical sciences and computer/engineering sciences is to initiate, facilitate or foster interactions among researchers with diverse backgrounds who seek to unravel the ill-understood dynamics of large-scale complex internetworks such as the Internet.
The global Internet is an example of a large-scale, highly interacting communication network that has experienced a fascinating evolution, even before the Web came along. Unprecedented in their growth, unparalleled in their heterogeneity, and unpredictable or even chaotic in their dynamics, large-scale networks of devices (e.g., computers, sensors) that communicate with one another using a set of protocols have become a gold mine for new, exciting and challenging mathematical problems, where scale, complexity, robustness, adaptivity, and dynamic play key roles and can no longer be ignored. Solving these problems can be expected to have profound implications for the efficient design and effective engineering, control, and management of future communication networks such as the next-generation Internet or networks of massively distributed, dynamic, and physically-embedded devices (e.g., sensor networks). The study of these networks is bound to rely more and more on an interdisciplinary approach that looks to other areas in the natural and social sciences where experimenting with and analyzing large-scale, complex, and highly interacting dynamical systems have a long tradition, e.g., biology, experimental physics, mathematical sciences, and economics. In addition, the networking application stands out in terms of the ability and potential for measuring, collecting and analyzing a wide range of extended, high quality data sets of many different types of network measurements that are unprecedented in other areas in the natural and especially in the social sciences.
For example, recent mathematical discoveries concerning various scaling properties of the temporal dynamics of Internet traffic as it traverses single links (e.g., self-similarity), or of some of the topological features associated with the physical structure of the Internet (i.e., links and routers) have revealed unprecedented opportunities for providing phenomenological explanations for these empirically observed phenomena (also known as ``emergent properties''). These explanations fully exploit the underlying networking context, are mathematically rigorous, can be validated against measured data, and provide a basis for developing a relevant, consistent, and verifiable theory of large-scale networks. This ability to move beyond the traditionally pursued descriptive nature of systems modeling (i.e., statistical model fitting) makes the networking application distinctly different from other fields in science and engineering. While many of these areas have a rich history in modeling complex systems (e.g., hydrology, turbulence, atmospheric sciences, freeway traffic, finance, biology), physical-based explanations are generally given only (if at all) at an intuitive level, without serious attempts for validating them empirically, mainly because the necessary information is not available or is hard (or impossible) to come by. Realizing this difference and the ensuing opportunities for new scientific discoveries due to access to extraordinary kinds of network-related measurements, researchers from many different scientific disciplines with a general dissatisfaction with the prevailing descriptive approaches to mathematical modeling have started to view the global Internet as a vast experimental playground, with unprecedented opportunities for measuring and collecting data that provide detailed information about so many different facets of user and network behaviors.
The goal of this program on the interface of the mathematical/physical/ biological sciences and computer/engineering science is to initiate, facilitate and foster interactions among the researchers whose backgrounds are highly diverse, who pursue the common but ambitious goal of unraveling the ill-understood dynamics of large-scale complex internetworks, and who fully recognize and appreciate the need for a full-fledged interdisciplinary effort for achieving their goal. To this end, the program is structured around three complementary emerging research topics that provide new and untested opportunities for ``hands-on'' workshops or activities consisting of a unique mixture of relevant experimental and theoretical research: next-generation network measurement infrastructures, next-generation network simulators, and massively distributed self-organizing systems such as sensor networks.
The first research topic is based on the DARPA-funded National Internet Measurement Infrastructure or NIMI project that concerns the development of a network-wide measurement infrastructure and that is geared towards collecting various types of measurements simultaneously from a set of potentially thousands of nodes within the network. NIMI nodes are equipped with ``probe platforms'' and engage in both independent and orchestrated measurements of network paths in an attempt to characterize the network's global behavior and to locate trouble spots. A small/medium-scale deployment of NIMI is imminent and the research topic in question deals with problems related to using NIMI-like infrastructures for studying and investigating networks at increasingly larger scales. An operational and tested version of NIMI can be expected to be available to the research community in 2001. The second research topic relates to another set of government-funded research projects and concerns the use, validation, and application of prototype next-generation network simulators that can successfully handle networks consisting of 100,000 realistically detailed multi-protocol hosts and routers. Various versions of these new simulation engines have already been made available to the research community and are currently undergoing a series of tests and validation procedures; for example, a reasonably reliable version of a scalable simulation platform for network simulations called SSFNET (Renesys Corp.) will become available to the research community during 2001, thereby extending the capabilities of the currently used NS-2 network simulator from Berkeley to handle truly large-scale networks. Finally, the third, more forward-looking research topic overlaps with a recent effort spear-headed by DARPA and concerns the design, control, robustness, and performance of sensor networks -- networks of massively distributed, physically-embedded, and highly heterogeneous devices that are typically small, low-power, mobile, and wireless. The three research topics have in common that they force us to face the paradoxical situation where experiments or simulations produce rich and detailed information, for which, at this point, practically no adequate analysis tools exist. Thus, the common glue between the three research areas is the need for innovative new approaches to data representation, analysis, and visualization, and one of the unifying mathematical research themes of this IPAM program will be the development and application of techniques from multiresolution analysis (MRA) and multiscale computations (MSC) to representing and analyzing the types of network-related measurements generated by large-scale networks of communicating sensors or computers.
To address this urgency for scientific advances in the different areas and to increase the likelihood that the proposed program will make some fundamental and high-impact contributions to networking research and to some of the other fields involved, we propose three ``experimental'' workshops that will bring together the leading scientists in these emerging research areas with a number of networking experts, researchers from mathematics, physics, and biology, and a group of experts on MRA/MSC-based approaches for representing and analyzing the new and challenging high-volume, high-quality, and high-complexity observational data resulting from these new application areas. The major objectives of these proposed workshop activities and of the resulting interactions is to design, run, analyze, and evaluate a set of experiments that - make use of the unique capabilities of these new technologies, - address fundamental and difficult networking problems, - have a high potential for advancing our understanding of large-scale complex internetworks, - are likely to generate new and interesting problems related to data processing, data analysis, visualization, or mathematical modeling, and - require expertise in the areas represented by the participants. Progress under this proposed new workshop format will be monitored in 4 stages: experiment-design (end-of-workshop), experiment (post-workshop), initial findings (concluding conference), further progress (post-program). In the long-run, the success of the proposed workshop format will be measured by how long after this program and at what level the different groups formed during the workshops will continue to work together and/or form new collaborations. To this end, IPAM has already budgeted for "reunion workshops" that will be held once during each of the two years immediately following the main program. |
|
|
