The Internet has an ever-expanding role in our daily lives; yet, it is arguably one of the most fragile components of our nation's critical infrastructure. The Internet was designed as a research network without the expectation that it would eventually be used for everything from banking, commerce, and telecommunications to the remote management of power networks. The scale and heterogeneity of the Internet have far surpassed all expectations, and the Internet is responding by showing signs of strain. Moreover, new applications heighten the need for security and network management capabilities, neither of which were major goals in the original design of Internet protocols.
DIMACS is hosting a 7-year special focus devoted to the study of algorithms and protocols for large-scale networks. The focus is scheduled to start in August 2007 and continue through July 2014. The special focus aims to enhance our understanding of the limitations of today's protocols, as well as the gains that new designs could achieve. This is an emerging cross-disciplinary area that requires expertise from several fields including networking, theory of computing, computer and communications security, and game theory. Research collaborations spanning these communities are crucial to making progress on the most challenging problems, and enabling these collaborations is a major goal of this special focus.
As the Internet continues to grow, more and more business-critical functions rely on its availability. One can easily envision a future in which the vast majority of communications traffic, including telephone, television, radio, business data, and government data, will rely on an Internet infrastructure that is available and secure. For the Internet to meet these challenges, we need a much deeper understanding of the properties of our existing protocols and the fundamental tradeoffs that should guide the design of the future Internet. Providing a strong algorithmic foundation for the Internet is especially timely, as the research community embarks on an ambitious rethinking of the Internet architecture.
There are many algorithms and protocols used in the Internet and its applications. Some adequately serve their desired purposes, while others need improvement. However, there is a disconnect between the methodology and results of algorithms research and the methodology and results used to guide the adoption of Internet protocol standards. On the one hand, traditional distributed-algorithms research does not adequately model the Internet's design goals, including autonomy, scalability, and privacy. On the other hand, protocol-adoption standards far too often rely on experimentation and testing by vendors and select customers, not on formal analysis. Protocols are often tweaked to add customer functionality without scrutinizing the resulting behavior in worst-case situations or proving any kind of correctness or security properties. Furthermore, these worst-case situations occur more often than expected, due to both the sheer size of the network and the fact that malicious agents can use security flaws to take control of significant parts of the Internet.
This special focus seeks to bridge the gap between networking research focused on the existing artifacts - the protocols and mechanisms underlying today's Internet - and the new work that needs to be done to lay a solid foundation for the design of a future Internet. Research focusing on today's network emphasizes characterization, primarily through measurement and prototyping, of existing protocols and mechanisms, in order to improve our understanding of the Internet and guide incremental changes to the system. Although algorithmic models have played a role in this work, the details of today's protocols and mechanisms often defy attempts to impose rigorous models "after the fact." The future Internet needs to be more secure, be easier to manage, and take greater advantage of new underlying technologies, such as sensor networks, wireless networks, and optical switching. This argues for the design of new protocols and mechanisms with their key properties in mind from the outset. An algorithmic mindset is an extremely important ingredient in this line of research.
This special focus is guided by a deep understanding of the current Internet but allows for the possibility of radical change where it is warranted. The focus seeks to analyze and design protocols, algorithms, and architectures for a future Internet that is based on sound mathematical and computational foundations, ensuring scalability, security, and manageability.
This special focus is particularly timely given NSF's and the computing research community's interest in developing the protocols and associated experimental infrastructure for the Internet's next generation. A special focus on algorithmic foundations of the Internet is especially timely in light of NSF's FIND (Future INternet Design) and GENI (Global Environment for Network Innovations) initiatives. The FIND program encourages the research community to take a clean slate approach to the design of a future Internet, without the constraints of backwards compatibility with the current protocols. FIND provides a unique opportunity to create a new network architecture built on a solid conceptual foundation. In addition, deeper understanding of the limitations of the existing protocols and of the fundamental trade-offs between different design goals can focus researchers' attention on the most promising directions for a new Internet architecture.
This special focus aims to facilitate research in this area through the broad participation from both the networking and theoretical computer science communities. In addition to encouraging collaborations between the two communities, focus activities can educate the participants and the larger community about the current research challenges and known results in the two areas. The special focus workshops and working groups target the "sweet spots" where the crucial challenges for designing the future Internet (e.g., security, network management, layer-2 technologies, and economic incentives) have rich connections to algorithms research (e.g., cryptography, formal analysis, game theory and mechanism design, and streaming algorithms).
The research ideas enabled by this DIMACS activity can lead to new protocols and architectures that can be evaluated on GENI. Focus workshops are encouraged to include presentations on the GENI facility in order to encourage experimentation on GENI and explore how to put the key design principles of GENI such as virtualization on a strong theoretical footing.