- Learning-oriented computer networks: The ``data pipe'' model used by the existing Internet protocol stack is no longer adequate for many emerging applications. Information, which can have different data representations, and knowledge, e.g., learned model parameters, have communication patterns and requirements different from those of point-to-point, reliable and elastic data transfer. Information and knowledge can be aggregated and transformed inside the network, due to latency constraints, and control scope is no longer limited to endpoint-to-endpoint only. On the other hand, new communication infrastructures such as 6G mobile communication systems, low-orbit satellites and underwater/ground communication systems are emerging rapidly, which offer additional communication capabilities with certain topological features. Now the network protocol stack between application requirements and communication services becomes a new bottleneck.
This research program focuses on a new computer network architecture serving learning-centric applications in data centers, around network edges and on mobile devices, by leveraging, improving and creating new learning techniques to determine and optimize protocol mechanisms and control policies. It will further our research in the last two decades on topology control, protocol design and performance evaluation for computer networks, to network a truly dependable cyber-infrastructure. The learning-oriented computer network can learn from applications and communications automatically and continuously while running on different infrastructures to support diverse requirements, and keep evolving its protocol mechanisms and control policies in an online manner, while maintaining security and preserving privacy, to learn and perform more effectively and efficiently.
- Topology control for computer networks: The world is connected and becomes increasingly so with the advances of information and communication technologies. Many natural phenomena and man-made systems can be captured, explained and improved by a better understanding of the structure and dynamics of the network topology and interactions. The proposed research program focuses on the topology control for computer networks, based on our early work on computational geometry and geometrical probability, and furthering into stochastic geometry with geographical perspectives, for both wireless communication networks and cloud computing networks. For example, in wireless communication networks, the received signal and interference strengths depend on the distances among transceivers and interferer, as well as shadowing and fading due to obstacles and multiple propagation paths. With cloud computing and distributed storage, user requests and associated data can be redirected to and stored at different locations, which also affect the service quality, system efficiency and the resilience against natural disasters and man-made attacks. Thus locations and distances play an increasingly important role in computer networks, particularly for topology control, based on which affect all upper-layer protocols and functions.
Starting from a graph-theoretic perspective, our research in computational geometry and geometrical probability led to a few breakthroughs including the random distance distributions associated with rhombuses, hexagons, regular and irregular triangles, and eventually arbitrary polygons with or without a reference point, with a wide variety of applications. In the proposed research program, we will continue with high-order, joint and conditional distance distributions in both physical and logical spaces, e.g., hop distance distributions with different mobility models and routing schemes. By taking geographical perspectives into account, our proposed research will also incorporate the spatiotemporal dynamics of user requests, computation and storage capacities, and energy supplies into the topology control process. The proposed research will also continue the training of highly-qualified personnel with high-impact research contributions and high-competition job placement to benefit the economy and society in Canada and around the world.
- Protocol design and performance evaluation:
The Internet has achieved great success in the last half century. The networking research community has
realized that we are now at the doorstep of designing and evaluating next-generation network protocols for the
next half century. There are many efforts already initiated, such as FIND/GENI/NetSE (US), FIRE (EU), JGN
(Japan), etc, driven by the emerging of cloud computing, smart grids, cyber-physical systems, online social
networks, etc, collectively under the name of network science and engineering. Besides the traditional demand
on networking research, such as performance, reliability, security and testability, these new networking
paradigms have created many new challenges such as energy-efficiency, ultra-scalability and heterogeneity, and
so on, among which the network topology plays an increasingly important role and affects all upper-layer
protocols. The proposed research program, together with other academic and industry research programs in
Canada, will keep Canadian society and economy on the leading edge of this international competition.
Particularly, traditional network protocols follow the "store-and-forward" principle. With the advance of
physical-layer technologies such as cognitive radio, compressive sensing, cooperative communication and
white-space networking, as well as the upper-layer technologies such as in-network processing and overlay
networking, the design principle is expanded to "store-carry-process-forward" in our research program, to fully
utilize predictable, controlled, or user mobility to carry information and to process data flows inside the
network to achieve the maximum capacity. This is the first time that such principle is proposed in the research
community, although the literature has been examining carry and process separately in last few years. Due to
the time/location-varying topology in many emerging, complex systems such as vehicular ad hoc networks,
wireless sensor networks, mobile social networks and peer-to-peer networks, we believe such a thorough
examination of store, carry, process and forward is very necessary to design and evaluate new protocols for the
critical infrastructures that we have to rely on.
- Networking a dependable cyberinfrastructure:
The daily life of ordinary people, ranging from banking to shopping and from learning to entertaining, is
increasingly dependent on the Internet-centric cyberinfrastructure. In the last decade, a vast amount of research
and investment has been devoted to improving the performance and reliability of the Internet, primarily in its
backbone. However, there are still many obstacles left in providing such reliability in access networks and
among Internet Service Providers. Denial-of-Service attacks, email spams and web scams have caught many
Internet users, both consumers and businesses, off-guard. The control-plane vulnerabilities in Domain Name
System (DNS), Border Gateway Protocol (BGP) and Session Initiation Protocol are even more alarming.
Existing schemes such as DNS Security and Secure BGP have high complexity and are rarely deployed. For
such a complex system as the Internet, it is very surprising to find that there are no systematic and online
testing capabilities embedded. In summary, the Internet is still much less dependable than what it should be and
most people have taken for granted.
Our research program on a dependable cyberinfrastructure focuses on the fundamental networking issues in
order to considerably improve the end-to-end reliability, security and testability of the Internet and its
accessories. The research explores three highly correlated topics: (1) We want to leverage the reliability at the
core Internet with that of wireless personal, local, metropolitan and even wide area networks in ad hoc or
hybrid mode. (2) We want to enhance the control plane security for the Internet, particularly in its naming,
routing and signaling infrastructures. (3) We want to embed the systematic and online testability into the
cyberinfrastructure, rather than being another add-on. The research is also expected to produce highly qualified
personnel for academia and industry with growing demands for such talent to meet this grand challenge during
the next major transformation of the Internet.
