Current Research Interests
Application-layer Internetworking: The explosive increase of network-enabled sensors, actuators, and other cyber-physical systems, referred to as the Internet of Things (IoT), together with the proliferation of proprietary network infrastructures that largely bypass the public Internet, requires new solutions for interconnecting large numbers of heterogeneous networks. We seek to develop an internetworking architecture that is based on selforganizing application-layer networks. Self-organizing means that devices use all their communication modalities to detect and establish connectivity with other devices to form a network. All nodes of an application-layer network participate in relaying data between sources and destinations of information. Application-layer internetworking refers to the interconnection of devices with attachment points in different network infrastructures at the level of application programs. The long-term goal is to explore the foundations of application-layer internetworking and develop new protocol solutions that enable and facilitate such internetworking. We address challenges of integrating low-power IoT devices into the novel internetworking architecture, and addresses scalable routing, resilience to failures, and network policies.
A major vehicle for developing, evaluating, and deploying solutions is an overlay network system, called HyperCast, that we have developed in my research group. HyperCast is an open source software system (of about 100,000 lines of code) for self-organizing application-layer overlay networks. Initially conceived for the empirical evaluation of large-scale reliable multicasting, the software has evolved into a programming platform for application-layer internetworking networks that accommodates different types of substrate networks and permits a variety of message semantics.
Designing traffic control algorithms with Network Calculus: Recently, there has been a new surge of interest in shaping and scheduling algorithms, for several reasons. The specification of an Ultra-Reliable Low-Latency Communication (URLLC) service category for 5G networks, with guaranteed latencies below 1 ms, has raised the ante on traffic control algorithms. New standardization efforts in the IEEE for Layer-2 networks and in the IETF for Layer-3 networks define a compatible framework for traffic control algorithms with assured latency bounds in local and wide area networks. Second, large content providers that manage traffic within and between servers, clusters, and data centers rely on fine-grain traffic control at multiple levels of aggregation, which creates a need for new hierarchical scheduling and shaping mechanisms. An additional factor for the current interest in scheduling are new architectures for line rate switches with programmable packet schedulers that permit a customization of scheduling algorithms to application requirements. The necessity to adapt traffic control to the demands of applications motivates the search for novel algorithms and methods for scheduling. We seek to develop new traffic control algorithms that are based on the sound mathematical principles of the network calculus and that allow an efficient implementation in deployed systems. Doing so we seek to develop theoretical underpinnings for scheduling and shaping algorithms in modern communication networks. By also implementations of the developed scheduling and shaping methods, we want to close the gap between theory and application of traffic control algorithms.
Communication-based Train Control (CBTC): Communications-Based Train Control (CBTC) is aimed at replacing conventional rail signaling with train control enabled by wireless communication between the train and a network of access points. CBTC systems are safety critical, meaning that their projected benefits have to be accompanied by high standards of reliability, beyond those of typical communication networks. We have shown that self-organizing networks can facilitate the switchover from a CBTC infrastructure network to a backup network in case of a communication failure. A second contribution is a novel handover algorithm for CBTC systems that takes into consideration the fixed mobility pattern of train systems (Prior handover algorithms for CBTC systems where based on received signal strength and did not consider train location when making handover decisions). As a third contribution, done in collaboration with Prof. Costats Sarris's research group, we propose a fundamental overhaul of CBTC network design. Conventional CBTC system design relies on a rather fragmented approach, where first a radio-frequency survey of the propagation channel is conducted to determine the location of access points. With the network infrastructure in place, communication protocols, including handover algorithms, are subsequently established in a second phase to meet CBTC service requirements. We propose a method that integrates physics-based and network protocol design, whereby the deployment of access points and the handover algorithms are designed in a single phase.