Dissertation defense announcement
In my dissertation research, I studied cognitive packet networks (CPN) which build networked learning systems that support adaptive quality of service-driven routing of packets. Routing in CPN differs from other algorithms in that no global information exchange among the nodes is required and that routing decisions can be easily tailored to pursue specific quality-of-service goals. In CPN, smart packets learn how to make intelligent decisions from the experience of previous packets. These packets employ spiked random neural networks and reinforcement learning as the core elements of an intelligent system that acquires, stores, and exploits knowledge with the aim of achieving pre-selected performance criteria.
My investigation initially was in the design of CPN and study of their performance over wired links, and later included tetherless, mobile ad hoc networks. My proposed ad hoc cognitive packet networks (AHCPN) algorithm is able to cope with node mobility and high error-prone networks. I introduced path availability, a new routing metric that quantifies a path via the probability to find nodes and links available for routing. When path availability is formulated as a function of battery lifetime at nodes and signal-to-noise ratio of communication channels, the result is a unique and synergetic routing solution that offers reliability, energy awareness, and link-quality awareness. Nodes which have low power are able to extend their working lifetime and routes are selected based on power considerations, leading to the establishment of more robust communication paths over time.
To conduct evaluation studies, I deployed various network testbeds and performed extensive experimentation. I proposed and implemented two operating architectures, both of which were integrated into the Linux 2.4 kernel. The two architectures differ in the way they coexist and interoperate with other network protocols such as TCP/IP. In the first implementation, an independent network stack resides in the kernel and interfaces with the BSD socket layer to supply communication support to applications. In the second approach, a virtual interface embeds the cognitive packet network algorithm, offering automatic interoperation with any layer 3 protocol, such as IP.
In addition, to gain a better understanding of the performance and dynamics of ad hoc CPN, I developed a simulation model of the algorithm for Network Simulator 2 (NS-2). Because of the resources involved (i.e. large number of resources and node mobility on a large terrain), the simulator allows us to gain valuable insight about the performance of the routing algorithm under situations difficult to replicate in a real testbed. Furthermore, it provides a simple framework for performance comparison studies with other ad hoc routing protocols available in NS-2.
My measurements on real systems and simulation studies revealed the great capacity for adaptability and successful performance of both CPN and AHCPN under different network conditions.
Knowing the growing necessity for networked multimedia applications, which rely on high performance communication technologies, I am motivated to continue my research within the area of computer networks. In the near future, I plan to work within the following areas:
Wireless networking. Wireless, ubiquitous, and pervasive networking is perhaps the next big step in the development of the Internet which will bring connectivity to a massive number of heterogeneous, small devices-for example monitoring systems. Many research issues arise from the expected internetwork. One of these issues is the optimization of TCP to operate over such unreliable and slow media.
Optical networks. Optical networks offer the high bandwidth that would be required in next-generation networks. Because this is an emerging technology, there are many challenges awaiting attention, such as, the design of scalable O-E-O switch fabrics.
Multicasting. As an alternative to unicast, multicast promises to diminish the increasing need of bandwidth for multimedia traffic on the Internet. One direct research extension of my dissertation is in the use of smart packets to handle multicast routing while still supporting the establishment of dynamic QoS-defined routes within a learning system.
In general, my research philosophy is a permanent feedback circle of examination of existing ideas from diverse fields, conceptual development and analysis, simulation, and experimentation. In the next years, I expect to continue engaging in significant research in computer networks while also exploring new areas of study.