Give


Graduate Exam Abstract


Pan Ho Lee

Ph.D. Preliminary

November 3, 2006, 9:00-11:00am

Engineering B3

Application-Aware Overlay Networking for Distributed Adaptive Sensing System


Abstract: Distributed Collaborative Adaptive Sensing (DCAS) systems are emerging for applications such as detection and prediction of hazardous weather using a network of radars. In the data distribution scenarios for the DCAS systems, a variety of data is distributed to geographically distant end-users, sensors, various processing nodes and storage units under dynamic network conditions including network congestion, link degradation/outage, and variable cross-traffic along wired and wireless links. Moreover, in many of these systems, the critical sensor data has to be distributed to multiple end users, and the end-users have differing QoS requirements for the data based on the ultimate use of the data. The DCAS systems with these properties need to employ a new data-dissemination mechanism that is able to adapt to the various network conditions in an application-specific manner to achieve best feasible QoS for the end users, and to provide acceptable communication quality and robustness for the DCAS data distribution. In this research we present an overlay-based data transport service framework for the DCAS systems. The distinguishing characteristic of our framework is application-awareness, where applications can decide how to address the various network conditions, and express their application-specific communication needs such as bandwidth requirement and latency constraint to steer overlay routing protocol in selecting paths. Our framework consists of a general-purpose overlay architecture and a programming interface. The architecture enables deploying application-aware in-network services in an overlay network to allow applications to best adapt to the network conditions. The application programming interface (API) facilitates development of applications within the architectural framework, and supports the configuration of overlay nodes for the in-network application-aware processing. The API also enables communication between the applications and the overlay routing protocol for the desired QoS support. We demonstrate the efficacy of the proposed framework by considering a DCAS application. In the application, high-bandwidth radar data is distributed to multiple end users with heterogeneous QoS requirements and network connectivity. A packet-marking scheme based on the application content is implemented that enables application-aware processing of the data in the overlay nodes. A token-bucket based rate control algorithm in conjunction with the proposed packet-marking scheme enables on-the-fly selection of data for forwarding to a particular end user at the desired transmission rate as well as dropping packets to avoid network congestion. We evaluate the schemes in a network emulation environment and in a world-wide Internet testbed. Under various network conditions, the proposed schemes are very effective in delivering high quality data to the multiple end users. In addition, we are planning to develop, model and evaluate a many-to-one application-aware transport service using the same API and architecture. The delivery of DCAS data often requires high bandwidth, low delay and loss rate in network transmission. Therefore, with the single best path overlay routing, the end users may not attain the required bandwidth, latency constraint and end-to-end communication availability for the DCAS applications. We consider the use of multi-path overlay routing to overcome the limitation of the single best path overlay routing. It is required to examine the performance gain and overhead of the multi-path overlay routing on various network settings before the routing algorithm is used in practice. On going research includes the evaluation of the performance of the multi-path overlay routing algorithm.

Adviser: Dr. Jayasumana
Co-Adviser:
Non-ECE Member: Dr. Malaiya, CS
Member 3: Dr. Chandrasekar, ECE
Addional Members:

Publications:


Program of Study: