(NSF Grant CNS-1217048)
PI: Eytan Modiano
This project develops a theory for understanding the requirements for control overheads in wireless networks and mechanisms for reducing the amount of these overheads. Network control mechanisms, such as scheduling, routing, and flow control, ensure effective data transport in a communication network, but also require the exchange of network state information, such as channel conditions and queue-length information, which amounts to “control overhead”. The project investigates the tradeoffs between the rate of sending such control information, and the ability to effectively control the network in terms of performance metrics such as throughput, stability, delay and network utility. The project takes a two-pronged approach: First, a rate-distortion framework is being developed for understanding the impact of degraded network state information on network performance. Second, mechanisms are being developed for reducing the amount of control overhead and the impact of these mechanisms on network performance is being investigated.
The project develops a fundamental understanding of the requirements for protocol overhead, which will lead to more efficient network control policies, with reduced overheads. Such improvements will have a significant impact on network performance, especially in wireless networks, which experience large overheads due to protocol inefficiencies. The results will be widely disseminated through publication in conferences, journals and the web to help advance the wireless networking field. The broader impacts include training of graduate and undergraduate students and technology transfer to industry and government laboratories.
Our research goals include:
- Develop a unifying rate distortion framework for understanding the relationship between control overhead and network performance.
- Investigate the impact of the rate of control information on throughput, delay, and network utility.
- Develop mechanisms for effectively controlling a network using the least amount of overhead.
- Study the role of control overheads in multi-hop networks, with emphasis on distributed control.
Prof. Eytan Modiano, PI
Matt Johnston, graduate student
Abhishek Sigha, graduate student
Isaac Keslassy, Technion
Yuri Polyanskiy, MIT
Some of the results developed under this project will be incorporated into graduate classes on data networks.
- Krishna Jaggannathan and Eytan Modiano, “The Impact of Queue Length Information on Buffer Overflow in Parallel Queues,” IEEE transactions on Information Theory, 2013.
- Krishna Jagannathan, Eytan Modiano, Lizhong Zheng, “On the Role of Queue Length Information in Network Control,” IEEE Transactions on Information Theory, September 2011.
- Matt Johnston, Eytan Modiano, “'Optimal Channel Probing in Communication Systems: The Two-Channel Case,” IEEE Globecom, December 2013.
- Mathew Johnston, Isaac Keslassy, Eytan Modiano, “Channel Probing in Communication Systems: Myopic Policies Are Not Always Optimal," IEEE International Symposium on Information Theory, July 2013.
- Matt Johnston, Eytan Modiano, Yuri Polyanskiy, "Opportunistic Scheduling with Limited Channel State Information: A Rate Distortion Approach,” IEEE International Symposium on Information Theory, Honolulu, HI, July 2014.