This project aims to leverage GENI's deeply programmable core and wireless edge networks to implement and evaluate recently proposed approaches for delivering predictable end-to-end performance to mobile devices. These approaches involve techniques such as inter-flow and intra-flow network coding, multipath routing, available bandwidth estimation, and spare bandwidth exploitation. While these techniques have been studied using simulations, it has never been possible to validate their models and their limits on an operational network without GENI's fine-grained measurement and control capabilities.
True ubiquitous broadband networking cannot be realized without overcoming the bandwidth bottleneck and mobility overheads of future Internet, both at the core and at the wireless edge. Early wireless access deployments in many instances resulted in only marginal improvements in end-to-end data rates to mobile users. Outcomes from this research may influence the ongoing broadband initiatives in the United States.
There were four major outcomes from this project. 1. Experiments conduted during the course of the project demonstrated that the commonly used transport layer scheme called Tranmission Control Protocol (TCP) does not perform well in emerging ultra-fast wireless access networks. The experiments also identified the primary cause of this underperformance. The large bandwidth-delay product in these networks adversely affect the feedback mechanism used in TCP to deal with fluctuations in available bandwidth. Even variants of TCP specifically designed for large bandwidth-delay products did not perform well because of the inherent timescales of the bandwidth fluctuations in these emerging wireless networks. 2. A new network layer protocol was developed to alleviate the underperformance problem of TCP. The new network protocol integrated several new ideas including network coding, spare bandwidth estimation, rate adaptation, and multiple route forwarding. Experiments demonstrated that, if these new network layer protocols are deployed, the performance of TCP does not degrade adversely and multi-gigabit end-to-end throughputs can be achieved. 3. By expoliting the deep programmability of Global Environment for Network Innovation (GENI) infrastructure, at-scale experiments spanning the large geographical extent were conducted to evaluate the effectiveness of the new network protocols. These experiments also demonstrated that these new network protocols can be easily integrated into existing network layer schemes and thereby support end-to-end multi-gigabit throughputs. 4. Multi-gigabit wireless networks cannot exist without access to adequate wireless spectrum. There is, however, an acute shortage of available wireless spectrum since, over time, large chunks of the spectrum have been allocated to various purposes. At the same time, studies have shown that at any given time instant, the utilization of the wireless spectrum is not that high. As a result, spectrum shortage can be alleviated if we create a marketplace for trading licensed spectrum among users. Such a marketplace requires pricing and purchase strategies. Novel wireless spectrum pricing and purchase strategies were developed as part of the project. Collectively, these two strategies bring forth increased predictability in network access which in turn, offers potential for better planning of network resource allocation in the backbone. 5. The additional predictability in network access as a result of the new pricing and purchase strategy was also exploited to developed new handoff mechanisms to support mobility in wireless network. As users move, the wireless nodes must to different access points. The latencies incurred when an access point has to be changed, adversely affect the performance of streaming and interactive applications (such as video streaming and multi-party conferencing). Using the new purchase strategy, mobile users can reserve resources into the future, not only in time but also in space. A new handoff strategy that takes advantage of this aspect and hides the effect of handoff latency was developed as part of this project.
