Skip to main content

Scheduling, routing, and transport challenges in multi-hop wireless networks

Final Report Summary - EFFICIENTMULTIHOP (Scheduling, routing, and transport challenges in multi-hop wireless networks)

Project objectives

Multi-hop wireless networks (MWNs) are networks where end-to-end paths consist of a number of consecutive, wireless hops. This new architecture enables a number of new applications, e.g. vehicular applications, environmental monitoring and disaster recovery communication, and improves the performance of existing services, e.g. Internet connectivity in airports, convention centres, and hospitals. MWNs differ significantly from wired and single-hop wireless networks (SWNs), and require fundamentally different approaches to operate efficiently.

The main goal of this project is to address fundamental architectural and design challenges of MWNs in all major networking layers, using formal mathematical tools, simulation, and experimentation. Specifically the first objective is to access the performance of scheduling protocols while taking into account both performance and implementation overhead. Special attention is paid to random access schedulers, as they have become the de facto standard, and their performance is studied via formal analysis and simulations. The second objective is to design, analyse, and implement neighbourhood-centric transport schemes for congestion control and rate allocation in the context of static MWNs. Both additive-increase multiplicative-decrease schemes, and explicit rate notification schemes are investigated, and fairness and efficiency issues are thoroughly studied. The third objective it to design, optimise and implement mobility-assisted routing schemes in the context of mobile MWNs. Since the performance of any such scheme depends on the constantly changing level of network connectivity, we also investigate automated distributed mechanisms that allow nodes to characterise on the fly how connected the network is.

Work performed and main results

We have worked on the performance analysis of scheduling protocols, and in particular on accessing the performance gap between the de facto standard 802.11 scheduler and the optimal scheduler under a variety of realistic MWN scenarios. Using formal theory and simulations we have established that the performance of 802.11 is very close to that of the optimal scheduler in practical scenarios of interest, and it achieves similar performance to greedy maximal scheduling, which is the best known distributed approximation of the optimal scheduler.

We have also worked on designing, implementing, and accessing the performance of transport schemes for congestion control and rate allocation. Specifically, we have tested a variety of transport schemes and have designed and implemented a novel transport scheme with very desirable features. Our novel scheme requires no modifications to transmission control protocol / Internet protocol (TCP / IP) and the 802.11 media access control (MAC), and is responsive to short flows, MAC-layer auto rate adaptation, and other dynamics, as we demonstrate in extensive experiments on real testbeds. Despite its simplicity, the technique is on average within 10 % of the max-min optimal allocation on several canonical topologies.

Last, we have worked on mobility-assisted routing schemes, and, in particular, we have designed efficient unicast and multicast routing protocols in the context of intermittently connected (or, delay tolerant) MWNs. We have also designed a methodology that predicts the current level of network connectivity and allows nodes to select the right routing protocol to be used each time. This adaptive approach significantly improves the performance and robustness of such networks.

Potential impact

Our work on scheduling offers formal results establishing the superiority of the 802.11 family of protocols in the context of MWNs, and gives insightful guidelines about how to design appropriate transport protocols that fully utilise the capacity offered by 802.11. Our work on transport layer issues resulted in a very efficient, backward compatible protocol, which allows MWNs to achieve their full performance potential. This, in turn, will facilitate and help materialise the plethora of applications that people have envisioned running on top of such networks. Last, our work on the routing layer for intermittently connected MWNs, offers a method to predict the connectivity level of the network and select the most efficient routing scheme among a family of such novel schemes. This significantly improves the performance and robustness of such networks, and will boost the performance of applications running on top of such networks, e.g. vehicular applications, environmental monitoring, and disaster recovery communications.

In summary, a number of useful applications (vehicular, military, environmental, disaster recovery, connectivity anywhere / anytime, etc.) have been envisioned to run on top of MWNs. The impact of this project is to significantly improve the performance of MWNs such that they can offer the required performance to all these applications.