Adaptive Quality of Service for Wireless Ad hoc Networks

Adaptive Quality of Service for Wireless Ad hoc Networks Seoung-Bum Lee This thesis contributes toward the design of a new adaptive quality of service (QOS) paradigm for wireless ad hoc networks. We address some of the key performance problems in the broader realm of wireless ad hoc networks, including mobile ad hoc networks and emerging wireless ad hoc sensor networks. Wireless ad hoc networks represent autonomous distributed systems that are infrastructureless, fully distributed, and multi-hop in nature. Over the last several years, wireless ad hoc networks have attracted considerable research attention in the general networking and performance community. This has been fueled by recent technological advances in the development of multifunctional and low-cost wireless communication devices. Wireless ad hoc networks have diverse applications spanning several domains, including military, commercial, medical, and home networks. The results of all this research activity the wireless ad hoc networks are starting to move from the research domain into the real world and are being gradually integrated into our daily lives. Projections indicate that this will accelerate later in the decade, to the point where some analysts predict that these types of self-organizing wireless devices will eventually become the dominant form of communications infrastructure. To cope with the unpredictable nature of this highly dynamic environment, wireless ad hoc networks need to be able to adapt to changes in resource availability (i.e., energy, bandwidth, processing power, network density, and topology changes) and overcome any unanticipated networking problems while satisfying a wide range of application requirements. Meeting these requirements in such an environment is very challenging because the performance observed by users, devices, and routing paths selected through the network will continuously change in response to the timevarying network dynamics. This thesis addresses some of the key issues needed to meet the requirements in support of the adaptive QOS for wireless ad hoc networks. They include (1) an adaptive QOS framework and signaling protocol for mobile ad hoc networks, (2) congestion mitigation in mobile ad hoc networks, and (3) a cost-efficient agile routing mechanism for wireless ad hoc sensor networks. In the contribution of this thesis, we study the technical challenges for QOS support in mobile ad hoc networks and propose the INSIGNIA QOS framework that is designed to support the adaptive service paradigm. The key component of the QOS framework is the INSIGNIA signaling system, an in-band signaling system specifically designed to address the adaptive QOS related challenges in mobile ad hoc networks. The INSIGNIA signaling system is recognized as one of the first QOS signaling protocols in mobile ad hoc networks. We also present a detailed performance evaluation of the IEEE 802.11 based INSIGNIA signaling system with a number of well-known MANET routing protocols. The INSIGNIA system shows operational transparency to a number of MANET routing protocols and offers significant performance gains for various TCP and UDP flows. Next, we investigate the MANET-specific congestion conditions called hotspots. A hotspot is defined as a node (or a group of nodes) experiencing flash congestion conditions or a period of excessive contention conditions in wireless ad hoc networks. Hotspots can exist even in lightly loaded ad hoc networks and can severely degrade the network performance. The existence of a hotspot is largely due to mobility in mobile ad hoc networks and related traffic patterns where the node mobility continuously changes the network topology and causes the on-going traffic to reroute. This effect varies the network loading conditions and produces transient congestion. These hotspots cause packet loss, increase in end-to-end delay, and even trigger route maintenance as they are often misinterpreted as routing failures. As a solution to this problem, we propose a Hotspot Mitigation Protocol (HMP) that works with best effort routing protocols. The HMP suppresses and disperses new/rerouted flows from hotspot regions to mitigate congestion conditions. HMP also provides a traffic throttling scheme that rate controls best effort TCP flows to relieve congestion. In the final contribution of the thesis, we shift our research focus to wireless ad hoc sensor networks, a new emerging frontier in wireless ad hoc networks. Based on the observation that current routing algorithms for sensor networks yield poor information delivery (i.e., poor fidelity as measured at the Internet gateway to the sensor network – typically called a sink), we investigate the problem and the solution space using the TinyOS embed operating system in an experimental testbed of Mica2 mote sensors. We show that the poor fidelity is largely due to the unresponsive nature of the route selection convention commonly in use in sensor networks. To resolve this problem, we propose an agile, cost effective, and high-fidelity yielding hop-by-hop routing protocol called Solicitation-based Forwarding (SOFA). SOFA achieves fast path convergence at network deployment time and acquires an alternative path quickly with minimal signaling overhead when faced with path changing conditions due to network dynamics. Path maintenance in SOFA is minimal and when a new sensor is added to the network, it is integrated quickly and seamlessly. SOFA shows significant reduction in energy consumption where the energy savings in SOFA network are primarily due to decrease in the signaling overhead. The on-demand nature makes SOFA cost effective; its agile self-adapting nature makes it resilient to network vagaries; and its use of timely solicitation-based handshakes make its forwarding decisions effective in data delivery.