basis of the mesh networking fundamentals.
Most of the existing solutions assume either general contention based channel access, or time division multiple access (TDMA) channel reservation strategies. IEEE 802.11s mesh networking framework, being a new standard in the wireless domain, is not characterized by any of the above mentioned works. The earlier works are difficult to implement over the IEEE 802.11s framework, as it provides a complete new set of protocols for the topology management, channel access and mesh path selection. Compatibility becomes a serious issue when dealing with multivariate optimization based solutions for the performance optimization in an IEEE 802.11s mesh network.
Therefore IEEE 802.11s, the new standard for mesh networking over the well-established Wireless Local Area Network (WLAN) technologies, demands focused researches to opti- mize the performance of a WMN and to provide extra functionality over the basic proto- cols. The standard has the capability of providing last mile ubiquitous broadband access through a cost-effective wireless solution, however need to be enhanced with modern de- mands of QoS assurance and new features like multi-rate support, for effective usage of the available channel resources.
Flow scheduling in the directional multi-interface IEEE 802.11s WMN suffers from the problem of unbalanced traffic allocation, because of the increased interference among mesh routers. Let us consider an end-to-end flow between two mesh routers2 ST AS and ST AD, that goes through intermediate routers ST AI1, ST AI2, ... and so on. In a directional network, the end-to-end flow ST AS → ST AD can be subdivided into several sub-flows based on the intermediate path, such asST AS →ST AI1,ST AI1 →ST AI2 and so on. The maximum end-to-end flow capacity for such a flow is the max-min capacity for all its sub-flows. However, in the IEEE 802.11s mesh network, intermediate mesh routers remain unaware of the maximum throughput that a flow can achieve, and inject packets more than the maximum capacity, which may cause throughput degradation for other flows. The upper layer end-to-end flow control fails to solve this problem, because an intermediate router may receive data from more than one flows, and may forward data to more than one next-hop routers. In this thesis, the problem of the balanced flow allocation is first modeled as a centralized convex optimization problem, considering the maximum end-to-end flow capacity and the network interference. Based on the convex nature of the solution set, the problem is decomposed into local sub-problems to calculate the solution at individual mesh routers, for their individual interfaces. The IEEE 802.11s MCCA based channel access mechanism is used to control the channel reservation, so that every interface can reserve maximum channel based on the solution of the flow-balancing problem. Further, the HWMP path selection metric has been augmented to cop up with directional multi-interface support to establish the optimal path based on the remaining bandwidth information obtained from the channel access protocol.
The above solution considers flow-balancing for the single-class homogeneous traffic only. The problem becomes more challenging when multiple classes of traffics are considered for the QoS based service differentiation. In this case, the minimum bandwidth demands for every classes of traffic need to be assured. Further the channel reservation should be based on the traffic class priority, i.e., a higher priority traffic flow should be able to reserve more amount of bandwidth compared to a lower priority traffic flow. This priority based bandwidth reservation strategy requires to support proportional fairness criteria among the contending flows. The centralized formulation for the proportional fairness criteria with the minimum traffic demand is known to be non- concave in nature [68]. Therefore, the optimization decomposition is difficult to design over a distributed network of mesh routers, considering the local information only. In this work, a sub-gradient optimization strategy is used to decompose the centralized
2According to IEEE 802.11s [25], mesh routers are termed as mesh stations (STA).
formulation, exploiting its log-convex property. The distributed decomposition enables local computation of the maximum capacity of every flows through every interface.
IEEE 802.11s MCCA protocol is augmented to support channel reservation based on the maximum capacity information. HWMP is enhanced with extra capabilities to take decision for the admission control, whenever a new flow is introduced in the network with a minimum traffic demand and a service class priority.
The next work in this direction augments the IEEE 802.11s mesh networking with multi-rate support. The classical rate adaptation algorithms, such as [69] and the references therein, consider the physical rate selection based on the channel fluctuation due to fading, shadowing and other physical effects. However, rate adaptation is more challenging in a mesh network due to the rate-hop-interference trade-off. Lower data rates can sustain for longer transmission ranges, which may improve network performance by reducing number of intermediate relays towards the destination router. However, in a moderate to high load network, longer transmission range can increase interference, resulting in throughput degradation. This work theoretically models this trade-off using the queuing network analysis over the IEEE 802.11s mesh architecture. A set of amendments are proposed over the standard MPM, MCCA and HWMP for multi-rate support. The interference information is obtained from the MCCA protocol, and the hop information is extracted from HWMP, in terms of the path quality metric. Based on these information, the optimum data rate is selected that can sustain over the current channel condition and provides improved performance in terms of data delivery. The optimum data rate, once selected, is used for the peer establishment using the MPM protocol.
In the next work, the thesis concentrates on the performance improvement in the path selection using HWMP. HWMP is based on a cooperation of two different path selection strategies - the proactive selection and the reactive selection. As mentioned earlier, proactive selection are prone to stale information due to the channel and the network fluctuations in a mesh environment, and reactive selection floods the network with control packets. This work proposes an alternative path selection strategy, called the‘Selective Greedy Forwarding’. In the proposed path selection strategy, the proactive HWMP is used to collect the initial information, where a set of potential forwarders are selected from the peer neighbors who can effectively forward the packets towards the destination. During the actual data transmission, the variability in the network and the channel conditions are explored to device a local greedy strategy to find out the best forwarder among the set of potential forwarders.
Finally the thesis reports the performance results of the IEEE 802.11s mesh protocols
from a practical indoor mesh testbed. The mesh testbed is built at the department of Computer Science and Engineering research labs, using IEEE 802.11n Multiple Input Multiple Output (MIMO) technology supported, dual interface Ra-Link [70] wireless chipset. The routers support open source Linux kernel, though the chipset driver is proprietary. Because of this limitation in the use of hardware drivers, some of the protocols, like the channel scheduling and mesh path selection are implemented as a loadable kernel module (LKM), that is executed when the specific functionality is required to be triggered.
The performance of the mesh network with improved channel access and mesh path selection support is analyzed from the results obtained from the testbed, and compared with the standard MCCA and HWMP.
In summary, the major contributions of this thesis are as follows.
The thesis proposes an improved mechanism for the capacity improvement in the IEEE 802.11s mesh networks, using multi-interface directional antenna support.
IEEE 802.11s MCCA protocol is enhanced to assist the balanced flow allocation over the directional interfaces.
IEEE 802.11s MCCA protocol is augmented for the QoS support during the channel reservation. The proposed modifications over the naive MCCA protocol enable service differentiation for the flows from different traffic classes, along with their minimum traffic demand. The compatibility of the proposed protocol with the standard EDCA based service differentiation is also analyzed in this context.
The rate-hop-interference trade-off in a multi-rate mesh network is analyzed using a queuing network modeling over IEEE 802.11s protocol. The modeling provides the theoretical basis for the rate adaptation in a mesh network.
An amendment is proposed over the standard IEEE 802.11s for multi-rate support with the physical layer rate adaptation techniques, based on the rate-hop-interference trade-off. MCCA and HWMP are augmented to collect necessary informations that are used to find out the optimal data rate in a particular network condition. The optimal data rate is used for the peer establishment during the operation of the MPM protocol.
The performance of HWMP is enhanced for the improved path selection strategy for a multi-interface mesh network, where the proactive approach is used for the initial information gathering, and a greedy selection mechanism is used during the actual path establishment. This provides an improved next-hop information, compared to
the naive HWMP approach considering the fluctuation in the network load, channel condition and interference.
Whenever necessary, the correctness of the proposed strategies are justified using mathematical proofs, theoretical arguments and modelings.
All the proposed schemes are implemented in Qualnet-5.0.1 network simulator framework [71], and the performance of the proposed schemes are evaluated using simulation results. Further, the performance metrics are also compared with the similar state-of-art schemes proposed in the literature.
The results from a practical indoor mesh testbed are used to analyze the performance improvement of a mesh network in terms of channel access and mesh path selection, and the results are compared with the naive MCCA and HWMP.