Scheduling, routing, and related cross-layer management through link activation procedures in wireless mesh networks

Abstract

In aWireless Mesh Network (WMN) [1] end users are provided with wireless broadband connectivity by means of a pre-defined system hierarchy. To describe this organization, several notations can be used. In the following, we adopt the terminology of [2]. The end terminals, also referred to as Mesh Clients (MCs), are connected to special nodes, denoted as Mesh Routers (MRs). These nodes do not generate traffic, since they are simply meant to relay the packets of their MCs. Additionally, some MRs, called Mesh Access Points (MAPs), can be provided with a wired connection, and can therefore act as gateways toward the Internet. The MAPs are also wirelessly interconnected to all the other MRs in a multi-hop fashion, without necessarily following pre-defined paths. Instead, an MC can interact only with the MR it is connected to. MRs form what is usually named as the backbone of the WMN, which can physically cover a large region in a wireless manner. This structure offers a good cost/benefit balance, since it almost entirely avoids cable set up. For this reason, it is deemed to be applicable in rural areas, where the deployment of wireline networks may be too expensive. WMNs can also be envisaged for dense residential or business areas, and in general, anyplace where the installation of cables is difficult because of physical obstacles. There are several possibilities to specify the Medium Access Control (MAC) used by a WMN. These are often related to existing standards, especially IEEE 802.11 [3] and IEEE 802.16 [4], parts of which are dedicated toWMNs. Actually, the first hop from any MC to its related MR is often assumed to employ a radio access interface different from the one used in the backbone, and entirely orthogonal (i.e., perfectly non-interfering) to it, e.g., since it uses another frequency, and possibly another technology. Moreover, the first hop may adopt management strategies typical of cellular networks [5], and is therefore conceptually simpler. For this reason, we will not investigate this part of the WMN in greater detail. Conversely, realizing the interconnections among MRs poses many theoretical challenges, most of which are common to all kinds of multi-hop networks, such as ad hoc and sensor networks. However, when revising them for WMNs, some important properties come into play. Usually, MCs can be portable devices, whereas MRs and MAPs are not mobile. Therefore, the backbone management does not suffer from most mobility issues, neither at the transport layer (i.e., paths do not need to be updated), nor at the physical layer (channel variability is relatively moderate). Moreover, communications in a WMN are usually to or from the Internet, thus all routes have either the source or the destination in a MAP. Finally, as MRs can be easily placed near to a power outlet, energy saving is not an issue. These properties considerably distinguish the backbone of WMNs from an ad hoc network (for what concerns pre-defined hierarchy and absence of mobility) or a sensor network (lack of terminal battery limitations). The issues which arise in the backbone management relate to different layers of the protocol stack. On the one hand, the creation of low-interference and high-rate paths to the MAPs is key to achieve good rates at each MR. This may also involve the exploitation of multiple channels as, for example, MRs can own several Network Interface Cards (NICs), which can simultaneously operate on different frequencies. On the other hand, the link layer needs to schedule packets over multiple links in order to achieve good transmission parallelism and possibly forward more data towards the MAPs at the same time. The main problems which will be investigated by our analysis are: routing algorithms, i.e., network level procedures to discover efficient paths which connect the ordinary MRs (and therefore their MCs) to the MAPs. Note that routing strategies designed for ad hoc networks usually admit also peer-topeer communications, which are not common for WMNs. Moreover, the goal in WMNs is more often to obtain high system throughput rather than maximizing battery lifetime. link scheduling, which involves medium access level procedures to activate communication links. Its goal is to ensure network connectivity while at the same time satisfying physical constraints related to technology, interference and network management. cross-layer management, operating at an intermediate level with both network and link layer procedures, jointly addressing these problems. The aforementioned issues involve other related topics, which are also worth discussing. In certain cases very broad subjects are involved, which will be discussed here only for what concerns their impact on the definition of routing and scheduling strategies. There are also other aspects of these matters which fall out of the scope of the present article, and therefore, will not be discussed here in detail. However, the reader will be addressed to external references to find further material on them. Some of the related problems which will be framed into our analysis are: channel assignment and node placement: in our analysis, these are considered to be aspects of network deployment, which means they have already been per formed at the time routing or scheduling strategies are sought. However, it will be briefly outlined how it is possible to incorporate them into the same cross-layer framework used for routing and scheduling with a modular approach, thus with no need for significant modifications of the reasonings presented in the rest of the article. models of wireless interference: for this point, two important considerations must be made. First of all, we propose a detailed review and classification of the possible approaches to characterize interference. We try to resolve terminology ambiguity due to the use of different names for the same model or of the same name for distinct models in the literature. Moreover, we discuss the choice of the model itself, which is driven by two contrasting aspects. On the one hand, the interference model should be as accurate as possible. In this sense, the use of heavily simplified interference models may end up in poor algorithm performance when applied to realistic cases. On the other hand, a certain degree of approximation is unavoidable as related to the properties of the Medium Access Control (MAC) protocol. In fact, in a layered network management, algorithms operating on top of the link layer necessarily abstract some aspects of the physical layer, such as interference. For these reasons, we will first concentrate our analysis on general results which hold true regardless of the interference model, such as theoretical performance bounds. Then, we will discuss how these findings translate to practical cases, at which point different interference models need to be taken into account. The rest of this chapter is organized as follows. In Section 8.2 we give a brief overview of the problem studied and we clarify terminology and notations employed in the rest of the chapter. In Section 8.3 we present a review of the works which discussed related topics in a way applicable to WMNs. In Section 8.4 we mathematically formalize the problem, in particular identifying the constraints determined by capabilities of the terminals and wireless interference. This latter aspect, in particular, is discussed in depth in Section 8.5, proposing a classification of interference models, and also touching MAC protocol issues. In Section 8.6 we give both theoretical and practical evaluations of the performance of WMNs. Even though the problem is NP-complete and exact approaches are hard, we present some original analytical results which determine both upper and lower performance bounds, and we give quantitative insights by applying them to sample WMN topologies. Finally, we present the conclusions. © 2007 Springer Science+Business Media, LLC. All rights reserved.