There are numerous routing protocols proposed for ad hoc networks in the literature. Because WMNs are multi-hop networks, the protocols designed for ad hoc networks also work well for WMNs. The main objective of those protocols is quick adaptation to the change in a path when there is path break due to mobility of the nodes. Current deployments of WMNs make use of routing protocols proposed for ad hoc networks such as AODV (Ad hoc On-Demand Distance Vector) [25], DSR (Dynamic Source Routing) [26], and TBRPF (Topology Broadcast based on Reverse Path Forwarding) [27]. However, in WMNs the mesh routers have minimal mobility and there is no power constraint, whereas the clients are mobile with limited power. Such difference needs to be considered in developing efficient routing protocols for WMNs. As the links in the WMNs are long lived, finding a reliable and high throughput path is the main concern rather than quick adaptation to link failure as in the case of ad hoc networks.
1.4.4.1 Routing Metrics for WMNs
Many ad hoc routing protocols such as AODV and DSR use hop count as a routing metric. This is not well suited for WMNs for the following reasons. The basic idea in minimizing the hop count for a path is that it reduces the packet delay and maximizes the throughput. But the assumption here is that links in the path either work perfectly or do not work at all and all links are of equal bandwidth. A routing scheme that uses the hop count metric does not take the link quality into consideration. A minimum hop count path has higher average distance between nodes present in that path compared to a higher hop count path. This reduces the strength of the signal received by the nodes in that path and thereby increases the loss ratio at each link [28]. Hence, it is always possible that a two-hop path with good link quality provides higher throughput than a one-hop path with a poor/lossy link. A routing scheme that uses the hop count metric always chooses a single- hop path rather than a two-hop path with good link quality. The wireless
links usually have asymmetric loss rate as reported in [29]. Hence, new routing metrics based on the link quality are proposed in the literature. They are ETX (Expected Transmission Count), per-hop RTT (Round-Trip Time), and per-hop packet pair. Couto et al. proposed ETX to find a high throughput path in WMNs [28]. The metric ETX is defined as the expected number of transmissions (including retransmissions) needed to successfully deliver a packet over a link. As per IEEE 802.11 standard, a successful transmission requires acknowledgment back to the sender. ETX considers transmission loss probability in both directions, which may not be equal as stated earlier. All nodes in the network compute the loss probability to and
from its neighbors by sending probe packets. If pf and pr are respectively
the loss probability in forward and reverse direction in a link, then the probability that a packet transmission is not successful in a link is given by
p = 1−(1− pf)(1− pr). The expected number of transmissions on that
link is computed as ETX = 1−1p. In [30] the routing metrics based on link
quality are compared with the hop count metric. The routing metric based on link quality performs better than hop count if nodes are stationary. The hop count metric outperforms the link quality metric if nodes are mobile. The main reason for this is that the ETX metric cannot quickly track the changes in the value of the metric. If the nodes are mobile, the ETX value changes frequently as the distance between the nodes changes.
As stated earlier, to improve the throughput the multi-radio multi-channel architecture is used in WMNs. In this case the routing metric based on link quality alone is not sufficient. It should also consider the channel diversity on the path. A new routing metric WCETT (Weighted Cumulative Expected Transmission Time) is proposed in [31], which takes both link quality and channel diversity into account. The link quality is measured by a per-link metric called ETT (Expected Transmission Time; expected time to transmit
a packet of a certain size over a link). If the size of the packet is S and
the bandwidth of the link is B, then ETT = ETX∗ SB. The channel diver-
sity in the path is measured as follows. If Xj is the sum of ETTs of the
links using the channel j in the path, then channel diversity is measured
as max1≤j≤kXj, wherek is the number of orthogonal channels used. The
path metric for pathpwithnlinks andkorthogonal channels is calculated
as WCETT(p)=(1−β)∗ n i=1 ETTi+β∗max1≤j≤kXj,
whereβ is a tunable parameter subject to 0 ≤β≤1. WCETT can achieve
a good trade-off between delay and throughput as it considers both link quality and channel diversity in a single routing metric.
The WCETT metric considers the quality of links and the intra flow interference along the path. But it fails to take into account inter flow
interference on the path. In [32], a new routing metric MIC (Metric of In- terference and Channel switching) is proposed for multi-channel multi- radio WMNs. This new metric considers the quality of links, inter flow interference, and intra flow interference altogether. This metric is based on Interference-Aware Resource Usage (IRU) and Channel Switching Cost (CSC) metrics to find the MIC for a given path. IRU captures the differences in the transmission rate and the loss ratios of the wireless link and the
inter flow interference. The IRU metric for a link k which uses channelc
is calculated as IRUk(c)=ETTk(c)∗Nk(c), whereETTk(c) is the expected
transmission time of the link kon the channelc, and Nk(c) is the number
of nodes interfering with the transmission of the linkk on channelc. The
CSC metric captures the intra flow interference along the path. CSC for a
node i is assigned a weight w1 if links in the path connected to it have
different channels assigned, andw2 if they are the same, 0≤w1 <w2. The
path metric for a given pathp, MIC(p), is calculated as follows:
MIC(p)=α∗
(link l ε p)
IRUl+
(node i ε p)
CSCi.
Hereαis a positive factor which gives a trade-off between benefits of IRU
and CSC.
1.4.4.2 Routing Protocols for WMNs
In [30], the authors proposed an LQSR (Link Quality Source Routing) pro- tocol. It is based on DSR and uses ETX as the routing metric. The main difference between LQSR and DSR is getting the ETX metric of each link to find out the path. During the route discovery phase, the source node sends a Route Request (RREQ) packet to neighboring nodes. When a node receives the RREQ packet, it appends its own address to the source route and the ETX value of the link in which the packet was received. The des- tination sends the Route Reply (RREP) packet with a complete list of links along with the ETX value of those links. Because the link quality varies with time, LQSR also propagates the ETX value of the links during the data transmission phase. On receiving a data packet, an intermediate node in the path updates the source route with the ETX value of the outgoing link. Upon receiving the packet, the destination node sends an explicit RREP packet back to the source to update the ETX value of links in the path. LQSR also uses a proactive mechanism to update the ETX metric of all links by piggybacking Link-Info messages to RREQ messages occasionally. This Link-Info message contains the ETX value of all the links incident on the originating node.
A new routing protocol for multi-radio multi-channel WMNs called Multi-Radio Link Quality Source Routing (MR-LQSR) is proposed in [31], which uses WCETT as a routing metric. The neighbor node discovery and
propagating the link metric to other nodes in the network in MR-LQSR are the same as that in the DSR protocol. But assigning the link weight and finding the path weight using the link weight are different from DSR. DSR uses equal weight to all links in the network and implements the shortest path routing. But MR-LQSR uses a WCETT path metric to find the best path to the destination.
In [32], the authors showed that, if a WCETT routing metric is used in a link state routing protocol, it is not satisfying the isotonicity property of the routing protocol and leads to formation of routing loops. To avoid the formation of routing loops by the routing metrics, they proposed Load and Interference Balanced Routing Algorithm (LIBRA) [32], which uses MIC as the routing metric. In LIBRA a virtual network is formed from the real network and decomposed the MIC metric into isotonicity link weight as- signment on the virtual network. The objective of MIC decomposition is to ensure that LIBRA can use efficient algorithms such as Bellman–Ford or Dijkstra’s algorithm to find the minimum weight path on the real network without any forwarding loops.