Wireless Mesh Networks A Survey on Routing Approaches Hamid Hamraz, Golnar Bidkhori Computer Science Department, University of Kentucky Davis Marksbury Building, 329 Rose Street Lexington, KY 40506
[email protected],
[email protected]
Abstract. Wireless Mesh Networks (WMNs) consist of two types of nodes: the mesh nodes that are generally stationary, serving as a multihop wireless backbone providing connectivity to an infrastructure, such as internet, for the usage of other mobile nodes connected wired or wirelessly to the meshes through possibly multiple hops. Obviously, routing is one of the main concerns, affecting the overall efficiency of the network. In this survey, we try to present some recent WMNs routing protocols, where we discovered two main types of routing mechanisms – traditional and opportunistic. We describe characteristics of each type, and present different sample protocols of each. There are some proposals that have the properties of both types, as well as the ones approaching routing from a completely different perspective – not easily classifiable as traditional or opportunistic. While the opportunistic approaches seem to be more recent and justified for wireless networking, research on traditional routing mechanisms is still on going. Depending on the particular properties of the wireless network on which the protocol is going to be implemented, both traditional and opportunistic routing mechanisms - as well as the other quite different approaches - have something to contribute, and may outperform each other. Keywords: Wireless Mesh Networks, Traditional Routing, Opportunistic Routing, Survey
1
Introduction
Wireless Mesh Networks (WMNs) [1] are a relatively new multi-hop wireless framework, having a lot in common with Mobile Ad hoc Networks (MANETs); WMNs are in fact a special type of MANETs. Like MANETs, they consist of a set of wireless nodes communicating with each other. Generally, each node can both be a host, and a wireless router. A WMN consists of adfa, p. 1, 2011. © Springer-Verlag Berlin Heidelberg 2011
a set of mesh nodes that are usually not resource-constrained as they are stationary or probably with little movements. Mesh nodes can serve as a wireless backbone for the entire network, where other nodes can connect to them through a wired or a wireless link. In most proposed applications, WMNs provide the connectivity to an infrastructure – usually the internet (Fig. 1). We will use gateway referring to mesh nodes providing internet connectivity whether directly or indirectly depending on the context.
Fig. 1. A typical WMN with wireless links between the meshes[2]. There are four types of links presented. Except for the intra-mesh links (that have to be wireless), all other links can be either wireless or wired.
The main advantage of WMNs in comparison to traditional broadband Internet access technologies (cable-modem and xDSL) is the remarkably reduced initial investment and deployment time. The main advantage in comparison to fixed wireless metropolitan area networks is the customizable coverage (especially in areas with significant obstructions – trees, high-rise buildings) and reliability (multiple available routes can avoid failed nodes and poor links). For instance broadband home networking, where locating the access points (Aps) is usually not straight forward. A home usually has many dead zones, where installing wired Aps incurs cost and inconvenience. By placing mesh routers in the range of each other; all dead zones can easily be covered. The main drawback of such deployments, however, is the reduced bandwidth experienced by users. Other examples of networking where flexible mesh connectivity can eliminate the cost and time required for laying wires are community and neighborhood, enterprise networking, security surveillance systems, and health and medical networking between the rooms in a hospital.
1.1
Routing
As mentioned earlier, from a taxonomic point of view, WMNs are a particular type of MANETS, i.e. they share the same multi-hop characteristics and mobility-related issues as MANETs. However, there are also remarkable differences between WMNs and general MANETs due to the presence of the gateways serving as the routing backbone of the network. Most WMNs are designed to provide connectivity to a distribution system (usually connected to the Internet) through the gateways. Moreover, in WMNs, most of the traffic is expected to flow between the clients and the Internet while in general MANETs, the assumption is that any node is equally likely to be the source or the destination in a traffic flow. Additionally, while in MANETs the mobility pattern is homogeneous, in WMNs the gateways are almost stationary and the other nodes are able to roam freely. Routing is a fundamental issue in any wireless network. Any strengths or weaknesses in the routing scheme directly affect the characteristics of the network. Several advantages of the WMNs are directly enabled by applying the appropriate routing protocol [2]. First of all, the routing protocol should be reliable, i.e., in case of a route failure, fast rerouting is necessary. For instance, if a gateway fails, the routing protocol should be able to quickly reconfigure the network connectivity, and take care of the orphan clients by redistributing them among other gateways. Secondly, the routing protocol should provide smooth hand-offs for the mobile nodes. Thirdly, if the routing protocol is not scalable and have large overhead, the deployment of the WMNs over a large-scale application won’t be possible. Last but not least, the routing protocol should provide Quality of Service (QoS) by selecting the best route depending on the traffic type. As far as we discovered, there are two main types of routing protocols for wireless networks: Traditional Routing: where the routing ideas are inspired by the routing in wired networks. The most important metric in this class is the hop-count (HC). In wired networks, each hop is deterministic, i.e. it works or it fails. This fact approves the hop-count as an effectively descriptive metric. The similar assumption remains true for the traditional routing where the dominant metric for comparing the routes is hop-count. Most of the traditional routing protocols adopt a flooding mechanism to discover the best route. When the flooding reaches the destination, the destination will reply back to the source and the route is established when the source receives the reply. After that, transmission of the data packets through the route starts. The main focus of the models proposed for this type of routing is on how to
reduce or eliminate the flooding in order to damp the overhead and improve the performance. Using the nodes’ location information to decrease the flooding diameter as in LAR [3], DREAM[4], and ZRP[5], or electing dedicated relay nodes as in RSPF [6] are main strategies for restricting the flooding. Opportunistic Routing: where the unpredictable nature of the wireless medium is taken into account. Unlike the wired networks, not all hops in a wireless network are equal. This challenges the hop-count to be the appropriate metric for routing as it has been used in traditional routing. Therefore, other metrics quantifying the quality of links using link loss rate, packet transmission time, or signal-to-noise ratios has been proposed. The most widely-used and effective metric of this sort is Expected Transmission Count (ETX)[7], where the reliability of a link is measured by sampling the number of transmissions required to reliably pass a packet through the link. As in EXOR [8](the seminal proposal), opportunistic routing exploits the broadcast nature of the wireless medium and does not commit to a particular route before data transmission. Instead, the sender broadcasts its data; among the nodes that hear the transmission, the one closest (measured by metrics aware of the link quality such as ETX) to the destination is selected to forward the data. In this way, opportunistic routing can effectively combine multiple weak links into a strong link and take advantage of transmissions that reach unexpectedly near or unexpectedly far.
There have been many routing algorithms proposed for MANETs, such as AODV [9],and DSR, [10]. However, the particular characteristics and requirements of WMNs prompt the development of new routing protocols that can significantly outperform the general MANET routing protocols. In this survey, we aim to describe several recently proposed routing protocols geared toward the WMNs. The rest of the paper is structured as follows. In section 2, we go through some traditional routing schemes, where we introduce MRP[2], AODV-DF[11], MMESH[12], and ORRP[13]. Then we switch to the opportunistic schemes, where we will describe SOAR[14], GPR[15], ROMER[16], and TDiCOR [17] in section 3. Section 4 introduces a few special routing mechanisms we couldn’t easily classify as one of the two types above. There, we introduce DART[18], HWMP [19], FBR [19], and oblivious routing[20]. We then go ahead and present a comparative study on the routing protocols introduced, in section 5. Finally, section 6 concludes the survey.
2
Traditional Routing for WMNs
In this section, we are going to present a few WMNs routing mechanism classified as traditional routing as described above. They are mainly focusing on adjusting the trade-off between elimination or reduction of flooding to decrease the overhead incurred on one side, and trying to improve the robustness of packet delivery by employing some control mechanism producing overhead on the other side.
2.1
MRP Wireless mesh networks routing protocol
In the design of MRP[2], the most common case of traffic, i.e., downloads to and from the internet, is considered. The gateways are assumed to be hierarchically connected to a super gateway which provides the connection to the internet. Any node in the WMN will only know how to reach one gateway and is, in general, reachable only from a gateway. Any small amount of clientto-client traffic can be routed through the common parent of the clients (potentially the gateway). In essence, the routes to and from the gateway form a tree rooted at the gateway. The tree-like structure will ensure that there are no cycles in the routes. Three versions of the protocol are introduced: 1. MRP On-Demand: a node trying to join the network locally broadcasts a route discovery message (RDIS). Any neighboring nodes hearing this, will reply with a route advertisement message (RADV) including its own route to a gateway, as well as some metrics about the route such as hop-count, route stability, minimum delay, maximum bandwidth, and minimum packet loss. The joining node will then collect all the RADVs, and chooses one of the neighbors with the best route toward the internet to connect. The route selection criteria can be easily tailored to the application. At this point, the node is in the half-connected state since it has a route toward the internet, but the hosts on the internet can’t reach it. The node then sends a registration request (RREG) to the gateway. The intermediate nodes set the reverse path while relaying the RREG. The gateway, upon the receipt of the RREG, will reply with a registration acknowledgment (RACK), and forwards up the RREG to the super gateway where it can set the return path for packets coming from the internet to the node. Upon receipt of the RACK, the node is in full-connected state where it can start sending and receiving data.
2. MRP Beacon in which any node in the full-connected state periodically broadcasts beacon messages (basically the same as RADV). A joining node will not need to initiate RDIS anymore. It simply listens to hear and collect beacons. The route selection and registration procedures are identical to MRP On-Demand. This version of protocol, although might have a little more control overhead, can detect route failures much quicker. 3. MRP Hybrid in which everything is quite similar to MRP Beacon except that the joining node does broadcast RDIS, and simply waits as much as the minimum of the beacon period and the waiting time for MRP On-Demand. In this version, the overhead is again a little bit increased. However, the route failure recovery and the joining procedure for a disconnected node get quicker.
In this protocol, when a gateway fails, all the nodes in its sub trees become disconnected and should reconnect to other gateways. A link failure prompts a similar situation where the downstream node and all of its children get disconnected, and each of them should individually reconnect. The paper has also proposed a strategy to distinguish temporal link failures and the permanent ones in order to avoid the costly tree reconfiguration to some extent. The main advantage of this protocol is the total elimination of flooding. However, in return the client –to-client routes is suboptimal. The proposed protocol is quite simple and straight forward to implement. Through extensive simulations, it has been verified that for WMNs, the proposed routing protocol outperforms general purpose MANET protocols in terms of routing overhead, packet delivery ratio, network throughput, end-to-end delay, and average hop-count.
2.2
Directional Flooding for Wireless Mesh Networks
The AODV-DF (Directional Flooding) protocol [11] is an improvement to AODV, in which the flooding is tried to be directed toward the destination, thereby drastically reduce the overhead incurred by the uncontrolled flooding scheme in AODV. Like MRP, the traffic from and toward the gateways is under consideration – for the peer-to-peer routing, the original AODV is utilized. To support DF in AODV, a gateway periodically broadcasts its identity. AODV-DF embeds the gateway’s address and hop-count information into the HELLO packets. There is no additional overhead here since original AODV
routing protocol periodically exchanges HELLO packets. Each node maintains an HC variable to indicate the number of hops to a gateway, which is initially set to be infinite. Only a gateway’s HC value is set to 0. The gateway hello messages flooded outward to let the other nodes in the WMN update their HC to the gateway. Sequence number in HELLO packet is used to determine the timeliness of each packet. The route discovery of AODV-DF is basically based on AODV. AODVDF uses a modified RREQ packet with an HC field to indicate the number of hops to a gateway. When a source node desires a route to a destination node for which it does not have a route, it broadcasts a RREQ packet to all its neighbors. If the destination is another peer within the WMN, RREQ packets are flooded in the same way as AODV. Otherwise, RREQ packets are directionally propagated toward the gateway (it is assumed that source nodes can determine whether their destinations locate within WMN or not). When a source node wants to send data to a destination node outside the WMN, if it does not have a route to the gateway, it broadcasts the RREQ with an HC set to its HC value. On receiving RREQ packets, a node compares the HC value on the RREQ with its HC. If the HC value on the RREQ is equal to or less than the node’s HC, the intermediate node discards RREQ packets. Otherwise, it replaces the HC value on the RREQ with its own HC value and then re-broadcasts the RREQ to all neighbors. If an intermediate node does not know its hop count to the gateway (i.e., HC is infinite) before gateway and hop count discovery, it broadcasts RREQ packet to all neighbors as the same way in AODV. When a node receives the RREQ, it establishes a reverse route to the RREQ source in its routing table, and it either replies to the RREQ if it has an entry for the gateway in the routing table or it forwards the RREQ. Eventually, the RREQ reaches the gateway and the gateway send back an RREP using unicast to confirm the path. The node receiving the RREP sets up a forward route to the gateway and desirable routes can be set. Route maintenance is similar to that of AODV. An existing routing entry may be invalidated if it is not used within a specified time interval, or if the next hop node is no longer reachable. In these cases, an invalidation notice is propagated to the neighbors that have used this node as the next hop. When a node detects that a route to a neighbor is no longer valid, it removes the invalid entry and floods a route error message to the nodes that are using the route. AODV-DF was compared to AODV via the NS-2 simulator. It is reported that the results verify that AODV-DF outperforms AODV, while at the same time incurs much less overhead.
2.3
Orthogonal rendezvous routing protocol for WMNs (ORRP)
Scalability and connectivity are the two important requirements for routing in wireless mesh networks. All proactive and reactive routing protocols flood so much information during rout dissemination and route discovery in the network. Flooding causes scalability problem. Orthogonal Rendezvous Routing Protocol (ORRP)[13] is a lightweight, but scalable routing protocol for wireless mesh networks which utilized directional communications to relax the requirement for information like node localization and coordinate space embedding. ORRP is an unstructured forwarding ORRP is based on two ideas: Local directionality is sufficient to maintain forwarding of a packet on a straight line with high probability even in sparse, bounded networks. Two sets of orthogonal lines in a plane intersect with high probability even in sparse, bounded networks.
ORRP assumes that each node has directional communication capacity so it has a local sense of direction. ORRP forwarded packets in orthogonal directions to find paths from a source to a destination and it does not make any assumption on location discovery. To do that, it makes three assumptions: Local Sense of Direction: Each node with transceivers/antennas has its own local perception of direction. So it can consistently send out orthogonal directions. This is done by selecting one of the antennas as the “local North” and assigning angels to the others according to the antenna which is selected. Neighbor Discovery: Any node knows i) its 1-hop neighbor and ii) the given interface to send packets to that neighbor. Ability to Transmit/Receive Directionally: Nodes should be able to communicate directionally over its antennas. This is done by different hardware such as antennas and FSO transceivers. ORRP source and ORRP destination send route discovery and rout dissemination packets in locally-chosen orthogonal directions. Connectivity occurs when these paths intersect. Even in sparse networks, ORRP can achieve connectivity with high probability.
2.4
Multi-Path Routing in WMNs
Since the main traffic in WMNs is between meshes and toward/from the internet, maintaining a single best route for a traffic flow causes some of the relaying meshes be the hotspot of the traffic, thereby degrading the performance. Resolving this, what Multi-path routing for MESH networks (MMESH) [12]proposes, is to maintain multiple routes between each pair of the meshes, and to synergistically transfer data along those paths to balance the traffic. Initially, all meshes having at least a route to the internet broadcast their route information along with some statistics about the routes. This information is received by next tier meshes, allowing them to choose some routes. After that, they go through a registration-acknowledgement procedure (like MRP) to establish their chosen routes to and from the internet. This phase is intended for the mesh nodes not connected directly to the internet to establish multiple routes toward the internet through gateways (mesh nodes directly connected to the internet). After initial route discovery phase, each mesh node continuously monitors the performance of all the active paths. On finding new routes or stale routes, a mesh node appropriately validates its routing tables and announces this information to its neighboring nodes. Neighboring nodes promptly update their routing table using this information and further propagate the changes. Note that since mesh nodes are generally static, this phase of the protocol is not done so often. However, some optimizations for discovering rout failures, and fast route recovery are discussed in the paper. Once the multiple routes setup, the task is to divide the traffic among these routes to balance the load, without degrading the performance. Simple and effective, each node sends every packet to a different next hop in a roundrobin fashion, thus uniformly distributing traffic over in the network. However this scheme suffers from setback due to possible out of order delivery of packets belonging to same flow. The transport layer protocols need to maintain a large buffer to cope up with out of order delivery and can lead to unnecessary loss with protocols such as TCP. Additionally, mitigating congestion and avoid routing traffic through congested routes, if it is found that a particular next hop node has high average queue length over a period of time (an indication of a congestion), that node is temporarily skipped and the traffic is sent through other neighboring nodes.
The authors report that the MMESH protocol has been implemented and compared to single-path routing counterparts in NS-2 simulator. Simulation results reveal that single path routing leads to severe packet loss and consequently very low end –to-end throughput for longer hop length flows. MMESH helps in dramatically increasing the performance of the longer hop length flows. Even with naive round robin scheduling, the throughput is substantially increased. Congestion aware routing improves the throughput by almost 100% compared to the pure round robin scheduling.
3
Opportunistic Routing Protocols for WMNs
The goal of opportunistic routing is to maximize the progress each transmission makes without causing duplicate transmissions or incurring significant coordination overhead [14]. In order to achieve this goal, several important design issues should be addressed. Forwarding Node Selection: While opportunistic routing defers the final route selection after data transmissions, the candidate forwarding nodes should still be selected in advance. This is necessary because the number of duplicate transmissions and coordination overhead tend to increase with the number of forwarding nodes. Without judicious forwarding node selection, the overhead of opportunistic routing might offset its benefits. Avoid Duplicate Transmissions: When multiple nodes overhear a transmission, it should be ensured that only the node closest to the destination forwards it. The best forwarding node should be selected in a cheap and distributed way. Loss Recovery: Since the MAC layer offers no reliability support for broadcast, it is important for opportunistic routing protocols to efficiently detect and recover packet losses. Rate Control: Determining an appropriate sending rate is important for opportunistic routing. Due to wireless interference, uncontrolled massive transmissions take away available bandwidth from the subsequent hops and significantly degrade performance. For multiple simultaneous flows, rate control is also useful to improve fairness and avoid starvation. While traditionally rate control is typically considered the responsibility of a transport layer protocol, previous works [21],[22],[23] have shown that the existing congestion control at the transport layer is not effective for multi-hop wireless networks. A joint optimization of rate limiting and routing is more promising.
3.1
Simple Opportunistic Adaptive Routing Protocol for WMNs
Simple Opportunistic Adaptive Routing protocol (SOAR) [14] is an optimization to its predecessor EXOR [8]that tries to optimize the opportunistic behavior by selecting the most appropriate neighbors for forwarding each packet. It also uses the ETX as the metric for measuring closeness in selecting the routes. It has the following major components for overcoming the opportunistic routing design challenges: Default Path Selection: Every node measures and maintains the network topology proactively. Each node measures the loss rate of its links to and from its neighbors by broadcasting one probe packet every second and counting the number of probes received in the last 10 seconds in order to calculate the link ETX. Each node maintains an exponentially weighted moving average of ETX samples. The default path is the shortest path between the source and destination in terms of ETX. Forwarding Node Selection: In order to leverage path diversity while avoiding duplicate transmissions, SOAR relaxes the actual route that data traverses to be along or near the default path. Different from existing opportunistic routing protocols, SOAR constrains the nodes involved in routing a packet to be near the default path. This prevents routes from diverging and minimizes duplicate transmissions. Moreover, this forwarding node selection also simplifies coordination since all the nodes involved are close to nodes on the default path and can hear each other with a reasonably high probability. Therefore, overheard transmissions can be used to coordinate between forwarding nodes in a cheap and distributed way. The forwarding nodes are ordered based on their proximity to the destination. The number of forwarding nodes is bounded to a maximum number. In addition to that, if the nodes closest to the destination have a sufficiently reliable virtual link, other forwarding neighbors may be pruned. The forwarding nodes selection is done by each hop on each packet, allowing for the forwarding list to quickly adapt to network conditions. The forwarding nodes list is to any data packet being broadcasted. Priority-based Forwarding: The neighboring nodes hearing the broadcast will drop the packet if they are not within the list of forwarding nodes. Otherwise, each forwarding neighbor sets a timer based on its position in the list to forward the packet – the farther in the list, the bigger the timer. This ensures that the closer nodes to the destination will broadcast sooner. In addition to that, other neighbors with bigger timers will drop the packet if they overhear the broadcast, which prevents duplicate transmissions.
Local Recovery: SOAR uses per-hop network-layer ACKs and retransmissions to provide the best effort reliability. Each node on the forwarding path will keep retransmitting a packet until either the packet is acknowledged by a node closer to the destination or the maximum retry count is reached. Rate Control: In SOAR, each data flow uses end-to-end ACKs from its destination node to control its source’s sending rate; the size of the transmission window is adjusted based on the ACKS from the destination received by the source.
The SOAR algorithm is reported to be implemented both on NS-2 and a real network of PCs. The experiment result shows that SOAR outperforms EXOR, as well as a shortest-path non-opportunistic counterpart in most scenarios.
3.2
Geographic Probabilistic Routing Protocol for Wireless Mesh Networks
Geographic Probabilistic Routing protocol (GPR) [15]is an opportunistic routing protocol working between mesh routers in WMNs. It is very much like the SOAR algorithm introduced above, where the geographical positions of nodes are also utilized. This is indeed a reasonable assumption because mesh nodes are usually stationary and maintaining the location of them as well as an image of the overall mesh nodes networked to each other is straight forward. Different to SOAR, the forwarding node selection is initially based on the geographical position. The forwarding nodes should fall within an angle less than 180 degrees, facing toward the destination node. To ensure packets do not cycle in a loop, every time the packet is going to be forwarded, it must never be send to nodes farther to the destination. There may be a lot of nodes can be chosen as relay nodes, and some of them have very low transmission reliabilities. Among the potential relay nodes, the ones with highest ETX to the destination are chosen. In other words, GPR chooses the relay node which can make the largest expected positive forward distance. Once get the forwarding neighbors subset, source sorts the candidate nodes in the decreasing order of their effective forward distance in the direction of Destination. The relay nodes set a timer to broadcast an ACK based on this order. So a candidate node making the largest forward distance in direction of the destination sends ACK firstly. This ACK helps to detect losses. Moreover, other relay nodes hearing the ACK drop the packet.
It has been reported that through simulation in the NS-2, it’s verified that GPR outperforms SOAR, incurs less overhead, thereby achieving more scalability to the larger networks.
3.3
Resilient Opportunistic Mesh Routing for Wireless Mesh Networks
Resilient Opportunistic MEsh Routing (ROMER) [16]balances long-term route stability and short-term opportunistic performance. It builds a runtime, forwarding mesh on a per-packet basis that offers a set of candidate routes. The actual forwarding path by each packet opportunistically adapts to the dynamic channel condition and exploits the highest-rate wireless channels at the time. To improve resilience against lossy links, mesh failures or mesh routers under DoS attacks, ROMER delivers redundant data copies in a controlled and randomized manner over the candidate forwarding mesh. ROMER has two components that work in concert for efficient performance tradeoff: Runtime Candidate Mesh: a credit-based approach that allows each packet to build its forwarding mesh on the fly. Each packet can specify its own credit based on its individual resilience requirement. The next-hop mesh is formed on the fly as the packet moves toward the destination; the packet may follow a subset of the candidate interleaved paths offered by the mesh to resist against channel and node outages. Opportunistic and Randomized Forwarding on the Mesh: to maximize the end-to-end throughput, ROMER uses greedy forwarding to opportunistically deliver the data packet along the dynamically highest-rate link with high probability, as well as other high-rate downstream links with fair probability. Controlling overhead and increasing robustness, ROMER also uses randomized forwarding at each intermediate node.
Comparing to the popular single-path routing protocol in mesh networks, ROMER enhances both resilience and performance. Other protocols rely on link-layer retransmissions to recover from link loss, incurring repetitive retransmissions over persistently poor links. Tackling this, ROMER exploits path diversity by transmitting to multiple receivers, at least one of which is more likely to be in good channel condition. Leveraging path diversity also enables ROMER to opportunistically select the highest-rate link. The authors reported that they compared ROMER to single path and two disjoint path conventional routing protocols via NS-2 simulator. The simula-
tion shows that ROMER is able to achieve up to 195% throughput gain compared to the conventional single-path routing. Its packet delivery ratio can be 40% more compared to the single-path and two-disjoint-path routing protocols, when the channel loss is about 5% and the node failure is about 10%.
3.4
Cooperative Opportunistic Routing using Transmit Diversity in Wireless Mesh Networks
Transmit Diversity Cooperative Opportunistic Routing (TDiCOR) [17] efficiently exploits multi-user and transmit diversity to improve the overall throughput in wireless multi-hop networks. Neighboring nodes cooperate through simultaneous transmissions in order to overcome the destructive effects of fading. TDiCOR uses distributed transmit diversity to increase the robustness of acknowledgements as well as data transmissions while preserving the opportunistic nature by using multiple forwarding nodes for packet relaying. The main contribution of TDiCOR is to incorporate transmit diversity into opportunistic routing, i.e. multiple sources transmit the same signal. This helps the receiver to get a stronger signal which is more robust to the issues arising from the unpredictable nature of wireless links, such as fading. In the other words, TDiCOR provides a many-to-many communication scheme, i.e., multiple sources as well as multiple forwarding nodes, in order to opportunistically overcome challenges arising from the wireless nature of the medium, thereby optimizing the network performance. Clearly, adjusting the trade-off between overhead and the network throughput is one of the major goals of TDiCOR. TDiCOR is reported to outperform traditional routing (i.e. DSR) in typical outdoor scenarios in terms of throughput by 30% and by 50% in indoor scenarios with high shadow fading, without consuming additional bandwidth or additional hardware resources.
4
Other Protocols
In this section, we are introducing some routing mechanism applicable to WMNs that cannot easily be classified as traditional or opportunistic. In fact, the following proposals are trying to approach the routing from a totally different viewpoint, such as assigning dynamic addresses for the mobile nodes
based on their locations, or using inspirations coming from the potential fields in physics, or cross-layer design.
4.1
Dynamic Address routing for scalable ad hoc and mesh networks
Most of the routing protocols for ad hoc networks use flat static addressing, so they have to keep track of each node which will make a huge overhead. Dynamic addressing can solve this problem and it enables scalable routing in ad hoc networks. Dynamic Address Routing Protocol (DART) [18] use a form of proactive distance-vector routing. DART protocol separates the routing address and the identifier of a node. The identifier is a unique number that does not change through the lifetime of a node. The routing address is dynamic and changes as the node moves in the network. There are three functions in DART: Address Allocation: Allocates an address to a node such that the address shows the location of the node in the network. Node Lookup: Is a lookup table which maps each node identifier to its current network address. Routing: Delivers packet from sender to receiver. Whenever a node joins a network it listens to the periodic routing updates of its neighboring nodes to find an unoccupied address. It first selects which neighbor to get the address from. It will choose the neighbor with the highest –level insertion point. The insertion point is the highest level for which no routing entry exists in a given neighbor’s routing table. Then this new node will register its unique identifier and the obtained address in the distributed node lookup table. All nodes participate in lookup table and each of them stores some entries. This node is called the anchor node of the identifier. There is a hash function that takes an identifier’s input and returns the address that the entry can be found. So when a node wants to send a packet to a node that it only knows its identifier it will use the lookup table to find its address. Each node also keeps some routing state about all of its Level-i siblings. The average size of the routing table is less than 2log n, where n is the number of the nodes in the network. The result of the Comparing DART protocol to reactive protocols (AODV, DSR) and a proactive routing (DSDV) shows that the dynamic addressing and
proactive routing together provides such a great scalability advantages. Also this comparison shows that proactive routing exhibits less overhead than its reactive counterpart.
4.2
Hybrid Wireless Mesh Protocol
Hybrid Wireless Mesh Protocol (HWMP) is a simple routing protocol for wireless mesh networks. HWMP is based on tree-based routing and AODV[19].In HWMP each mesh point discovers and monitors neighbor nodes, so it relies on the management protocol peer link Since HWMP supports two types of path selection protocols, it is a hybrid protocol. Since HWMP is based on AODV it has some problems like, when a congested hot spot which is responsible for packet losses and long delays forms in a path, the congestion will remain until new routes are established. Also a node failure creates requests for route re-establishment and it will make high traffic volume and degrades the performance of the network.
4.3
Field Based Routing (FBR)
Field based routing protocol is inspired by the potential fields in physics which describes the potential energy which is associated with each point in space [19]. A potential field models the “energy” of a node by creating a scalar field around each destination node. This field will decreases according to the distance. The path which has the lowest energy required is the path with steepest slope, so the packets will be routed along that path. So the direction of the steepest is to the neighbor which has the greatest potential value. But it is possible that different nodes have the same potential field. If this happens the potential field will be obtained in the same manner as in the physical creating the linear superposition of the influence of individual nodes of the group. If a node wants to send a packet to the Internet, it takes the neighboring node with the greatest potential field as its default gateway so the Internet gateway is always achieved.
Fig. 2. Field Based Routing in wireless mesh networks
4.4
Distributed potential field based routing and autonomous load balancing for WMNs
Distributed potential-field-based routing scheme for any cast wireless mesh networks is robust to sudden traffic and network perturbations. It balances the load among multiple gateways and mesh nodes with little control overhead. In conventional field-based routing, in order to form a field gradient, scalar values are assigned to each node, so the packets will be forwarded to a node with the lowest gradient. The distance to a destination will be considered in assigning scalar filed value to each node. But in this protocol scalar field values are determined by a distance to a destination and also a degree of congestion. To achieve the global congestion control, only one-hop neighbor’s potential information is considered.
4.5
Cross layer
Cross layer [24]is a new on-demand, fully distributed cross layer integrated QoS routing algorithm. In this routing algorithm we assume that each WMR has a fixed transmission power and certain signal receiving threshold. Each mesh router maintains its immediate neighboring nodes. Also time division duplex (TTD) is used so each link can only transmit or receive signal but not both at the same time. It
is required that in routing discovery phase each receiving of each link records one-hop delay, link throughput and PER information. Routing discovery procedure is initialized when new traffic flows are accepted by certain nodes. When a node serves as the source, it will generate a request packet REQ which contains the QoS flow constraints and will start a timer when sending the REQ through the allocated time slot in the control channel to its one-hop neighbors. If this router does not receive a reply message before the timer expires, it will generate a request packet and broadcast it to the whole network because of the possible packet loss. When a node which is the neighbor of the sender router receives REQ, it will average previous one-hop delay, link throughout and PER measurements and then piggybacks this information in REQ and sends it to all its one-hop neighbors through the allocated time slots in the control channel. All the nodes in the network will repeat these procedures until gateway node receives the request message, and then it will send a reply packer REP back through different routes to the mesh router which initialized this procedure. When that node will choose the best time by calculating the QoS performance metric before the timer expires.
4.6
Oblivious routing for wireless mesh networks
Traffic routing has an important role in determining the performance of a WMN. The routing algorithms which are proposed are heuristic routing algorithms or theoretical studies. Heuristic routing protocols adapt well to the dynamic environment of wireless networks but they do not have the theoretical foundation to analyze that how is the performance of the networks. Theoretical studies formulate mesh routing algorithms as optimization problems. They presume that traffic demand is static. But actually the traffic demand is dynamic and it is hard to estimate the traffic demand. Oblivious routing [20]protocol can provide the optimal worst-case performance on all of the possible routing traffic demands.
5
Comparative Study
Comparing the two general routing schemas presented, i.e., traditional and opportunistic, we would say that both classes are trying to optimize the same
thing through two vastly different methodologies. In traditional routing, the main goal is to adjust the trade-off between routing performance and the overhead incurred. This is done by focusing on how to decrease the control overhead in a way to not to degrade the network throughput. The opportunistic approaches, on the other hand, are trying to improve the performance by introducing some extra minimal control overhead in order to probabilistically deliver the packets as fast and as reliable as possible. MRP, for instance, provides sub-optimal peer-to-peer routing in order to eliminate flooding. SOAR, on the other hand, tries to reliably and quickly deliver packets by redundant transmissions. AODV-DF tries to limit the flooding area incurred by the general AODV protocol, using proximity information based upon HC. Similarly, minimizing the flooding overhead, ORRP direct the flooding for route discovery in two vertical directions. ROMER builds a runtime, forwarding mesh on a per-packet basis that offers a set of candidate routes. To improve resilience against lossy links or mesh failures, ROMER delivers redundant data copies in a controlled and randomized manner over the candidate forwarding mesh nodes. TDiCOR has also utilized opportunism in two different ways, i.e., it utilizes multiple sources trying to transmit the same signal to improve the signal robustness, as well as multiple forwarding nodes to improve packet progress toward the destination. An exception to this general and fuzzy pattern we sketched above is the MMESH algorithm we classified as traditional routing. Interestingly, MMESH does not utilize any opportunism to deliver the data. Instead, it tries to provide route reliability and quicker delivery, using more than one definitely chosen paths toward destination, distributing packets among them in a round robin fashion. It, however, introduces some randomness on which path to choose for each packet to overcome some unexpected packet loss and delay due to the wireless nature of the medium. Although the two classes of routing protocols seem to optimize performance in quite different fashions, there are some similarities between the two classes. Similar to the traditional routing which is trying to restrict the diameter of flooding, the opportunistic routing tries to restrict the diameter of opportunistic forwarding. For example, GPR, classified as an opportunistic approach functioning similar to SOAR, tries to limit the diameter of opportunistic forwarding using the location information of the mobile nodes.
6
Conclusions
In this survey, we aimed to provide an overview of different routing approaches for wireless networks in order to hopefully be useful for future research on this end. The specific focus of the survey was on routing for WMNs, even though some protocols are also applicable to general MANET structures. We noticed that there are two major class of routing mechanism for wireless networks: the traditional routing schemes, and the opportunistic routing ones. The main ideas utilized in the traditional approaches have been inspired by the wired networks’ routing schemes, where the links are deterministic, and using HC is justifiable as the dominant routing metric. Opportunistic routing, however, challenges the HC as a sufficiently descriptive metric for wireless networks. Instead, other link quality aware metrics, such as ETX, has been used. Some protocols presented have some characteristics of both classes, while for some of them, the routing proposal were approached from a totally different perspective – not easily classifiable as traditional or opportunistic. Classifying and comparing several recent routing proposals for WMNs, the final conclusion we draw is that while the opportunistic approaches seem to be more justified for wireless networking, research on traditional routing mechanisms is still on going. Depending on the particular properties of the wireless network on which the protocol is going to be implemented, both traditional and opportunistic routing mechanisms - as well as the other quite different approaches - have something to contribute, and may outperform each other.
References
1. Akyildiz, I.F., Wang, X., and Wang, W.: ‘Wireless mesh networks: a survey’, Computer networks, 2005, 47, (4), pp. 445-487 2. Jun, J., and Sichitiu, M.L.: ‘MRP: Wireless mesh networks routing protocol’, Computer Communications, 2008, 31, (7), pp. 1413-1435 3. Ko, Y.-B., and Vaidya, N.H.: ‘Location-aided routing (LAR) in mobile ad hoc networks’, in Editor (Ed.)^(Eds.): ‘Book Location-aided routing (LAR) in mobile ad hoc networks’ (ACM, 1998, edn.), pp. 66-75
4. Basagni, S., Chlamtac, I., Syrotiuk, V.R., and Woodward, B.A.: ‘A distance routing effect algorithm for mobility (DREAM)’, in Editor (Ed.)^(Eds.): ‘Book A distance routing effect algorithm for mobility (DREAM)’ (ACM, 1998, edn.), pp. 76-84 5. Haas, Z.J., Pearlman, M.R., and Samar, P.: ‘The zone routing protocol (ZRP) for ad hoc networks’, draft-ietf-manet-zone-zrp-04. txt, 2002 6. Jiang, Q., Finkel, R., Manivannan, D., and Singhal, M.: ‘RPSF: A routing protocol with selective forwarding for mobile ad-hoc networks’, Wireless Personal Communications, 2007, 43, (2), pp. 411-436 7. De Couto, D.S., Aguayo, D., Bicket, J., and Morris, R.: ‘A high-throughput path metric for multi-hop wireless routing’, Wireless Networks, 2005, 11, (4), pp. 419434 8. Biswas, S., and Morris, R.: ‘ExOR: opportunistic multi-hop routing for wireless networks’, in Editor (Ed.)^(Eds.): ‘Book ExOR: opportunistic multi-hop routing for wireless networks’ (ACM, 2005, edn.), pp. 133-144 9. Perkins, C.E., and Royer, E.M.: ‘Ad-hoc on-demand distance vector routing’, in Editor (Ed.)^(Eds.): ‘Book Ad-hoc on-demand distance vector routing’ (IEEE, 1999, edn.), pp. 90-100 10. Johnson, D.B., and Maltz, D.A.: ‘Dynamic source routing in ad hoc wireless networks’, Kluwer International Series in Engineering and Computer Science, 1996, pp. 153-179 11. Kum, D.-W., Le, A.-N., Cho, Y.-Z., Toh, C.K., and Lee, I.-S.: ‘An efficient ondemand routing approach with directional flooding for wireless mesh networks’, Communications and Networks, Journal of, 2010, 12, (1), pp. 67-73 12. Nandiraju, N.S., Nandiraju, D.S., and Agrawal, D.P.: ‘Multipath routing in wireless mesh networks’, in Editor (Ed.)^(Eds.): ‘Book Multipath routing in wireless mesh networks’ (IEEE, 2006, edn.), pp. 741-746 13. Cheng, B.-N., Yuksel, M., and Kalyanaraman, S.: ‘Orthogonal rendezvous routing protocol for wireless mesh networks’, IEEE/ACM Transactions on Networking (ToN), 2009, 17, (2), pp. 542-555 14. Rozner, E., Seshadri, J., Mehta, Y., and Qiu, L.: ‘SOAR: Simple opportunistic adaptive routing protocol for wireless mesh networks’, Mobile Computing, IEEE Transactions on, 2009, 8, (12), pp. 1622-1635 15. Xiao, N., Ding, L., Li, M., and Wu, M.: ‘Geographic probabilistic routing protocol for wireless mesh network’: ‘Advanced Parallel Processing Technologies’ (Springer, 2007), pp. 477-486 16. Yuan, Y., Yang, H., Wong, S.H., Lu, S., and Arbaugh, W.: ‘ROMER: resilient opportunistic mesh routing for wireless mesh networks’, in Editor (Ed.)^(Eds.): ‘Book ROMER: resilient opportunistic mesh routing for wireless mesh networks’ (2005, edn.), pp. 17. Kurth, M., Zubow, A., and Redlich, J.-P.: ‘Cooperative opportunistic routing using transmit diversity in wireless mesh networks’, in Editor (Ed.)^(Eds.): ‘Book Cooperative opportunistic routing using transmit diversity in wireless mesh networks’ (IEEE, 2008, edn.), pp. 1310-1318 18. Eriksson, J., Faloutsos, M., and Krishnamurthy, S.V.: ‘DART: dynamic address routing for scalable ad hoc and mesh networks’, Networking, IEEE/ACM Transactions on, 2007, 15, (1), pp. 119-132
19. Shakya, S., and Pokhrel, S.R.: ‘Global optimization of field based routing in wireless mesh network (GOFBR-WMN)’, in Editor (Ed.)^(Eds.): ‘Book Global optimization of field based routing in wireless mesh network (GOFBR-WMN)’ (IEEE, 2012, edn.), pp. 1-5 20. Wellons, J., and Xue, Y.: ‘Oblivious routing for wireless mesh networks’, in Editor (Ed.)^(Eds.): ‘Book Oblivious routing for wireless mesh networks’ (IEEE, 2008, edn.), pp. 2969-2973 21. Fu, Z., Zerfos, P., Luo, H., Lu, S., Zhang, L., and Gerla, M.: ‘The impact of multihop wireless channel on TCP throughput and loss’, in Editor (Ed.)^(Eds.): ‘Book The impact of multihop wireless channel on TCP throughput and loss’ (IEEE, 2003, edn.), pp. 1744-1753 22. Rangwala, S., Gummadi, R., Govindan, R., and Psounis, K.: ‘Interference-aware fair rate control in wireless sensor networks’, in Editor (Ed.)^(Eds.): ‘Book Interference-aware fair rate control in wireless sensor networks’ (ACM, 2006, edn.), pp. 63-74 23. Li, Y., Qiu, L., Zhang, Y., Mahajan, R., and Rozner, E.: ‘Predictable performance optimization for wireless networks’, in Editor (Ed.)^(Eds.): ‘Book Predictable performance optimization for wireless networks’ (ACM, 2008, edn.), pp. 413-426 24. Liu, C.H., Leung, K.K., and Gkelias, A.: ‘A novel cross-layer QoS routing algorithm for wireless mesh networks’, in Editor (Ed.)^(Eds.): ‘Book A novel crosslayer QoS routing algorithm for wireless mesh networks’ (IEEE, 2008, edn.), pp. 1-5
