Advanced Wireless Networks - 4G Technologies

488 AD HOC NETWORKS

S

C

A

B

Figure 13.22 Loop-back termination.

Later

Qquery

Earlier

Qquery

S

A

B

C

Figure 13.23 Early termination.

13.2.2 Early termination

When the ability to terminate route query threads is limited to peripheral nodes, threads are allowed to penetrate into previously covered areas, which generates unnecessary control traffic. This excess traffic can be eliminated by extending the thread termination capability to the intermediate nodes that relay the thread. This approach is referred to as early termination (ET). Figure 13.23 illustrates the operation of the ET mechanism. Node S broadcasts a route query with node C as one of the intended recipients. Intermediate node A passes along the query to B. Instead of delivering the query to node C, node B terminates the thread because a different thread of this query was previously detected. Intermediate nodes may terminate existing queries but are restricted from issuing new queries. Otherwise, the ZRP would degenerate into a flooding protocol. The ability to terminate an overlapping query thread depends on the ability of nodes to detect that a routing zone they belong to has been previously queried. Clearly, the central node in the routing zone (which processed the query) is aware that its zone has been queried. In order to notify the remaining routing zone nodes without introducing additional control traffic, some form of ‘eavesdropping’ needs to be implemented. The first level of query detection (QD1) allows the intermediate nodes, which transport queries to the edge of the routing zone, to detect these queries. In single channel networks, it may be possible for queries to be detected by any node within the

Figure 13.24 Query detection (QD1/QD2).

range of a query-transmitting node. This extended query detection capability (QD2) can be implemented by using IP broadcasts to send route queries. Figure 13.24 illustrates both levels of advanced query detection. In this example, node S broadcasts to two peripheral nodes, B and D. The intermediate nodes A and C are able to detect passing threads using QD1. If QD2 is implemented, node E will be able to ‘eavesdrop’ on A’s transmissions and record the query as well.

The techniques just discussed improve the efficiency of the IERP by significantly reducing the cost of propagating a single query. Further improvements in IERP performance can be achieved by reducing the frequency of route queries, initiating a global route discovery procedure only when there is a substantial change in the network topology. More specifically, active routes are cached by nodes: the communicating end nodes and intermediate nodes. Upon a change in the network topology, such that a link within an active path is broken, a local path repair procedure is initiated. The path repair procedure substitutes a broken link by a minipath between the ends of the broken link. A path update is then generated and sent to the end points of the path. Path repair procedures tend to reduce the path optimality (e.g. increase the length for shortest path routing). Thus, after some number of repairs, the path endpoints may initiate a new route discovery procedure to replace the path with a new optimal one.

13.2.3 Selective broadcasting (SBC)

Rather than broadcast queries to all peripheral nodes, the same coverage can be provided by broadcasting to a chosen subset of peripheral nodes. This requires IARP to provide network topology information for an extended zone that is twice the radius of the routing zone.

A node will first determine the subset of other peripheral nodes covered by its assigned inner peripheral nodes. The node will then broadcast to this subset of assigned inner peripheral nodes which forms the minimal partitioning set of the outer peripheral nodes.

This is illustrated in Figure 13.25. S’s inner peripheral nodes are A, B and C. Its outer peripheral nodes are F, G, H, X, Y and Z. Two inner peripheral nodes of B (H and X) are also inner peripheral nodes of A and C. S can then choose to eliminate B from its list of broadcast recipients since A can provide coverage to H and C can cover X.

The position of the routing functions in the protocol stack are illustrated in Figure 13.26. Route updates are triggered by the MAC-level neighbor discovery protocol (NDP). IARP is notified when a link to a neighbor is established or broken. IERP reactively acquires routes

490 AD HOC NETWORKS

X C

B S

H

A

G

Figure 13.25 Selective broadcasting.

Figure 13.26 The position of the routing functions in the protocol stack.

Table 13.1 Simulation parameters

to nodes beyond the routing zone. IERP forwards queries to its peripheral nodes (BRP) keeping track of the peripheral nodes through the routing topology information provided by IARP.

Pearlman and Haas [38] present the performance evaluation of the hybrid protocol, described above for the simulation set of parameters given in Table 13.1. From Figure 13.27 one can see that the IARP control traffic per node is increased as the radius of the zone is increased.

At the same time IERP traffic generated per zone would be reduced. So, the total ZRP traffic has a minimum value for some zone diameter r, which is demonstrated in Figure 13.28.

Zone radius

Figure 13.27 IARP traffic generated per neighbor.

Zone radius

Figure 13.28 ZRP traffic per node (N = 1000 nodes, v = 0:5 neighbors/s).

13.3 SCALABLE ROUTING STRATEGIES

13.3.1 Hierarchical routing protocols

A hierarchical approach to routing, often referred to as hierarchical state routing (HSR), has been a traditional option when the network has a large number of nodes. The approach has a lot in common with routing with aggregation presented in Chapter 7. Common

492 AD HOC NETWORKS

Figure 13.29 The network hierarchical structure.

Figure 13.30 Three layers network hierarchical structure.

table-driven protocols and on-demand protocols are for flat topologies and thus have a scalability problem when the network is large. For table-driven protocols there is high volume of overhead transmissions. On the other hand, for on-demand protocols there is large discovery latency.

The experience gained in wired networks suggests the use of a hierarchical structure to address the scalability problem. The use of hierarchical routing protocol in ad-hoc networks reduces overhead traffic and discovery latency but it has drawbacks, such as: suboptimal routes and complex management of the network hierarchical structure due to its dynamic nature. The basic idea is to divide the network into clusters or domains, as illustrated in Figures 13.29 and 13.30.

The mobile nodes are grouped into regions, regions are grouped into super-regions, and so on, as shown in Figure 13.30. A specific mobile host is chosen as the clusterhead for each region. In hierarchical routing, mobile nodes know how to route packets to their destination

494 AD HOC NETWORKS

Chapter 7. The virtual link can be viewed as a ‘tunnel’ implemented through lower level nodes.

Applying the aforementioned clustering procedure (aggregation) recursively, new cluster heads are elected at each level and become members of the higher level cluster (e.g. node 1 is elected as a cluster head at level 1 and becomes a member of level 2 cluster C2–1).

Nodes within a cluster exchange virtual LS information as well as summarized lowerlevel cluster information. After obtaining the LS information at this level, each virtual node floods it down to nodes within the lower level cluster. As a result, each physical node has a ‘hierarchical’ topology information, as opposed to a full topology view as in flat LS schemes. The hierarchy so developed requires a new address for each node, the hierarchical address. There are many possible solutions for the choice of the hierarchical address scheme. In hierarchical state routing (HSR), we define the hierarchical ID (HID) of a node as the sequence of the MAC addresses of the nodes on path from the top hierarchy to the node itself. For example, in Figure 13.31 the hierarchical address of node 6 [called HID(6)], is 3, 2, 6. In this example, node 3 is a member of the top hierarchical cluster (level 2). It is also the cluster head of C1–3. Node 2 is member of C1–3 and is the cluster head of C0–2. Node 6 is a member of C0–2 and can be reached directly from node 2. The advantage of this hierarchical address scheme is that each node can dynamically and locally update its own HID upon receiving the routing updates from the nodes higher up in the hierarchy. The hierarchical address is sufficient to deliver a packet to its destination from anywhere in the network using HSR tables.

Referring to Figure 13.31, consider for example the delivery of a packet from node 5 to node 10. Note that HID(5) =<1,1,5> and HID(10) =<3,3,10>. The packet is forwarded upwards (to node 1) to the top hierarchy by node 5. Node 1 delivers the packet to node 3, which is the top hierarchy node for destination 10. Node 1 has a ‘virtual link’, i.e. a tunnel, to node 3, namely, the path (1, 6, 2, 8, 3). It thus delivers the packet to node 3 along this path. Finally, node 3 delivers the packet to node 10 along the downwards hierarchical path, which in this case reduces to a single hop.

Gateways nodes can communicate with multiple cluster heads and thus can be reached from the top hierarchy via multiple paths. Consequently, a gateway has multiple hierarchical addresses, similar to a router in the wired Internet (see Chapter 1), equipped with multiple subnet addresses.

13.3.2 Performance examples

Performance analysis of the system described above can be found in Iwata et al. [41]. In most experiments, the network consistes of 100 mobile hosts roaming randomly in all directions at a predefined average speed in a 1000 × 1000 m area(i.e. no group mobility models are used). A reflecting boundary is assumed. The radio transmission range is 120 m. A free space propagation channel model is assumed. The data rate is 2 Mb/s. The packet length is 10 kb for data, 2 kb for a cluster head neighbor list broadcast, and 500 b for MAC control packets. Transmission time is 5 ms for a data packet, 1 ms for a neighboring list, and 0.25 ms for a control packet. The buffer size at each node is 15 packets. Figures 13.32 and 13.33 illustrate the tradeoffs between throughput and control overhead (O/H) in HSR when the route refresh rate is varied.

496 AD HOC NETWORKS

13.3.3 FSR (fisheye routing) protocol

This protocol represents a different way to reduce (aggregate) the amount of information used for routing purposes. In Kleinrock and Stevens [42], a ‘fisheye’ technique was proposed to reduce the size of information required to represent graphical data. The eye of a fish captures with high detail the pixels near the focal point. The detail decreases as the distance from the focal point increases. In routing, the fisheye approach translates to maintaining accurate distance and path quality information about the immediate neighborhood of a node, with progressively less detail as the distance increases.

The FSR scheme presented in Iwata et al. [41] is built on top of another routing scheme called ‘global state routing’ (GSR) [43]. GSR is functionally similar to LS routing in that it maintains a topology map at each node. The key is the way in which routing information is disseminated. In LS, LS packets are generated and flooded into the network whenever a node detects a topology change. In GSR, LS packets are not flooded. Instead, nodes maintain an LS table based on the up-to-date information received from neighboring nodes and periodically exchange it with their local neighbors only (no flooding).

Through this exchange process, the table entries with larger sequence numbers replace the ones with smaller sequence numbers. The GSR periodic table exchange resembles the DSDV, discussed earlier in this chapter, where the distances are updated according to the time stamp or sequence number assigned by the node originating the update. In GSR (like in LS), LSs are propagated, a full topology map is kept at each node, and shortest paths are computed using this map.

In a wireless environment, a radio link between mobile nodes may experience frequent disconnects and reconnects. The LS protocol releases an LS update for each such change, which floods the network and causes excessive overhead. GSR avoids this problem by using periodic exchange of the entire topology map, greatly reducing the control message overhead [43]. The drawbacks of GSR are the large size update message that consumes a considerable amount of bandwidth and the latency of the LS change propagation, which depends on the update period. This is where the fisheye technique comes to help, by reducing the size of update messages without seriously affecting routing accuracy.

Figure 13.34 illustrates the application of fisheye in a mobile wireless network. The circles with different shades of gray define the fisheye scopes with respect to the centre node (node 11). The scope is defined as the set of nodes that can be reached within a given number of hops. In our case, three scopes are shown for one, two and three hops, respectively. Nodes are color-coded as black, gray and white, accordingly. The reduction of update message size is obtained by using different exchange periods for different entries in the table. More precisely, entries corresponding to nodes within the smaller scope are propagated to the neighbors with the highest frequency. Referring to Figure 13.35, entries in bold are exchanged most frequently. The rest of the entries are sent out at a lower frequency. As a result, a considerable fraction of LS entries are suppressed, thus reducing the message size. This strategy produces timely updates from near stations, but creates large latencies from stations that are far away. However, the imprecise knowledge of the best path to a distant destination is compensated for by the fact that the route becomes progressively more accurate as the packet gets closer to its destination.

In summary, FSR scales well to large networks, by keeping LS exchange overhead (O/H) low without compromising route computation accuracy when the destination is near. By retaining a routing entry for each destination, FSR avoids the extra work of ‘finding’ the