23-05-2012, 03:08 PM
Zone Routing Protocol (ZRP) ANALSIS
ZRP ANALSIS.docx (Size: 112.38 KB / Downloads: 30)
Overview
This document describes and analyzes the Zone Routing Protocol or ZRP for Mobile Ad-Hoc Networks (MANETs). The basics of MANET and the implications on routing in particular are briefly covered in Section 2 to provide an introduction to the problems resulting from a rapidly changing topology without a fixed router.
As we will see in Section 3, ZRP, in contrast to other MANET routing protocols, utilizes a hybrid pro-active/re-active approach to maintain valid routing tables without too much overhead. Furthermore, ZRP does not provide a single protocol, but rather outlines a routing framework suitable for inclusion and extension of other existing protocols.
In describing the protocol, we will consider some specific examples in order to visualize how ZRP's characteristics influence it's performance (Section 4).
Section 5analyzes the performance of the ZRP as well as discusses various scenarios, and in Section 6 the information provided in this document is briefly summarized.
MANET in general
A Mobile Ad-Hoc Network (MANET) is a decentralized network of autonomous mobile nodes able to communicate with each other over wireless links. Due to the mobility of the nodes, the topology of the network may rapidly be changing, making it impossible to use conventional routing tables maintained at fixed points (routers). Instead, each node is required to determine the best route to a given destination node by itself.
Given their dynamic nature, route discovery in a MANET differs significantly from the more or less static routes in wired networks: Not all nodes in a MANET necessarily have the same capabilities. Two nodes, even if they are direct neighbours, may differ with respect to signal strength, available power, reliability etc.
These differences require much more complicated and particularly more active distributed algorithms in order to maintain an accurate picture of the networks topology, while at the same time providing scalability for potentially large (and ever-growing) networks. At the same time, route discovery must not use up the majority of the often limited bandwidth available to today’s mobile devices.
Furthermore, it is important to point out an important difference to conventional routing approaches: In wired networks, each link is bi-directional. If a node A can send packets to a node B, we know that node B can send packets back to node A, and a reverse path can be entered. This is not necessarily the case in a wireless network, where the physical location and the individual power resources have great influence upon a nodes transmission capacity and signal strength.
MANET routing protocols are IP based and may use unicast, multicast or hybrid approaches and should allow for interaction with standard wired IP services rather than being regarded as a completely separate entity.
A detailed yet not overly complex overview of the various aspects of Mobile Ad-Hoc Networking is given in [7].
Pro-active vs. reactive
The IETF MANET Working Group has researched and developed a number of protocols for mobile ad-hoc networks, which have been described in [8], [9], [10], [11] and [12]. These protocols can generally be categorized into two groups: pro-active and re-active protocols.
Pro-active protocols follow an approach similar to the one used in wired routing protocols. By continuously evaluating the known and attempting to discover new routes, they try to maintain the most up-to-date map of the network. This allows them to efficiently forward packets, as the route is known at the time when the packet arrives at the node.
Pro-active or table-driven protocols, in order to maintain the constantly changing network graph due to new, moving or failing nodes, require continuous updates, which may consume large amounts of bandwidth - clearly a disadvantage in the wireless world, where bandwidth is often sparse. Even worse so, much of the accumulated routing information is never used, since routes may exist only for very limited periods of time.
The family of Distance-Vector protocols, including Destination-Sequenced Distance-Vector Routing ([13]), fall into the category of pro-active protocols.
In contrast, reactive protocols determine the proper route only when required, that is, when a packet needs to be forwarded. In this instance, the node floods the network with a route-request and builds the route on demand from the responses it receives. This technique does not require constant broadcasts and discovery, but on the other hand causes delays since the routes are not already available. Additionally, the flooding of the network may lead to additional control traffic, again putting strain on the limited bandwidth.
These reactive (or on-demand) protocols include Dynamic Source Routing (DSR) [9] and Ad-hoc On demand Distance Vector Routing (AODV) [8], as well as the classical flooding algorithms. [1]
Introduction to ZRP
As explained above, both a purely pro-active and purely reactive approaches to implement a routing protocol for a MANET have their disadvantages. The Zone Routing Protocol, or ZRP, as described in this document combines the advantages of both into a hybrid scheme, taking advantage of pro-active discovery within a node's local neighbourhood, and using a reactive protocol for communication between these neighbourhoods.
In a MANET, it can safely be assumed that the most communication takes place between nodes close to each other. Changes in the topology are most important in the vicinity of a node - the addition or the removal of a node on the other side of the network has only limited impact on the local neighbourhoods.
As mentioned earlier, the ZRP is not so much a distinct protocol as it provides a framework for other protocols. The separation of a nodes local neighbourhood from the global topology of the entire network allows for applying different approaches - and thus taking advantage of each technique's features for a given situation. These local neighbourhoods are called zones (hence the name); each node may be within multiple overlapping zones, and each zone may be of a different size. The ``size'' of a zone is not determined by geographical measurement, as one might expect, but is given by a radius of length , where is the number of hops to the perimeter of the zone.
By dividing the network into overlapping, variable-size zones, ZRP avoid a hierarchical map of the network and the overhead involved in maintaining this map. Instead, the network may be regarded as flat, and route optimization is possible if overlapping zones are detected.
While the idea of zones often seems to imply similarities with cellular phone services, it is important to point out that each node has its own zone, and does not rely on fixed nodes (which would be impossible in MANETs). Figure 1 shows an example routing zone with .
Figure 1: Routing Zone of node A with .
Note that in this example node A has multiple routes to node F, including one that has a hop countof . Since it also has a route with , F still belongs to A's zone. Node G is out of A's zone,
The nodes on the perimeter of the zone (i.e. with a hop count ) are referred to as peripheral nodes (marked gray), nodes with are interior nodes.
Obviously a node needs to first know about itsneighbours before it can construct a routing zone and determine its peripheral nodes. In order to learn about its direct neighbours, a node may use the media access control (MAC) protocols directly. Alternatively, it may require a Neighbour Discovery Protocol (NDP). Again, we see that ZRP, as a framework, does not strictly specify the protocol used but allows for local independent implementations.
Such a Neighbour Discovery Protocol typically relies on the transmission of ``hello'' beacons by each node. If a node receives a response to such a message, it may note that it has a direct point-to-point connection with this neighbour. The NDP is free to select nodes on various criteria, such as signal strength or frequency/delay of beacons etc. Once the local routing information has been collected, the node periodically broadcasts discovery messages in order to keep its map of neighbours up to date. In doing so, it is assumed that these ``link-layer (neighbour) unicasts are delivered reliably and in-sequence.''[1]
If the MAC layer of the nodes does not allow for such a NDP, the Intrazone Routing Protocol must provide the possibility of direct neighbour discovery. This protocol is responsible for determining the routes to the peripheral nodes and is commonly a pro-active protocol. The Intrazone Routing Protocol, or IARP, is described in more detail in in Section 3.1.
Communication between the different zones is guarded by the Interzone Routing Protocol, or IERP, and provides routing capabilities among peripheral nodes only. That is, if a node encounters a packet with a destination outside its own zone - i.e. it does not have a valid route for this packet - it forwards it to its peripheral nodes, which maintain routing information for the neighbouring zones, so that they can make a decision of where to forward the packet to. Through the use of a bordercast algorithm rather than flooding all peripheral nodes, these queries become more efficient. The Interzone Routing Protocol and the Bordercast Resolution Protocol are presented in Sections 3.2 and 3.3.
Figure 2: ZRP components
As we can see, the Zone Routing Protocol consists of several components, which only together provide the full routing benefit to ZRP. Each component works independently of the other and they may use different technologies in order to maximize efficiency in their particular area. For example, a reactive protocol such as AODV might be used as the IARP, while the IERP is most commonly a pro-active protocol such as OLSR [14].
Figure 2, as adapted from [1], illustrates the different protocols and their interactions.
Even though the hybrid nature of the ZRP seems to indicate that it is a hierarchical protocol, it is important to point out that the ZRP is in fact a flat protocol. In hierarchical network architecture, two different protocols are maintained for communication among (a) each individual cluster's nodes and (b) the different clusters. The main difference here is that in the ZRP there is a one-to-one correspondence between nodes and routing zones, causing overlapping zones maintained by each individual nodes (see [1] for details).