An Efficient Routing Protocol for Wireless Ad hoc Networks - CiteSeerX

4 downloads 233 Views 244KB Size Report
An Efficient Routing Protocol for Wireless Ad hoc. Networks. A. N. Al-Khwildi, K. K. Loo and H. S. Al-Raweshidy. Wireless Network and Communications Group ...
SETIT 2007

4th International Conference: Sciences of Electronic, Technologies of Information and Telecommunications March 25-29, 2007 – TUNISIA

An Efficient Routing Protocol for Wireless Ad hoc Networks A. N. Al-Khwildi, K. K. Loo and H. S. Al-Raweshidy Wireless Network and Communications Group (WNCG), School of Engineering & Design, Brunel University, West London, UB8 3PH [email protected] Jonathan.Loo @brunel.ac.uk Abstract: A new routing protocol which increases the effectiveness of the routing protocol within a mobile ad-hoc network is proposed. The proposed on-demand link weight (ODLW) routing protocol is targeted for real-time multimedia applications where QoS parameters such as bandwidth, delay and node lifetime are considered. The ODLW is essentially a succession of on-demand and link-state routing protocols. It was shown that the unique route discovery mechanism of the ODLW outperformed Dynamic Source Routing protocol (DSR) and Optimized Link State Routing Protocol (OLSR) in reducing the end-to-end delay by 80% and 40% respectively as well as the packet drop by 30% and 20% respectively. The simulation is based on the application scenario model implemented on Opnetworks 11.0. Key words: MANET, QoS, Routing Protocols. addition to being a source or destination node[1][2][3].

1. Introduction

The purpose of this paper is to examine the route discovery procedure of various ad hoc wireless routing protocols. DSR [4], OLSR [5] and newly developed ad hoc protocol, ODLW [6] are selected for this purpose. The paper is organized as follows: Section I presents a thorough discussion on ad hoc protocol categories, Proactive, Reactive and Hybrid. Section II discusses Reactive DSR and ODLW, and Proactive (OLSR) protocol in detail. Section III presents the comparison between ODLW, DSR and OLSR. The simulation model and simulation results present in sections IV and V respectively. Finally the conclusion will present in section VI.

Wireless industry has seen exponential growth in last couple of years. The advancement in growing availability of wireless networks and emergence of handheld computer, PDAs and Cell phones is now playing very important role in our daily routines. Surfing internet from railway station, airport, cafes, public locations, internet browsing on cell phones, and information or file exchange between devices without wired connectivity are just few examples. All this ease is the result of mobility of wireless devices while being connected to a gateway to access the internet or information from fixed or wired infrastructure (called Infrastructure based wireless network) or ability to develop an on demand, selforganizing wireless network without relying on any available fixed infrastructure (called Ad hoc networks). Typical example of first type of network is office wireless networks (WLANs) where wireless access point services to all wireless devices within the radius. Example of MANET can be describe as a group of soldiers in a war zone, wirelessly connected to each other with the help of limited battery power devices and efficient ad hoc routing protocol that helps them to maintain quality of communication, while they are changing their position rapidly. Therefore routing in ad hoc wireless networks plays an important role for data forwarding, where each mobile node can act as a relay in

2. Ad hoc Routing Protocol Classification In general, ad hoc routing protocols are categorized into three categories: proactive (table-driven) protocols, reactive (on-demand) protocols and hybrid protocols. This classification differentiates the routing protocols according to their technique, hop count, link state and QoS in route discovery. In protocols based on hop count technique, each node contains next hop information in its routing table, to the destination. While link state routing protocols keep a routing table for complete topology, which is built up by finding shortest path of link costs. QoS routing is the process of selecting the path to be used by the packets of a flow, based on its QoS -1-

SETIT2007

2.3. Hybrid Routing Protocols

requirements eg bandwidth, delay etc. The proposed ODLW protocol is on-demand and selects the best path based on link weight parameters (Bandwidth, Delay, Node life time). Figure 1 show some ad hoc routing protocols sorted according to this classification [7][8].

Hybrid protocols inherit the advantage of high-speed routing form proactive and less overhead control messages from reactive protocols. The characteristics of proactive and reactive routing protocols can be integrated to achieve hybrid routing technique. Hybrid routing protocols may exhibit proactive or reactive behavior depending on the circumstance, hence allow flexibility based on the wireless network. The most typical protocols are ZRP and ZHLS [15],[16].

AD-HOCMOBILEROUTING PROTOLOCS

Hybrid

On-Demand

HopCount Link-State

AODV DSR TORA GBRP

SOAR OLIVE

QoS Routing

AQOR MP-DSR

TableDriven

QoS Routing

ZRP ZHLS

CEDAR DQRA

The most typical hybrid one is zone routing protocol. As to the major division of routing protocols, Table I gives a comparison of Proactive, Reactive and Hybrid routing protocols.

Link-State HopCount

OLSR DSDV

TABLE I CHARACTERISTIC COMPARISON OF PROACTIVE, REACTIVE AND HYBRID ROUTING PROTOCOL

WRP CGSR

Proactive

Reactive

Hybrid

Flat and hierarchical structure

Mostly Flat

Hierarchical

Topology dissemination

Periodical

On-Demand

Both

Route Latency

Mobility Handling

Periodical updates

Available when needed Up to few hundred nodes, depend on (traffic level, number of hops) Route maintenance

Both

Scalability

Always available Up to hundred nodes

High

Low

ODLW

Routing Structure

Figure 1. Categorization of ad hoc routing protocol 2.1. Proactive Routing Protocol This type of routing protocols is very familiar in fixed wired networks. In this approach, each ad hoc node consists of a topology table, which contains the up to date networks nodes interaction information. This table is updated all the time and it gives the proactive protocols another name of table-driven. One or more routing tables are maintained at each node and are exchanged periodically to share the topology information with the neighboring nodes in order to maintain a consistent network view. Ad hoc network based on proactive protocols, power and bandwidth consumption increased due to topology table exchange among nodes after each changing in nodes location. This takes place even if the network is in stand-by mode (e.g. no data transmissions in the network). The best network context for proactive protocols is the low (or no) mobility networks. The most accepted proactive protocols are DSDV, OLSR, CGSR and WRP [9] [5] [10] [11].

Communicati on overhead

Designed for up to 1000 or more nodes

Both Medium

3. A description of the reactive & proactive routing protocols In this section, we discuss about DSR, OLSR and proposed ODLW protocol in detail like route discovery mechanism, route reply and maintenance.

2.2. Reactive Routing Protocol Reactive routing techniques, also called on-demand routing, take different approach for routing than proactive protocols. Routes to the destination are discovered only when actually needed. When source node needs to send packet to some destination, it checks it routing table to determine whether it has a route. If no route exists, source node performs route discovery procedure to find a path to the destination. Reactive routing protocols can dramatically reduce routing overhead because they do not need to search for and maintain the routes on which there is no data traffic. This property is very appealing in the resource-limited environment. The most accepted reactive protocols are DSR, AODV,TORA and GBRF [4] [12] [13] [14].

3.1. Dynamic Source Routing Protocol DSR is a reactive flat protocol. The main difference between DSR and all other reactive protocols is that it is based on source routing scheme. In source routing, the source node specifies the intermediate nodes sequence. In a DSR protocol, mobile nodes are required to maintain route caches that contain the source routes to all mobile nodes that it is aware. The entries in the route cache are updated as new routes are learned. When a node requires sending data to a destination node, it first refers its routing cache to determine if it has a route to the destination. If an unexpired route exists, it will use the route to send the packet. The data packet carries the

-2-

SETIT2007

source route in the packet header. There are two major phases in this protocol. The first is Routing Discovery, which is achieved by flooding the network with Route Request packets as shown in Figure 2(a). The destination node, on receiving a (RREQ) responds by sending Route Reply (RREP) packet back to the source, which carries the route traversed by (RREQ) packet. Route to the destination at the cache will be unicast to the specific nodes as showing in Figure 2(b) and 2(c). (d) Path of the Route Reply through piggyback on a route request Figure 2. DSR Routing Mechanism

If a route to the destination does not exist, route discovery is initiated by broadcasting a route request. Within the route request, contains the address of the destination along with sources node’s address and a unique identification number. All nodes receiving this route request will check whether it knows of a route to the destination. If it does not have a route, it will add its own address in to the route record of the packet and further forward the packet along its outgoing links. To limit the number of route requests propagating on the outgoing link of a node, the node would only forward a route request only if the request is yet to be seen by the mobile and only if the mobile address does not already appear in the route record [17] as shown in Figure 2(d).

3.2. Optimized Link State Routing OLSR is a link-state and proactive based protocol. Unlike distance victor protocols, link-state protocols do not relay on number of hops to the destination node. Instead, link-state algorithms determined the best route according to the link load, delay, bandwidth etc. Although calculating the best available route by this approach is more complicated than hop count, it is approved that link-state routes are more accurate and stable . Control overheads information is compact and retransmission number to flood these control messages is reduced comparing to pure link-state protocols. The perfect network context for OLSR is low mobile and dense ad hoc networks. OLSR overhead control signals do not require for a reliable transmission link, which is very suitable for wireless networks. OLSR supports node mobility as far as it is traceable by its neighbors. Overhead control signals are periodically broadcasted by each node in the network. The period between each signal is determined according to the nodes expected speed. Each node N in the network, selects a set from the next hop neighbors called multipoint relay nodes. Multipoint relay nodes set covers all nodes two hops away from N. Figure 3 illustrates an example of an ad hoc network based upon OLSR protocol. The concept of multipoint relay set is to reduce the overhead control messages and support the optimization. Only multipoint relay set nodes can retransmit node N broadcast control messages. Other one-hop neighbors receive messages and update their information accordingly but do not retransmit them [18]. The route is built using the routing table, which is based on the topology table and the neighbor list saved in each node. The topology table is created by periodically broadcasting control messages called Topology Control, which contain each node multipoint relay set. This makes the multipoint relay set for each node available for all other nodes in the network. The concept of having multipoint relay set in OLSR is to avoid sending the same overhead control message twice to the same node. This will optimise the network bandwidth and power consumption. The Route discovery procedure is described following:

(a) Route Request Propagation

(b) Route Reply with reference of destination route cache.

(c) Path of the Route Reply through route record.



-3-

Every node broadcasts HELLO messages that contain one-hop neighbor information periodically.

SETIT2007

The receiving node will only forward the RREQM message from the sender where its route meets the QoS requirements. At the receiving end, the destination node evaluates all RREQM messages which contain QoS information from its neighboring senders and return a RREPM message to the sender where its route meets the required bandwidth, low accumulated delay and long route lifetime. This RREPM message is unicast back to the source following the route recorded in the route list. If one or more routes met the required bandwidth, a route with low accumulated and longer route lifetime will be chosen [19][20].

The TTL of HELLO message is 1, so they are not forwarded by its neighbors. With the aid of HELLO messages, every node obtains local topology information; •

A node (also called selector) chooses a subset of its neighbors to act as multi-point relaying nodes for it based on the local topology information, which are specified in the periodic HELLO messages later. MPR nodes have two roles: o

When the selector sends or forwards a broadcast packet, only its MPR nodes among all its neighbors forward the packet;

o

The MPR nodes periodically broadcast its selector list throughout the MANET (again, by means of MPR flooding). Thus every node in the network knows which MPR nodes could reach every other node. Note that a) reduces the number of retransmissions of topology information broadcast, and b) reduces the size of broadcast packet. As result, much more bandwidth is saved compared with original link state routing protocols.

o

With global topology information stored and updated at every node, a shortest path from one node to very other node could be computed node to very other node could be computed with Dijkstra’s algorithm, which goes along a series of MPR node.

(a)

Referring to Figure 4 assume a source node is “A”, destination node is “J” and the required data bandwidth size is 2Mbps. Node “A” broadcast a RREQM, which is received by three relay nodes, “I”, “E” and “B”. RREQM contains link requirements Bandwidth, Delay, Node Life time (NL) for node A. In this example, relay nodes I, E and B will be broadcasting RREQM containing route information in the form of table as : {[A,5,1,3], [B,2,2,2], [A,5,1,2]} respectively to the neighbors. The table will be extracted from RREQM, for example [A,5,1,3] where A refers to the originating nodes of RREQM, 5 represents the total of bandwidth available through this path, 1 represents the delay in transmitting the data from node to its neighbors and 3 represents the minimum node life time. Also the nodes will generate a table containing a node-ID and link weight parameters for the node that has sent the RREQM. This process is repeated and RREQM fields will be updated from node to node until the destination is reached. After the destination node receives the RREQM’s, it would unicast back a RREPM by following a reverse methodology path from destination to source node. A destination has list of the qualified routes through nodes H, F and C. In this case the destination chooses the best optimum path which meets the requirements. The best optimum path will be through node H, although it has longer route, but it meets the required bandwidth in addition of that it has a better node life time and better delay.

(b)

The other path through node C is one of the other available paths to reach the source node A but the node life time is on level 5 which shows node will last less then 2 minutes as can be seen from node life time table in Figure 4. The node life time is very important because if the node runs out of battery source node need to rebroadcast again to find the path to the destination and will affect all the routing process. Node H receives the RREPM sent by destination, it checks the request-ID to search for the corresponding table ID, and then update the fields of the RREPM and unicast again. This process is repeated and RREPM fields will be updated from node to node until the original source is reached. The best optimum route from destination J to source A is (J-H-GD-B-A) as it meets best requirements compare to other available routes. In case of failure of this primary route, a secondary route via node F will be chosen. In overall, the

Figure 3. OLSR Routing Mechanism 3.3. On-Demand Link Weight Protocol The ODLW routing protocol finds the best route with QoS assurance by using a route request (RREQM) message and a route reply (RREPM) message as shown in Figure 4. The RREQM message records the QoS parameters such as required bandwidth, minimum available route bandwidth, accumulated route delay and minimum route lifetime, from node to node, so that a QoS route list can be constructed at each node. When a source node wants to send data to a destination node, it broadcast a RREQM message to its neighbors. Receiving node updates the RREQM message fields and compares among all other received RREQM messages from the neighboring senders.

-4-

SETIT2007

(QoS) support in terms of bandwidth and bounded endto-end delay. It is more difficult to provide QoS support in MANET than in wired wide area networks due to MANET's dynamic topology, time-varying link capacity, unpredictable channel condition and shared wireless medium. Existing routing protocols in MANET, such as DSR is designed primarily to carry best-effort traffic, while OLSR is designed for best effort flooding, to provide connectivity between the nodes.

route discovery mechanism of ODLW achieves the required bandwidth or larger bandwidth, low accumulated delay and reliable route based on node life time. Route maintenance procedure triggers when ever selected route between source and destination is break or changed due to the nodes mobility. In proposed ODLW ad hoc routing protocol, destination compares multiple received RREQM, by analyzing the required bandwidth, end to end delay and in the end node life time information contained in the RREQM. Once selection is made, destination node starts a timer to keep track of the availability of the selected route, and RREPM is unicast to the first available intermediate node. When intermediate node receives RREPM, it check the REQUEST_ID field in the packet header of RREPM to search for the corresponding TABLE_ID, it had created in its memory. Once table related to REQUEST_ID is found, it updates the INTERMEDIATE_ID field in RREPM, and continues unicasting (to original source). This process continues until the RREPM reach the original source node. After receiving the RREPM, source node selects next hop who best satisfy the QoS requirements and transmit data packet to the next hop. If data packets do not arrive to the destination node and timer expires at destination node, it is assumed that the selected route between source and destination is lost or break. In this case destination node selects alternative calculated route and unicast a new RREPM after starting the timer again. The alternative route is available for all the nodes, who received the RREQM. DST 2Mbps

G,4,7,3 C,3,4,5

4,7 F

H,4,10,3 F,4,10,4 C,2,7,5

3,1 G C

D,4,5,3 C,3,6,5

3,3

5,2 3,1 4,2

5,6

E B,2,2,2 A,5,3,4

D

2,1 5,3

5,1

5,1

NL

B,5,3,3 C,2,4,5

5,2

Link Weight BW Delay

Node Lifetime (NL) 1 = INF 2 =