+AODV With Sufficient Bandwidth Aware Routing Protocol
Content
AODV with Sufficient Bandwidth Aware Routing Protocol
Pattana Wannawilai
Faculty of Information Technology King Mongkut’s Institute of Technology Ladkrabang Bangkok 10520, Thailand Tel: +66-81658-4732
Chanboon Sathitwiriyawong
Faculty of Information Technology King Mongkut’s Institute of Technology Ladkrabang Bangkok 10520, Thailand Tel: +66-2723-4911
[email protected] ABSTRACT
Congestion is a major problem in mobile ad hoc network (MANET) which causes long delay and significant loss of data packets, and increases the routing overhead and battery power consumption. The shortest path route seldom offers the optimal route, especially when it traverses the congested area of the mobile network. This paper proposes a novel AODV with Sufficient Bandwidth Aware (AODV+SBA) routing protocol which significantly improves the performance of on-demand routing protocols by discovering better routes to avoid congestion and reducing excessive routing overhead. The ns-2 simulation results illustrate the improvement of network performance and stability by reducing data packet delay and routing overhead and increasing packet delivery ratio, under high traffic load. Moreover, in case of low-to-medium traffic load, its performance is close to the popular AODV routing protocol and the quality of its properties is maintained.
[email protected]
information sharing in a conference, military actions and disaster rescues. However, multi hop routing, random movement of mobile nodes and other features unique to MANET lead to enormous control overhead for route discovery and maintenance. In some scenarios, the routing maintenance overhead may consume so much resource that it seriously compromises long term efficiency. Furthermore, compared with the traditional networks, MANET suffers from the resource constraints in energy, computational capacities and bandwidth. All of these make routing in MANET a very challenging problem. To address the routing challenge in MANET, many approaches have been proposed in the literature. Based on the routing mechanism for the traditional networks, the proactive approaches attempt to maintain routing information for each node in the network at all times (e.g., [1] [7]) whereas the reactive approaches only find new routes when required (e.g., [2] [3]). Other approaches make use of geographical location information for routing (e.g., [9]). In most of the previous works, the number of hops is the most common criterion to determine routing. However, in MANET, shortest path (or minimum hop count) routings such as DSDV [1] and AODV [2] produce some areas of the network that are likely to have higher data loads than other areas, especially at the central network. This can make certain areas prone to congestion, thus decreasing the overall network performance. In such case, the criteria or metric based on the number of hops will not suffice for making routing decisions. In this paper, we propose a new improved version of Ad hoc Ondemand Distance Vector (AODV) that uses a light-weight mechanism to determine network congestion. It is based on the information acquired from the MAC layer, to improve algorithm performance. This algorithm which we call AODV+SBA, uses the concept of congestion avoidance that prohibits the new route to allow additional traffic coming into the congested area. Our algorithm adopts the cross-layer design approach by utilizing parameters from different layers to achieve overall system optimization. The parameters for measuring local network congestion around a node depend largely on the MAC layer. In this paper, we focus on the IEEE 802.11 DCF mode, since it is the most widely used wireless LAN standard. By using the wireless medium information from the MAC layer, AODV+SBA prevents the discovery of routes over which it is undesirable to carry additional data and routing traffic since the wireless medium over those hops is already very busy. The simulation results show many significant improvement of network performance and stability as well as a noticeable increase of data delivery ratio and decrease of data packet delay and routing overhead, especially in stressful network situations. The organization of the rest part of the paper is as follows. Section II briefly reviews related works. Section III provides Channel Free Time metric. Section IV describes the details of proposed AODV+SBA
Categories and Subject Descriptors
C.2.1 [Computer-Communication Networks]: Network Architecture and Design – Wireless communication; C.2.2 [ComputerCommunication Networks]: Network Protocols – Routing protocols
General Terms
Algorithms, Performance, Experimentation
Keywords
Mobile ad hoc network, on-demand routing, congestion, sufficient bandwidth.
1. INTRODUCTION
A mobile ad hoc network (MANET) is a mobile wireless network that is formed spontaneously. It is a collection of autonomous mobile computing nodes that communicate with each other over packet radio and without using any existing network infrastructure, and thus to be self creating, self organizing, and battery-powered. Unlike the traditional wireless networks, communication in such a decentralized network is typically multi-hop, with the nodes using each other as relay routers without any fixed infrastructure. This kind of network is very flexible and suitable for applications such as temporary
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee "IWCMC’10, June 28– July 2, 2010, Caen, France. Copyright © 2010 ACM 978-1-4503-0062-9/10/06/...$10.00"
281
routing protocol. Section V demonstrates the performance evaluation, and finally section VI concludes the paper.
2. RELATED WORKS
In recent years, many researchers have strived to improve the performance of MANET. They have developed numerous protocols for efficient flooding such as DLAR [10] and LWR [11], and have improved the routing metric for congestion detection such as ABR [12], CADV [13] and QAR [14], and CRP [15]. DLAR scheme uses the number of packets buffered in the interface as the primary route selection criteria. LWR scheme considers local information such as channel utilization, queue size, the number of active neighbors and the value of back-off time. ABR scheme considers the routing load as the secondary metric. CADV uses expected delay to measure congestion. QAR uses bandwidth estimation as the metric of the admission schema. CRP uses the knowledge of each node about its packet buffering status, such as the average queue length and the average packet delay, as a congestion indicator. Due to the limit and shared bandwidth among nodes located within their transmission ranges, the packet buffering status does not always reflect the availability of their outgoing links. The proposed AODV+SBA routing protocol offers on-demand routing with congestion mitigation by considering a channel free time representing network capacity still available, as a metric, during the route establishment phase. Node drops route request packets when the channel free time is less than the specified threshold.
each data transmission. This overhead makes it impossible in a distributed MAC competition scheme to fully use the available bandwidth for data transmission. The available period is idle time or CFT. The IEEE 802.11 uses RTS/CTS to detect free channel when the following three requirements are met. • • • NAV value is less than the current time. Receiving state is idle. Sending state is idle.
The MAC claims that the channel is busy when one of the following occurs: • • •
Sender Receiver
NAV sets a new value. Receive state changes from idle to any other state. Send state changes from idle to any other state.
DIFS RTS SIFS CTS SIFS DATA SIFS ACK
Other
RTS-NAV CTS-NAV
DIFS
RTS = Request to Send SIFS = Short Inter Frame Space
CTS = Clear to Send DIFS = Distributed Inter Frame Space
NAV = Network Allocation Vector
Figure 1. Protocol timing of the IEEE 802.11 RTS/CTS.
3. CHANNEL FREE TIME
Congested area defense is a lightweight method to improve network performance and stability. Congestion in MANET causes long delay, high packet loss rate, and high routing overhead incurred from frequent rerouting. Node movement and media sharing produce dynamic bandwidth in high density area of nodes. All nodes in this area share the channel and increase the chance of congestion and instability. While the high priority of routing packets makes congestion worse. Therefore, the shortest route is seldom the best route when packets traverse the congested area. The residual bandwidth is used to protect the congested area. In case the area has a sufficient bandwidth to accept new data traffic and does not effect the current communication, the new data traffic can traverse it. Otherwise, the new data traffic is prohibited. Therefore, no additional new traffic traverses the congested area to worsen the situation. But so far, the difficulty of calculating the residual bandwidth using the IEEE 802.11 is still a challenging problem since the bandwidth is interfered and shared among neighbors. Moreover, each node does not have knowledge of traffic status of its neighbors. The estimation of the network capacity available can be done by calculating the Channel Free Time (CFT). The status flags in the IEEE 802.11 can determine the free and busy times. Fig. 1 illustrates the protocol timing of the IEEE 802.11 RTS/CTS. The sender is busy while sending RTS until ACK is responded. Likewise, the receiver is busy while receiving RTS until ACK is successfully sent. Other nodes receive RTS and/or CTS that specify the Virtual Carrier Sense or Network Allocation Vector (NAV) to announce the channel busy time. The DIFS, SIFS and backoff scheme represent overhead, which must be accounted for in
4. PROPOSED AODV with SUFFICIENT BANDWIDTH AWARE (AODV+SBA) ROUTING PROTOCOL
The proposed AODV with Sufficient Bandwidth Aware (AODV+SBA) routing protocol is the extension to the well-known AODV routing protocol [2] that add CFT constraint to the route establishment phase for performance improvement and network stability. It is a reactive single path routing protocol comprising the following three divisions.
4.1 Channel Free Time Status Monitoring
The CFT is a metric used in the MAC layer. It can be used to make routing decision that taking into account the status of local wireless medium, which helps to avoid choosing routes through the congested area of the network. This metric provides the node with a view of current status of the shared wireless medium. The instantaneous MAC layer utilization at the node is either 1 (busy) or 0 (idle). In our algorithm, we accumulate the time (Tbusy) when the medium is busy over a period of time (Tmac) to reflect the usage of the wireless medium around the node as follows:
CFT = 1 −
∑ Tbusy Tmac
, T busy ⊆ Tmac
(1)
The status of congestion (Cs) can be indicated by two levels: forward and drop. The forward level mean the Route Request (RREQ) packet can be processed and broadcasted to the next neighbor. The drop level means there is no way to process and broadcast packet, just drop it. • • If CFT >= Threshold, Cs = Forward level. If CFT < Threshold, Cs = Drop level.
282
The threshold is a tuning parameter that helps to improve network performance and its value is in the range [0, 1]. If the threshold is equal to zero, the AODV+SBA routing protocol works the same as the original AODV routing protocol. Our preliminary simulation, the threshold used in AODV+SBA is set to 0.1.
When RREQ packet received. If (Cs = drop level) { drop RREQ packet received; }
4.2 Route Discovery
Fig. 2 illustrates the extension procedure for creating and sending a Route Request (RREQ) packet for route discovery. During routing discovery, when a source requires a path to its destination, it determines the level of congestion. If Cs is in the forward level, Route Request (RREQ) packet will be created and broadcasted to its neighbor as in the original AODV routing protocol. Otherwise, RREQ packet will be dropped to avoid the entry of routing and data packets into the congested area.
When RREQ packet is created and broadcasted. If (Cs = drop level) { stop this process; } Else { create RREQ packet; broadcast RREQ packet to its neighbors; }
Else { If (the first copy) { create reversed route to the source; } If (the destination node or any intermediate node has an up-to-date route) { send RREP packet to the source; } Else { update RREQ packet and rebroadcast it to its neighbors; } }
Figure 3. Decision to process route request packet in AODV+SBA.
5.1 Simulation Parameters
The simulated network model comprised 50 nodes in a 1,500m x 300m rectangular field that can easily simulate congestion scenario. The shared-media radio model was Lucent’s WaveLAN operating at a nominal bit rate of 2 Mbps and a nominal radio range of 250 meters. The MAC layer was based on the IEEE 802.11 Distributed Coordination Function (DCF). The channel propagation model we used combined both the free-space and two-ray ground reflection models. The same configuration parameters were used as in ns-2 version 2.30. An interface queue at the MAC layer could hold up to 50 packets before they were sent out to the physical link. A routing buffer at the network layer could store up to 64 data packets. This buffer keeps data packets waiting for a route, such as packets for which route discovery had started but no reply arrived yet. The mobility model uses the random waypoint model [4] while Mobgen-ss [8] is used instead of Setdest [6] to decrease the transient time and increase the confidence of simulation results. Lu et at. [15] suggested that, by setting the maximum node velocity to 4 meters per second, we could cover most mobility effects by just vary the pause period. Therefore, we used this maximum speed and considered different pause period: 800 seconds (lowest mobility) and zero second (highest mobility). For each off these cases, two simple ways can be used to illustrate different traffic loads: 1) fix the packet rate and vary the number of connections or 2) fix the number of connection and vary the packet rate. We employed the latter approach. Each simulation run lasted for 800 seconds, during which 20 connections were generated and remained open until the simulation ended. For each connection the source generated 512-byte data packets at a constant bit rate (CBR). This rate was varied among 1, 5, 10, 15, 20, and 40 packets per second.
Figure 2. Decision to create and broadcast route request packet in AODV+SBA. Fig. 3 illustrates the extension procedure to receive a RREQ packet. When a node receives a RREQ packet, it first checks the value of Cs to decide whether to drop or to continue the RREQ packet before proceeding to the next process. If the value of Cs is the drop level, the packet is dropped to reduce routing overhead and block data traffic that may traverse the congested area in the future. Otherwise, the node processes the RREQ packet the same as the original AODV routing protocol. Each node responds to the first copy of the RREQ packet by creating the reversed route to the source. The destination node or any intermediate node has an up-to-date route sending back to the source by Route Reply (RREP) packet. Along the route, the RREQ packet is subsequently updated and rebroadcasted to its next neighbor towards the source.
4.3 Route Maintenance
Routing maintenance phase is the same as the original AODV. Each node uses periodic Hello packet probe to detect the link failure and keep track of its immediate neighbors. If a node does not receive Hello packets from its next neighbors within a certain time, the node will invalidate all the routes containing this link by sending the Route Error (RERR) packet to all the upstream nodes that use this link. Once a source node receives RERR packet, it re-initiates a route discovery process to search for a new route.
5. PERFORMANCE EVALUATION
The performance study uses the network simulator (ns-2) [5] with CMU monarch wireless extension [6] to compare the proposed AODV+SBA with the original AODV.
5.2 Performance Metrics
Three important performance metrics are evaluated:
283
Data Packets Delivery Ratio (%)
(1) Packet Delivery Ratio: Percentage of data packets received at the destinations out of the number of data packets generated by the CBR traffic sources. (2) End-to-End Delay: The accumulative delay in data packet delivery due to buffering of packets, new route discoveries, queuing delay, MAC-layer retransmission, and transmission and propagation delays. (3) Normalized Routing Overhead: The ratio of the amount in bytes of control packets transmitted to the amount in bytes of data received. A desirable routing protocol should offer high packet delivery ratio, low end-to-end delay and low routing overhead.
Pause Time = 800 seconds 110 100 90 80 70 60 50 40 30 20 10 0 5 10 15 20 25 30 35 40 Packet rate per source (packets/sec) AODV+SBA AODV
5.3 Simulation Results
5.3.1 Lowest Mobility (mobility pause = 800 seconds)
By setting the pause period to 800 seconds, the packet loss in the steady network was only due to channel error and network congestion, not mobility. Therefore, since AODV+SBA distributes routes and traffic overall network, we expected that AODV+SBA would offer the best delay and throughput. This hypothesis indeed tested true in our simulation. Fig. 4 shows the results for the average delay. In the steady network with little traffic (one packet per second), the remaining bandwidth is sufficient for sending data and routing packets. AODV+SBA routing performs the same as the original AODV. While, the sending rate is increased to 5, 10, 15, 20 and 40, AODV+SBA was clearly better than AODV. For instance, when the packet rate was 20 packets per seconds, the average delay only 33 percent of the original AODV. Fig. 5 shows the results for the data delivery ratio. In a steady network with little traffic, all protocols performed well when 99.99 percent of packets were delivered successfully. However, AODV+SBA delivered more data as the packet rate was increased. Obviously, in highly stressful network (40 packets per second), the data delivery ratio is increased up to 149 percent of that AODV. In term of normalized routing overhead (Fig. 6), the AODV+SBA performs the same as the original AODV because route re-discovery is not required in the steady network. Routing overhead is caused by the Hello beacons mechanism which performs local connectivity management.
Pause Time = 800 seconds 3 AODV+SBA AODV 2.5 2 1.5 1 0.5 0 0 5 10 15 20 25 30 35 Packet rate per source (packets/sec) 40
Figure. 5. Data packet delivery ratio and confidence interval.
Pause Time = 800 seconds 0.055 0.05 AODV+SBA AODV
Normalized Routing (%)
0.045 0.04 0.035 0.03 0.025 0.02 0.015 0 5 10 15 20 25 30 35 40 Packet rate per source (packets/sec)
Figure 6. Normalized routing overhead and confidence interval.
5.3.2 Highest Mobility (mobility pause = 0 seconds)
The average node velocity is 4 meters per second without pause time or continues mobility which is the worst case of dynamic topology change in MANET. The packets are lost or dropped not only due to network congestion but also mobility. Fig. 7 shows the average delay and confidence interval period. The proposed AODV+SBA outperforms the original AODV, especially when the network traffic is heavily loaded. In case of small sending rates, 1 and 5 packets second, the remaining bandwidth supports thepresence of all data and routing packets. Therefore, for normal traffic loading the proposed AODV+SBA routing protocol keeps up the property of the original AODV routing protocol. More impressively, when the sending rates are increased to 10, 15 and 20 packets per second, AODV+SBA is significantly better than AODV. For instance, when the packet rate is 10 packets per second, the average delay and confidence interval of AODV+SBA is less than 50 percent and 39 percent of the original AODV, respectively. In regard of data packet delivery ratio (Fig. 8), when packet rate was small (1 and 5 packet per seconds), AODV+SBA and AODV delivered similar load of packets. This was because network traffic was not yet heavy, but if more data traffic were transmitted from the source, AODV+SBA delivered more data successfully. For instance, CRP successfully delivered at most 144 percent more data than AODV when the packet rate was 40 packets per second. In comparing their routing overhead, AODV+SBA was much more lightweight than the original AODV. When the traffic load was small (1 packet per second), AODV+SBA performs the same as the original AODV routing protocol. However, when the traffic was heavy, the
Figure 4. Average end-to-end delay and confidence
interval.
Delay (sec)
284
routing overhead is decreased up to 37 percent of the AODV as show in Fig. 9.
Pause Time = 0 second 3 AODV+SBA AODV 2.5 2 1.5 1 0.5 0 0 5 10 15 20 25 30 35 40 Packet rate per source (packets/sec)
• Protocol overhead: AODV+SBA is significantly better than AODV when the network is heavily loaded and highly mobile. The proposed AODV+SBA routing protocol works better than the original AODV routing protocol under highly stressful network since it reduces the control overhead and establishes a route to avoid the congested area.
Delay (sec)
6. CONCLUSIONS
This paper has presented an AODV with Sufficient Bandwidth Aware (AODV+SBA) routing protocol. Several contributions have been made. First, it significant increases the network performance and stability when the network is heavily loaded. Second, it establishes a better route by avoiding the congested area. Third, it reduces the routing overhead as well as the battery power consumption to enhance the network lifetime. Finally, it preserves compatibility with the wellknown AODV routing protocol without modification of the original routing packets and internal processes.
Figure 7. Average end-to-end delay and confidence interval.
Pause Time = 0 second
Data Packets Delivery Ratio (%)
90 80 70 60 50 40 30 20 10 0 5 10 15 20 25 30 35 40 Packet rate per source (packets/sec) AODV+SBA AODV
7. REFERENCES
[1] Perkins, C. E., and Bhagwat, P. 1994. Highly dynamic destination-sequenced distance-vector routing (DSDV) for mobile computers. Proc. ACM SIGCOMM. 234-244. [2] Perkins, C. E., Belding-Royer, E. M., and Das, S. 2003. Ad hoc on-demand distance vector (AODV) routing. RFC 3561. [3] Johnson, D., Hu, Y. and Maltz, D. 2007. The dynamic source routing protocol for mobile ad hoc networks for IPv4. RFC 4728. [4] Broach, J., Maltz, D., Johnson, D., Hu, Y. and Jetcheva, J. 1998. A performance comparison of multi-hop wireless ad hoc network routing protocols. Proc. ACM/IEEE MOBICOM. 85-97. [5] ns-2 : Network Simulator 2. http://www.isi.edu/nsnam/ns/ [6] CMU. 1999. The CMU Monarch project: wireless and mobility extensions to NS. http://www.monarch.cs.rice.edu/cmu-ns.html. [7] Clausen, T. and Jacquet, P. 2003. Optimized link state routing protocol (OLSR). RFC 3626. [8] Navidi, W. and Camp, T. 2004. Stationary distributions for the random waypoint mobility model. IEEE Trans. on Mobile Computing. 3, 1 (Jan-Mar 2004), 99-108. [9] Ko, Y. and Yaidya, N. 1998. Location-Aided Routing (LAR) in Mobile Ad Hoc Networks. Proc. ACM/IEEE MOBICOM. 66-75. [10] Lee, S. J. and Gerla, M. 2001. Dynamic load-aware routing in ad hoc networks. Proc. IEEE ICC. 3206-3210. [11] Yi, Y., Kwon, T. J. and Gerla, M. 2001. A load aware routing based on local information. Proc. IEEE PIMRC. G-65-G-69. [12] Toh, C.-K. 1997. Associatively-based routing for ad-hoc mobile networks. Wireless Personal Comm. Journal. 4, 2 (Mar. 1997), 103-139. [13] Lu, Y., Wang, W., Zhong, Y. and Bhargave, B. 2003. Study of distance vector routing protocols for mobile ad hoc networks. Proc. IEEE PERCOM. 187-194. [14] Geng, R. and Li, Z. 2006. QoS-aware routing based on local information for mobile ad hoc networks. LNCS. 159-167. [15] Tran, D. A. and Raghavendra,H. 2006 Congestion adaptive routing in mobile ad hoc network. IEEE Trans. on Parallel and Distributed systems. 17, 11 (Nov. 2006), 1294-1305.
Figure 8. Data packet delivery ratio and confidence interval.
Pause Time = 0 second 0.16 0.14
Normalized Routing (%)
AODV+SBA AODV
0.12 0.1 0.08 0.06 0.04 0.02 0 0 5 10 15 20 25 30 35 40 Packet rate per source (packets/sec)
Figure 9. Normalized routing overhead and confidence interval.
5.4 Comparison Remarks
The following highlights are concluded from the performance evaluation: • End-to-end delay: Consistently in simulation runs, AODV+SBA provides an average delay shorter than AODV. • Data packet delivery ratio: When the network is heavily loaded, whether the network is steady or highly mobile, AODV+SBA performs better than AODV. In the other cases (only a few), they performs similarly.
285
Pattana Wannawilai
Faculty of Information Technology King Mongkut’s Institute of Technology Ladkrabang Bangkok 10520, Thailand Tel: +66-81658-4732
Chanboon Sathitwiriyawong
Faculty of Information Technology King Mongkut’s Institute of Technology Ladkrabang Bangkok 10520, Thailand Tel: +66-2723-4911
[email protected] ABSTRACT
Congestion is a major problem in mobile ad hoc network (MANET) which causes long delay and significant loss of data packets, and increases the routing overhead and battery power consumption. The shortest path route seldom offers the optimal route, especially when it traverses the congested area of the mobile network. This paper proposes a novel AODV with Sufficient Bandwidth Aware (AODV+SBA) routing protocol which significantly improves the performance of on-demand routing protocols by discovering better routes to avoid congestion and reducing excessive routing overhead. The ns-2 simulation results illustrate the improvement of network performance and stability by reducing data packet delay and routing overhead and increasing packet delivery ratio, under high traffic load. Moreover, in case of low-to-medium traffic load, its performance is close to the popular AODV routing protocol and the quality of its properties is maintained.
[email protected]
information sharing in a conference, military actions and disaster rescues. However, multi hop routing, random movement of mobile nodes and other features unique to MANET lead to enormous control overhead for route discovery and maintenance. In some scenarios, the routing maintenance overhead may consume so much resource that it seriously compromises long term efficiency. Furthermore, compared with the traditional networks, MANET suffers from the resource constraints in energy, computational capacities and bandwidth. All of these make routing in MANET a very challenging problem. To address the routing challenge in MANET, many approaches have been proposed in the literature. Based on the routing mechanism for the traditional networks, the proactive approaches attempt to maintain routing information for each node in the network at all times (e.g., [1] [7]) whereas the reactive approaches only find new routes when required (e.g., [2] [3]). Other approaches make use of geographical location information for routing (e.g., [9]). In most of the previous works, the number of hops is the most common criterion to determine routing. However, in MANET, shortest path (or minimum hop count) routings such as DSDV [1] and AODV [2] produce some areas of the network that are likely to have higher data loads than other areas, especially at the central network. This can make certain areas prone to congestion, thus decreasing the overall network performance. In such case, the criteria or metric based on the number of hops will not suffice for making routing decisions. In this paper, we propose a new improved version of Ad hoc Ondemand Distance Vector (AODV) that uses a light-weight mechanism to determine network congestion. It is based on the information acquired from the MAC layer, to improve algorithm performance. This algorithm which we call AODV+SBA, uses the concept of congestion avoidance that prohibits the new route to allow additional traffic coming into the congested area. Our algorithm adopts the cross-layer design approach by utilizing parameters from different layers to achieve overall system optimization. The parameters for measuring local network congestion around a node depend largely on the MAC layer. In this paper, we focus on the IEEE 802.11 DCF mode, since it is the most widely used wireless LAN standard. By using the wireless medium information from the MAC layer, AODV+SBA prevents the discovery of routes over which it is undesirable to carry additional data and routing traffic since the wireless medium over those hops is already very busy. The simulation results show many significant improvement of network performance and stability as well as a noticeable increase of data delivery ratio and decrease of data packet delay and routing overhead, especially in stressful network situations. The organization of the rest part of the paper is as follows. Section II briefly reviews related works. Section III provides Channel Free Time metric. Section IV describes the details of proposed AODV+SBA
Categories and Subject Descriptors
C.2.1 [Computer-Communication Networks]: Network Architecture and Design – Wireless communication; C.2.2 [ComputerCommunication Networks]: Network Protocols – Routing protocols
General Terms
Algorithms, Performance, Experimentation
Keywords
Mobile ad hoc network, on-demand routing, congestion, sufficient bandwidth.
1. INTRODUCTION
A mobile ad hoc network (MANET) is a mobile wireless network that is formed spontaneously. It is a collection of autonomous mobile computing nodes that communicate with each other over packet radio and without using any existing network infrastructure, and thus to be self creating, self organizing, and battery-powered. Unlike the traditional wireless networks, communication in such a decentralized network is typically multi-hop, with the nodes using each other as relay routers without any fixed infrastructure. This kind of network is very flexible and suitable for applications such as temporary
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee "IWCMC’10, June 28– July 2, 2010, Caen, France. Copyright © 2010 ACM 978-1-4503-0062-9/10/06/...$10.00"
281
routing protocol. Section V demonstrates the performance evaluation, and finally section VI concludes the paper.
2. RELATED WORKS
In recent years, many researchers have strived to improve the performance of MANET. They have developed numerous protocols for efficient flooding such as DLAR [10] and LWR [11], and have improved the routing metric for congestion detection such as ABR [12], CADV [13] and QAR [14], and CRP [15]. DLAR scheme uses the number of packets buffered in the interface as the primary route selection criteria. LWR scheme considers local information such as channel utilization, queue size, the number of active neighbors and the value of back-off time. ABR scheme considers the routing load as the secondary metric. CADV uses expected delay to measure congestion. QAR uses bandwidth estimation as the metric of the admission schema. CRP uses the knowledge of each node about its packet buffering status, such as the average queue length and the average packet delay, as a congestion indicator. Due to the limit and shared bandwidth among nodes located within their transmission ranges, the packet buffering status does not always reflect the availability of their outgoing links. The proposed AODV+SBA routing protocol offers on-demand routing with congestion mitigation by considering a channel free time representing network capacity still available, as a metric, during the route establishment phase. Node drops route request packets when the channel free time is less than the specified threshold.
each data transmission. This overhead makes it impossible in a distributed MAC competition scheme to fully use the available bandwidth for data transmission. The available period is idle time or CFT. The IEEE 802.11 uses RTS/CTS to detect free channel when the following three requirements are met. • • • NAV value is less than the current time. Receiving state is idle. Sending state is idle.
The MAC claims that the channel is busy when one of the following occurs: • • •
Sender Receiver
NAV sets a new value. Receive state changes from idle to any other state. Send state changes from idle to any other state.
DIFS RTS SIFS CTS SIFS DATA SIFS ACK
Other
RTS-NAV CTS-NAV
DIFS
RTS = Request to Send SIFS = Short Inter Frame Space
CTS = Clear to Send DIFS = Distributed Inter Frame Space
NAV = Network Allocation Vector
Figure 1. Protocol timing of the IEEE 802.11 RTS/CTS.
3. CHANNEL FREE TIME
Congested area defense is a lightweight method to improve network performance and stability. Congestion in MANET causes long delay, high packet loss rate, and high routing overhead incurred from frequent rerouting. Node movement and media sharing produce dynamic bandwidth in high density area of nodes. All nodes in this area share the channel and increase the chance of congestion and instability. While the high priority of routing packets makes congestion worse. Therefore, the shortest route is seldom the best route when packets traverse the congested area. The residual bandwidth is used to protect the congested area. In case the area has a sufficient bandwidth to accept new data traffic and does not effect the current communication, the new data traffic can traverse it. Otherwise, the new data traffic is prohibited. Therefore, no additional new traffic traverses the congested area to worsen the situation. But so far, the difficulty of calculating the residual bandwidth using the IEEE 802.11 is still a challenging problem since the bandwidth is interfered and shared among neighbors. Moreover, each node does not have knowledge of traffic status of its neighbors. The estimation of the network capacity available can be done by calculating the Channel Free Time (CFT). The status flags in the IEEE 802.11 can determine the free and busy times. Fig. 1 illustrates the protocol timing of the IEEE 802.11 RTS/CTS. The sender is busy while sending RTS until ACK is responded. Likewise, the receiver is busy while receiving RTS until ACK is successfully sent. Other nodes receive RTS and/or CTS that specify the Virtual Carrier Sense or Network Allocation Vector (NAV) to announce the channel busy time. The DIFS, SIFS and backoff scheme represent overhead, which must be accounted for in
4. PROPOSED AODV with SUFFICIENT BANDWIDTH AWARE (AODV+SBA) ROUTING PROTOCOL
The proposed AODV with Sufficient Bandwidth Aware (AODV+SBA) routing protocol is the extension to the well-known AODV routing protocol [2] that add CFT constraint to the route establishment phase for performance improvement and network stability. It is a reactive single path routing protocol comprising the following three divisions.
4.1 Channel Free Time Status Monitoring
The CFT is a metric used in the MAC layer. It can be used to make routing decision that taking into account the status of local wireless medium, which helps to avoid choosing routes through the congested area of the network. This metric provides the node with a view of current status of the shared wireless medium. The instantaneous MAC layer utilization at the node is either 1 (busy) or 0 (idle). In our algorithm, we accumulate the time (Tbusy) when the medium is busy over a period of time (Tmac) to reflect the usage of the wireless medium around the node as follows:
CFT = 1 −
∑ Tbusy Tmac
, T busy ⊆ Tmac
(1)
The status of congestion (Cs) can be indicated by two levels: forward and drop. The forward level mean the Route Request (RREQ) packet can be processed and broadcasted to the next neighbor. The drop level means there is no way to process and broadcast packet, just drop it. • • If CFT >= Threshold, Cs = Forward level. If CFT < Threshold, Cs = Drop level.
282
The threshold is a tuning parameter that helps to improve network performance and its value is in the range [0, 1]. If the threshold is equal to zero, the AODV+SBA routing protocol works the same as the original AODV routing protocol. Our preliminary simulation, the threshold used in AODV+SBA is set to 0.1.
When RREQ packet received. If (Cs = drop level) { drop RREQ packet received; }
4.2 Route Discovery
Fig. 2 illustrates the extension procedure for creating and sending a Route Request (RREQ) packet for route discovery. During routing discovery, when a source requires a path to its destination, it determines the level of congestion. If Cs is in the forward level, Route Request (RREQ) packet will be created and broadcasted to its neighbor as in the original AODV routing protocol. Otherwise, RREQ packet will be dropped to avoid the entry of routing and data packets into the congested area.
When RREQ packet is created and broadcasted. If (Cs = drop level) { stop this process; } Else { create RREQ packet; broadcast RREQ packet to its neighbors; }
Else { If (the first copy) { create reversed route to the source; } If (the destination node or any intermediate node has an up-to-date route) { send RREP packet to the source; } Else { update RREQ packet and rebroadcast it to its neighbors; } }
Figure 3. Decision to process route request packet in AODV+SBA.
5.1 Simulation Parameters
The simulated network model comprised 50 nodes in a 1,500m x 300m rectangular field that can easily simulate congestion scenario. The shared-media radio model was Lucent’s WaveLAN operating at a nominal bit rate of 2 Mbps and a nominal radio range of 250 meters. The MAC layer was based on the IEEE 802.11 Distributed Coordination Function (DCF). The channel propagation model we used combined both the free-space and two-ray ground reflection models. The same configuration parameters were used as in ns-2 version 2.30. An interface queue at the MAC layer could hold up to 50 packets before they were sent out to the physical link. A routing buffer at the network layer could store up to 64 data packets. This buffer keeps data packets waiting for a route, such as packets for which route discovery had started but no reply arrived yet. The mobility model uses the random waypoint model [4] while Mobgen-ss [8] is used instead of Setdest [6] to decrease the transient time and increase the confidence of simulation results. Lu et at. [15] suggested that, by setting the maximum node velocity to 4 meters per second, we could cover most mobility effects by just vary the pause period. Therefore, we used this maximum speed and considered different pause period: 800 seconds (lowest mobility) and zero second (highest mobility). For each off these cases, two simple ways can be used to illustrate different traffic loads: 1) fix the packet rate and vary the number of connections or 2) fix the number of connection and vary the packet rate. We employed the latter approach. Each simulation run lasted for 800 seconds, during which 20 connections were generated and remained open until the simulation ended. For each connection the source generated 512-byte data packets at a constant bit rate (CBR). This rate was varied among 1, 5, 10, 15, 20, and 40 packets per second.
Figure 2. Decision to create and broadcast route request packet in AODV+SBA. Fig. 3 illustrates the extension procedure to receive a RREQ packet. When a node receives a RREQ packet, it first checks the value of Cs to decide whether to drop or to continue the RREQ packet before proceeding to the next process. If the value of Cs is the drop level, the packet is dropped to reduce routing overhead and block data traffic that may traverse the congested area in the future. Otherwise, the node processes the RREQ packet the same as the original AODV routing protocol. Each node responds to the first copy of the RREQ packet by creating the reversed route to the source. The destination node or any intermediate node has an up-to-date route sending back to the source by Route Reply (RREP) packet. Along the route, the RREQ packet is subsequently updated and rebroadcasted to its next neighbor towards the source.
4.3 Route Maintenance
Routing maintenance phase is the same as the original AODV. Each node uses periodic Hello packet probe to detect the link failure and keep track of its immediate neighbors. If a node does not receive Hello packets from its next neighbors within a certain time, the node will invalidate all the routes containing this link by sending the Route Error (RERR) packet to all the upstream nodes that use this link. Once a source node receives RERR packet, it re-initiates a route discovery process to search for a new route.
5. PERFORMANCE EVALUATION
The performance study uses the network simulator (ns-2) [5] with CMU monarch wireless extension [6] to compare the proposed AODV+SBA with the original AODV.
5.2 Performance Metrics
Three important performance metrics are evaluated:
283
Data Packets Delivery Ratio (%)
(1) Packet Delivery Ratio: Percentage of data packets received at the destinations out of the number of data packets generated by the CBR traffic sources. (2) End-to-End Delay: The accumulative delay in data packet delivery due to buffering of packets, new route discoveries, queuing delay, MAC-layer retransmission, and transmission and propagation delays. (3) Normalized Routing Overhead: The ratio of the amount in bytes of control packets transmitted to the amount in bytes of data received. A desirable routing protocol should offer high packet delivery ratio, low end-to-end delay and low routing overhead.
Pause Time = 800 seconds 110 100 90 80 70 60 50 40 30 20 10 0 5 10 15 20 25 30 35 40 Packet rate per source (packets/sec) AODV+SBA AODV
5.3 Simulation Results
5.3.1 Lowest Mobility (mobility pause = 800 seconds)
By setting the pause period to 800 seconds, the packet loss in the steady network was only due to channel error and network congestion, not mobility. Therefore, since AODV+SBA distributes routes and traffic overall network, we expected that AODV+SBA would offer the best delay and throughput. This hypothesis indeed tested true in our simulation. Fig. 4 shows the results for the average delay. In the steady network with little traffic (one packet per second), the remaining bandwidth is sufficient for sending data and routing packets. AODV+SBA routing performs the same as the original AODV. While, the sending rate is increased to 5, 10, 15, 20 and 40, AODV+SBA was clearly better than AODV. For instance, when the packet rate was 20 packets per seconds, the average delay only 33 percent of the original AODV. Fig. 5 shows the results for the data delivery ratio. In a steady network with little traffic, all protocols performed well when 99.99 percent of packets were delivered successfully. However, AODV+SBA delivered more data as the packet rate was increased. Obviously, in highly stressful network (40 packets per second), the data delivery ratio is increased up to 149 percent of that AODV. In term of normalized routing overhead (Fig. 6), the AODV+SBA performs the same as the original AODV because route re-discovery is not required in the steady network. Routing overhead is caused by the Hello beacons mechanism which performs local connectivity management.
Pause Time = 800 seconds 3 AODV+SBA AODV 2.5 2 1.5 1 0.5 0 0 5 10 15 20 25 30 35 Packet rate per source (packets/sec) 40
Figure. 5. Data packet delivery ratio and confidence interval.
Pause Time = 800 seconds 0.055 0.05 AODV+SBA AODV
Normalized Routing (%)
0.045 0.04 0.035 0.03 0.025 0.02 0.015 0 5 10 15 20 25 30 35 40 Packet rate per source (packets/sec)
Figure 6. Normalized routing overhead and confidence interval.
5.3.2 Highest Mobility (mobility pause = 0 seconds)
The average node velocity is 4 meters per second without pause time or continues mobility which is the worst case of dynamic topology change in MANET. The packets are lost or dropped not only due to network congestion but also mobility. Fig. 7 shows the average delay and confidence interval period. The proposed AODV+SBA outperforms the original AODV, especially when the network traffic is heavily loaded. In case of small sending rates, 1 and 5 packets second, the remaining bandwidth supports thepresence of all data and routing packets. Therefore, for normal traffic loading the proposed AODV+SBA routing protocol keeps up the property of the original AODV routing protocol. More impressively, when the sending rates are increased to 10, 15 and 20 packets per second, AODV+SBA is significantly better than AODV. For instance, when the packet rate is 10 packets per second, the average delay and confidence interval of AODV+SBA is less than 50 percent and 39 percent of the original AODV, respectively. In regard of data packet delivery ratio (Fig. 8), when packet rate was small (1 and 5 packet per seconds), AODV+SBA and AODV delivered similar load of packets. This was because network traffic was not yet heavy, but if more data traffic were transmitted from the source, AODV+SBA delivered more data successfully. For instance, CRP successfully delivered at most 144 percent more data than AODV when the packet rate was 40 packets per second. In comparing their routing overhead, AODV+SBA was much more lightweight than the original AODV. When the traffic load was small (1 packet per second), AODV+SBA performs the same as the original AODV routing protocol. However, when the traffic was heavy, the
Figure 4. Average end-to-end delay and confidence
interval.
Delay (sec)
284
routing overhead is decreased up to 37 percent of the AODV as show in Fig. 9.
Pause Time = 0 second 3 AODV+SBA AODV 2.5 2 1.5 1 0.5 0 0 5 10 15 20 25 30 35 40 Packet rate per source (packets/sec)
• Protocol overhead: AODV+SBA is significantly better than AODV when the network is heavily loaded and highly mobile. The proposed AODV+SBA routing protocol works better than the original AODV routing protocol under highly stressful network since it reduces the control overhead and establishes a route to avoid the congested area.
Delay (sec)
6. CONCLUSIONS
This paper has presented an AODV with Sufficient Bandwidth Aware (AODV+SBA) routing protocol. Several contributions have been made. First, it significant increases the network performance and stability when the network is heavily loaded. Second, it establishes a better route by avoiding the congested area. Third, it reduces the routing overhead as well as the battery power consumption to enhance the network lifetime. Finally, it preserves compatibility with the wellknown AODV routing protocol without modification of the original routing packets and internal processes.
Figure 7. Average end-to-end delay and confidence interval.
Pause Time = 0 second
Data Packets Delivery Ratio (%)
90 80 70 60 50 40 30 20 10 0 5 10 15 20 25 30 35 40 Packet rate per source (packets/sec) AODV+SBA AODV
7. REFERENCES
[1] Perkins, C. E., and Bhagwat, P. 1994. Highly dynamic destination-sequenced distance-vector routing (DSDV) for mobile computers. Proc. ACM SIGCOMM. 234-244. [2] Perkins, C. E., Belding-Royer, E. M., and Das, S. 2003. Ad hoc on-demand distance vector (AODV) routing. RFC 3561. [3] Johnson, D., Hu, Y. and Maltz, D. 2007. The dynamic source routing protocol for mobile ad hoc networks for IPv4. RFC 4728. [4] Broach, J., Maltz, D., Johnson, D., Hu, Y. and Jetcheva, J. 1998. A performance comparison of multi-hop wireless ad hoc network routing protocols. Proc. ACM/IEEE MOBICOM. 85-97. [5] ns-2 : Network Simulator 2. http://www.isi.edu/nsnam/ns/ [6] CMU. 1999. The CMU Monarch project: wireless and mobility extensions to NS. http://www.monarch.cs.rice.edu/cmu-ns.html. [7] Clausen, T. and Jacquet, P. 2003. Optimized link state routing protocol (OLSR). RFC 3626. [8] Navidi, W. and Camp, T. 2004. Stationary distributions for the random waypoint mobility model. IEEE Trans. on Mobile Computing. 3, 1 (Jan-Mar 2004), 99-108. [9] Ko, Y. and Yaidya, N. 1998. Location-Aided Routing (LAR) in Mobile Ad Hoc Networks. Proc. ACM/IEEE MOBICOM. 66-75. [10] Lee, S. J. and Gerla, M. 2001. Dynamic load-aware routing in ad hoc networks. Proc. IEEE ICC. 3206-3210. [11] Yi, Y., Kwon, T. J. and Gerla, M. 2001. A load aware routing based on local information. Proc. IEEE PIMRC. G-65-G-69. [12] Toh, C.-K. 1997. Associatively-based routing for ad-hoc mobile networks. Wireless Personal Comm. Journal. 4, 2 (Mar. 1997), 103-139. [13] Lu, Y., Wang, W., Zhong, Y. and Bhargave, B. 2003. Study of distance vector routing protocols for mobile ad hoc networks. Proc. IEEE PERCOM. 187-194. [14] Geng, R. and Li, Z. 2006. QoS-aware routing based on local information for mobile ad hoc networks. LNCS. 159-167. [15] Tran, D. A. and Raghavendra,H. 2006 Congestion adaptive routing in mobile ad hoc network. IEEE Trans. on Parallel and Distributed systems. 17, 11 (Nov. 2006), 1294-1305.
Figure 8. Data packet delivery ratio and confidence interval.
Pause Time = 0 second 0.16 0.14
Normalized Routing (%)
AODV+SBA AODV
0.12 0.1 0.08 0.06 0.04 0.02 0 0 5 10 15 20 25 30 35 40 Packet rate per source (packets/sec)
Figure 9. Normalized routing overhead and confidence interval.
5.4 Comparison Remarks
The following highlights are concluded from the performance evaluation: • End-to-end delay: Consistently in simulation runs, AODV+SBA provides an average delay shorter than AODV. • Data packet delivery ratio: When the network is heavily loaded, whether the network is steady or highly mobile, AODV+SBA performs better than AODV. In the other cases (only a few), they performs similarly.
285
