You are on page 1of 6

AN ARCHITECTURE FOR INTEGRATING MOBILE AD HOC NETWORKS WITH THE INTERNET USING MULTIPLE MOBILE GATEWAYS

Mieso K. Denko Department of Computing and Information Science University of Guelph, Guelph, Ontario,N1G 2W1 denko@cis.uoguelph.ca Abstract
Mobile Ad hoc Networks (MANETs) are typically considered as stand-alone, autonomous networks that support multihop communication without relying on any existing infrastructure. However, the integration of MANETs and infrastructure networks such as the Internet extends the network coverage and increases the application domain of ad hoc networks. In this paper we propose an architecture for integrating MANETs and the Internet using multiple mobile gateways. We used an extended Ad hoc On demand Distance Vector (AODV) routing protocol and Mobile IP (MIP) to achieve the integration. The proposed architecture has two main features. First, it provides global communication between MANETs and the Internet using a subset MANET nodes called Mobile Gateways (MGs) and the MIP foreign agents. The MGs are selected among MANET nodes based on the node stability, load and distance metrics. Second, it allows the MANET nodes to maintain multiple routes to the MGs using hybrid gateway discovery mechanisms. The simulation results of the proposed architecture indicate that the use of multiple mobile gateways and hybrid gateway discovery mechanisms enhance the network performance while providing bi-directional Internet connectivity. Keywords: Mobile Internet, wireless networks, ad hoc networks, mobile gateways, mobile computing

Chen Wei Department of Computing and Information Science University of Guelph, Guelph, Ontario,N1G 2W1 chenw@uoguelph.ca
ad hoc networks with the Internet. In [2], an architecture called MIPMANET which uses MIP foreign agent AODV was proposed. MIPMANET allows a mobile node to switch from home agent domain to foreign agent domain. In [10], an integrated architecture that allows a gateway to decide time to live value for packet propagation was proposed. A hybrid gateway discovery mechanism that uses AODV and MIP foreign agents was presented. Two foreign agents and one MANET were used in their simulation experiments. In [7], a three-tier architecture that uses mobile gateways between nodes within MANETs and the Internet was proposed. The proposed scheme uses MIP and the Destination Sequenced Distance Vector (DSDV) routing protocol. It considers only communication initiated by MANET nodes. In this proposal, both foreign agents and mobile gateways broadcast agent advertisements to indicate their availability. In this paper, we propose an architecture for integrating MANETs and the Internet using multiple mobile gateways. The distinctive features of our proposal are: First unlike existing proposals which consider MANET as access networks to the fixed networks, we propose an architecture that supports bi-directional communication. Second, we consider multiple gateways and maintain multiple routes to gateways to provide load balancing and improve handover. Third, we use hybrid gateway discovery mechanisms to reduce communication overheads. The rest of this paper is organized as follows. Section 2 presents the description of the proposed architecture; Section 3 discusses the global communication scenarios; Section 4 presents results of performance evaluation and; Finally, Section 5 presents conclusions and future work.

1. Introduction
Several approaches have been proposed for integrating MANETs with infrastructure networks in recent years. However, these proposals consider MANETs as access networks to the Internet and provide only unidirectional Internet connectivity. Fixed gateways are used to achieve the integration task. Moreover, most existing proposals do not consider multiple gateways and hence lack mechanisms for load balancing and scalability. The Network Address Translator (NAT) and MIP [1], [2],[6], [13] are among the commonly used approaches for integrating ad hoc networks with the Internet. In [1], a mechanism that uses Routing Information Protocol (RIP) was proposed to integrate

2. Description of the Proposed Architecture 2.1. The Architecture and Functionality


Unlike existing proposals that consider MANET as access network to the Internet, we propose an architecture that enables global and bi-direction connectivity between MANETs and the Internet. We use mobile gateways to extend the coverage of

0-7803-8886-0/05/$20.00 2005 IEEE CCECE/CCGEI, Saskatoon, May 2005

1097

MIP agents. Figure 1 shows the components of the proposed architecture which consists of the MGs, MIP foreign agents, Mobile Nodes (MNs) and Correspondent nodes (CNs). The CNs can be static or mobile. In this paper, the terms mobile node and MANET nodes are used interchangeably while visiting mobile node refers to the Internet mobile nodes. The MG is a mobile node that is directly connected to the foreign agent. Other mobile nodes reach the mobile gateway over multiple hops. The MIP was designed to support one-hop communication and it uses proactive messages for location and address management. Ad hoc on demand routing protocols, on the other hand, operate reactively. To overcome this functionality disparity, we extended the MIP and AODV [12] protocols. The mobile gateway runs an extended AODV and MIP protocols to support mobility management, handover and hybrid gateway discovery.

nodes. A mobile gateway that did not registered with another foreign agent or had its registration lifetime expired sends registration request message to foreign agent. Also mobile gateways can broadcast solicitation message to learn the availability of foreign agents. In such a case, a foreign agent that receives the message unicasts the foreign agent advertisement to the mobile gateway. If a new MANET node seeks Internet connectivity it discovers mobile gateway as described in Section 2.5. After registration, a mobile node receives information such as the registration lifetime, IP address of mobile gateway and mobile gateway load in a route reply packet.

2.4. Selection of a Mobile Gateway


If a mobile gateway can reach multiple foreign agents, it will register with one having lower load and can adjust its selection based on the agent advertisement. To be selected as a mobile gateway, a node should be one-hop away from the foreign agent, should have relatively low mobility and load compared to other nodes. If there are multiple reachable mobile gateways, selection of a suitable gateway is made based on hop length and load metrics. A mobile node also maintains multiple mobile gateways.

2.5. Mobile Gateway Discovery


In [5],[8], proactive discovery and reactive discovery mechanisms were proposed for integrated networks using fixed gateways. In proactive discovery mechanism, the gateway periodically broadcasts the agent advertisement to other nodes in the network. Any mobile node that wants to interact with the Internet nodes detects this packet and begins registration. In reactive gateway discovery, a mobile node initiates the gateway discovery. The proactive discovery mechanism reduces the average delay compared to the reactive discovery but incurs higher communication overhead. To reduce the shortcomings of the proactive and reactive discover mechanisms a hybrid gateway discovery mechanism was proposed for fixed gateways in [11]. We propose both reactive and hybrid gateway discovery mechanisms for proposed architecture. The hybrid scheme is more suitable for relatively larger number of registered mobile nodes while the reactive scheme favors low number of registered nodes. The proposed gateway discovery mechanisms are described below. 2.5.1 Reactive scheme. The MG maintains connectivity information to foreign agents but does not broadcast its availability to mobile nodes. Nodes that are close to the mobile gateway use the proactive mechanism for gateway discovery while nodes that are far away from the gateway and seek Internet connectivity discover the gateway using a pure reactive mechanism. 2.5.2 Hybrid scheme. The mobile gateway broadcasts its availability to k-hop neighbors. Nodes that are not reachable by

Figure 1. The proposed architecture

2.2. Packets Processing


Incoming packets from the Internet are delivered to MANETs if the destination is at home or forwarded to the mobile gateway and tunneled to the destination network if the destination is not on network. All outgoing packets are routed along the mobile gateway and delivered to the corresponding foreign agents for further processing. The MG maintains a list of internal nodes and visiting mobile nodes. We assume that mobile gateways exchange information about all visiting mobile nodes and MANET nodes currently being served by them. A visiting mobile node can choose a particular MANET based on its proximity. A visiting node can be from another MANET or an Internet mobile node. When a node joins a MANET, it registers with the nearest mobile gateway to get Internet connectivity.

2.3. The Operation of the Mobile Gateways


The communication between foreign agents and mobile gateways can be reactive or proactive. The MIP foreign agents broadcast periodic agent advertisement to reachable MANET

1098

the MG, discover the nearest mobile gateway using reactive discovery mechanism. This scheme is an attempt to reduce communication overhead as not all mobile nodes want the Internet connectivity. Moreover, even if the number of nodes that register with a mobile gateway can be high, only few of them need Internet connection at the same time. Thus a hybrid scheme that favors reactive mechanism is more suitable in such scenarios.

4. Performance Evaluation 4.1. The simulation environment


4.1.1 Simulation scenarios. To evaluate the proposed architecture, we have carried out simulation experiments using the Network Simulator, NS-2[9]. The simulation experiments consist of up to 50 mobile nodes, five mobile gateways, three CNs, and three foreign agents. The simulation area used for these experiments is a square of 1000mx1000m with the radio transmission range of 250m. Up to 50% of the mobile nodes seek Internet connectivity. In the simulation experiments, hybrid 1and hybrid 2 refer to hybrid gateway discovery mechanism based on 1-hop and 2-hop neighbors respectively. Reactive refers to the pure reactive gateway discovery mechnism. 4.1.2 Mobility and communication models. Mobile nodes in MANETs move according to the Random Waypoint mobility model [3]. The speed of non-gateway MANET nodes is randomly chosen between 0 m/s and 20 m/s while that of mobile gateways is between 0 m/s and 5 m/s. The pause time was 10 and 50 seconds for mobile nodes and mobile gateways respectively. In the simulation, 20 CBR sources are used to generate packets at the rate of 4 packets per second. The packet size of 512 bytes was used throughout the simulation. All simulation experiments were run for 900 seconds of simulated time and each data point in the graphical results is based on the average of five simulation runs. 4.1.3. Performance metrics. We have used, average end-toend delay, throughput, mobile gateway overhead, nodegateway connectivity ratio and gateway lifetime as performance metrics. 1. The end-to-end delay is defined as the difference between the time the packet is received by the destination and the time the packet is generated by the source. 2. The normalized throughput is defined as the ratio of the individual throughputs and the total throughput. 3. The mobile gateway overhead is defined as the total number of registration and agent advertisement messages generated by the mobile gateway. 4. The node-gateway connectivity ratio is defined as the percentage of time when at least one gateway is reachable to the mobile nodes during the simulation time. 5. The gateway lifetime is defined as the length of time for which a node had valid path to the mobile gateway. The metrics gateway connectivity and gateway lifetime measure the effectiveness of the proposed mobile gateway approach in the integrated network.

3. Global Communication Scenarios


Packets are exchanged between the Internet and MANET nodes using one or a combination of three kinds of routing schemes: traditional IP routing, MIP routing and AODV-based ad hoc routing. In this section, we describe two main communication scenarios.

3.1. MANET-Internet Communication


When a static Internet node wishes to communicate with MANET, it uses the IP routing in the Internet, MIP routing between foreign agent and the mobile gateway and the AODV protocol within MANETs. On the other hand, when a MANET node wishes to communicate with the Internet, the packet is forwarded to the gateway which in turn forwards it to a foreign agent for further forwarding. A MANET node may also move from its home network to another MANET. For example, in Figure 1, when MN2 moves from the home network (MANET1) to foreign the foreign network (MANET2), it registers with the mobile gateway in MANET2. After the mobile node finishes the registration process, all packets destined to it will be encapsulated by home agent and forwarded to the hosting foreign agent. When it reaches the destination network, it is decapsulated and delivered to the destination through the serving mobile gateway.

3.2. Inter-MANET communication


When a communication is intended among independent MANETs that do not rely on the Internet, direct communication can be established if they are reachable. In such multiple MANET scenarios, the boundary between them is defined in terms of the hop length from a gateway. To achieve a global connectivity, communication between MANETs where source and destination nodes are located on two different networks is needed. Consider the communication between MANET2 and MANET3 in Figure 1. In this scenario, packets from mobile nodes are transmitted to the mobile gateway and foreign agent using AODV, transmitted through the Internet using the IP routing until they reache the foreign agent. From the foreign agent, it will be forwarded to the mobile gateway serving the destination MANET using MIP routing and finally delivered to the required destination using the AODV protocol.

1099

4.2. Discussion of Simulation Results


Throughput(%)

For each gateway discovery architecture, the effects of the nodes, the relative speed of advertisement intervals on the investigated.

mechanism in the proposed number of MGs, registered the MGs and the agent performance metrics were

100% 95% 90% 85% 80% 5 10 15 20 25 30 40 50 Registered nodes (%)


Figure 3. Average throughput

Hybrid 1 Hybrid 2 Reactive

4.2.1 The percentage of registered nodes. Figures 2-4 show the effect of registered nodes on the throughput, delay and gateway overhead. When the percentage of registered nodes increases, the throughput slightly decreases while the delay rises. When the number of registered nodes increases, a single mobile gateway can serve more nodes resulting in relatively lower throughput. The performance of hybrid schemes was observed to be slightly better than the pure reactive scheme. In the hybrid discovery scheme, as mobile gateways broadcast advertisements within periodic intervals, the overhead slightly increases. However, the overhead is higher in pure reactive scheme compared to the hybrid schemes when the number of registered nodes increases. As the number of registered nodes increases, the mobile gateway overhead slightly increases. However, using the hybrid scheme with lower hop length for agent advertisement can improve performance. 4.2.2 Number of mobile gateways. The effect of the number of mobile gateways on the throughput delay and node-gateway connectivity is shown in Figures 5-7. When the number of mobile gateways increases, the throughput increases. No significant difference was observed among all the schemes. However, there is relatively higher delay for reactive schemes than the hybrid schemes. The presence of multiple mobile gateways and the maintenance of multiple routes increased connectivity for all schemes. For the reason of excessive flooding overhead, less mobile gateways should be applied if only few nodes in the MANET require communication with Internet. It was also noted that an increase in the number of foreign agents improves performance. This occurs because mobile gateways have higher probability of registering with foreign agents in such cases. On the other hand while the increasing in the number of gateways can provide better load balancing mechanism, agent advertisements traffic can overload the network.
End-to-end delay (ms)

Gateway overhead

1500 1000 500 0 5 10 15 20 25 30 40 50 Registered nodes (%) Hybrid 1 Hybrid 2 Reactive

Figure 4. Average gateway overhead


120 100 80 60 40 20 0 1 2 3 4 5 Number of mobile gateways

Throughput (%)

Hybrid 1 Hybrid 2 Reactive

Figure 5. Average throughput


End-to-end delay (ms)

50 40 30 20 10 0 5 10 15 20 25 30 40 50 Registered nodes (%) Hybrid 1 Hybrid 2 Reactive

30 25 20 15 10 5 0 1 2 3 4 5 Number of mobile gateways

Hybrid 1 Hybrid 2 Reactive

Figure 6. Average end-to-end delay

Figure 2. Average end-to-end delay

1100

120 100 80 60 40 20 0 1 2 3 4 5 Number of mobile gateways

Connectivity ratio

Gateway overhead

1200 1000 800 600 400 200 0 5 10 20 30 40 50 Advertisement interval Hybrid 1 Hybrid 2 Reactive

Hybrid 1 Hybrid 2 Reactive

Figure 11. Average gateway overhead Figure 7. Average node-gateway connectivity


100%

120 100 80 60 40 20 0 0 1 2 3 4 5 Speed of mobile gatew ays (m/s)

Throughput (%)

Throughput (%)

95% 90% 85% 80% 5 10 20 30 40 50 Advertisement interval

Hybrid 1 Hybrid 2 Reactive

Hybrid 1 Hybrid 2 Reactive

Figure 12. Average throughput

End-to-end delay(ms)

Figure 8. Average throughput


120
Connectivity ratio

40 30 20 10 0 5 10 20 30 40 50 Advertisement interval Hybrid 1 Hybrid 2 Reactive

100 80 60 40 20 0 0 1 2 3 4 5 Speed of mobile gateways (m/s) Hybrid 1 Hybrid 2 Reactive

Figure 13. Average end-to-end delay

Figure 9. Average node-gateway connectivity

800 600 400 200 0 0 1 2 3 4 5 Speed of mobile gateways (m/s) Hybrid 1 Hybrid 2 Reactive

4.2.3 The effect of mobile gateway speed: Figures 8-10 show the effect of the speed of a gateway on the throughput, nodegateway connectivity and gateway lifetime. The throughput and node-gateway connectivity decrease slightly with increase in gateway speed. The gateway lifetime decrease with increase with the speed of the gateway. This occurs because when the speed of mobile gateways increases, more links will be broken resulting in shorter gateway lifetime. However, the effect of the motion of the gateway on performance was reduced due to the use of multiple routes. 4.2.4 The advertisement interval. Figures 11-13 show the effect of agent advertisement interval on the throughput, delay and gateway overhead. We have used an advertisement interval of 5-50 seconds in the simulation. When the advertisement interval increases, mobile gateway overhead decreases since there is less frequent traffic. The throughput in the reactive

Gateway lifetime (sec)

Figure 10. Average gateway lifetime

1101

scheme is lower than the hybrid schemes and less affected while compared to the hybrid schemes. Shorter advertisement intervals have shown better throughput than the longer intervals. The end-to-end delay increases slightly in hybrid schemes but still lower than reactive scheme. The end-to-end delay in reactive scheme is much affected by the advertisement interval. The increase in the advertisement interval reduces the benefit of the proactive route maintenance that occurs at nodes close to the gateway. Also too small advertisement interval may cause congestions due excessive traffic.

[7]

[8] [9] [10]

5. Summary and Future Work


In this paper we proposed a mobile gateway-based architecture for integrating MANETs with the Internet. The effect of gateway mobility on performance was reduced through multiple route maintenance and alternative gateways. The performance of the proposed architecture and gateway discovery mechanisms was evaluated using simulation. The simulation results confirm that the proposed architecture has relatively lower overhead and can be used to extend network coverage and support load balancing mechanisms. To reduce the communication overhead and improve the overall network performance, determining a suitable hop length for propagating agent advertisement is necessary. This depends on parameters such as network size, number of gateways and advertisement interval. Future research work is being carried out to develop and evaluate mobility management and load balancing schemes in the integrated network environment.

[11] [12]

[13]

H. Ammari and H. El-Rewini, Design and Performance Evaluation of a Novel Architecture to the Integration of the Internet and Mobile Ad Hoc Networks, proceedings of International Conference on Wireless Networks, June 2004. R. Wakikawa; et al. Global Connectivity for IPV6 Mobile Ad Hoc Networks, IETF Internet draft, Oct 2003. S. McCanne and S.Floyd. Network Simulator. http://www.mash.cs.berkeley.edu/ns/. P. Ratanchandani and R. Kravet,. A Hybrid Approach to Internet Connectivity for Mobile Ad Hoc Networks, proceedings of IEEE WCNC03, New Orleans, Louisiana, USA, March, 2003. M. Ghassemian, et al, "PerformanceAnalysis of Internet Gateway Discovery Protocols in Ad Hoc Networks, WCNC04, 2004. C.E. Perkins, et. al, "Ad Hoc On Demand Distance Vector (AODV) Routing", IETF Internet draft, draftperkins-manet-aodvbis-00.txt, Oct 2003 (Work in Progress). C. E. Perkins, "Mobile IP: Design Principles and Practices", Addison-Wesley: Reading, MA, 1997.

Reference
[1] [2] H. Lei, C.E. Perkins, Ad Hoc Networking with Mobile IP, proceedings of the (EPMCC'97), Bonn, September. 1997. U.Jonsson, et al, MIPMANET-Mobile IP for Mobile Ad Hoc Networks, IEEE/ACM annual Workshop on Mobile Ad Hoc Networking and Computing (MobiHOC02), Boston, Massachusetts, USA, August 2000. J. Broch, D. Maltz, and D. Johnson, Supporting Hierarchy and Heterogeneous Interfaces in Multi-Hop Wireless Ad Hoc Networks, proceedings of IEEE Workshop on Mobile Computing, June 1999. M.Belding-Royer, Y.Sun, and C.E.Perkins, Global Connectivity for Ipv4 Mobile Ad Hoc Networks, IETF Internet Draft, draft-royer-manet-globalv4-00.txt, Work-in-progress, November 2001. Y.Sun, E.M.Belding-Royer, and C.E.Perkins, Internet Connectivity for Ad Hoc Mobile Network International Journal of Wireless Information Networks, vol.9, April 2002. Y.-C. Tseng, et. al, Mobile IP and Ad Hoc Networks: An Integration and Implementation Experience, IEEE Computer, 36(5), May 2003, pp. 48--55.

[3]

[4]

[5]

[6]

1102

You might also like