Path Load Balanced-Fuzzy Logic Based Adaptive ...

Path Load Balanced-Fuzzy Logic Based Adaptive Gateway Discovery in Integrated Internet-MANET Rafi U Zaman

Khaleel Ur Rahman Khan

A. Venugopal Reddy

CSED, Muffakham Jah College of Engineering and Technology Hyderabad, India [email protected]

CSED, Ace College of Engineering Hyderabad, India [email protected]

CSED, University College of Engineering, Osmania University Hyderabad, India [email protected]

Abstract— In this paper, we propose a modified form of the hybrid gateway discovery in integrated Internet-MANET. A modified hybrid gateway discovery approach is called an adaptive gateway discovery mechanism. Three issues are considered together in the proposed adaptive gateway discovery: the load balancing in routing of packets, the range, and periodicity of a gateway advertisement message. We address these three issues by making use of existing approaches. Based on simulation studies, we show that our approach gives better results than the existing approach. Keywords- Mobile Ad Hoc Network, Integrated InternetMANET, Adaptive Gateway Discovery, Path Load Balancing, Fuzzy Logic

I.

INTRODUCTION

Integrated Internet-MANET [1] is a heterogeneous network which is formed by the interconnection of the wired Internet and the wireless mobile ad hoc network. Heterogeneous wireless networks are envisioned as an integral component of the future 4G networks [2]. Figure 1 depicts a snapshot of an integrated Internet-MANET architecture.

Figure 1. Architecture of Integrated Internet-MANET.

Mobile nodes in the ad hoc network need to discover and then register with gateways in order to obtain Internet connectivity. There are three different gateway discovery mechanisms defined in the literature [3]. In the proactive gateway discovery approach, the gateway periodically transmits gateway advertisement messages (GW_ADV) containing relevant information using which a mobile node can register with the gateway. Only mobile nodes within the proactive range of the gateway receive GW_ADV message and con obtain Internet connectivity by responding to a GW_ADV message. In the reactive gateway discovery approach, the gateway does not send GW_ADV messages. Whenever a mobile node desires Internet connectivity, it broadcasts a gateway solicitation message (GW_SOL). When the GW_SOL message reaches a gateway, it sends a GW_ADV message to the mobile node, which then registers with the gateway. In the hybrid approach, mobile nodes in a part of the mobile ad hoc network use the proactive approach for gateway discovery and the rest of the nodes outside the proactive range use the reactive approach. In traditional hybrid approach, the proactive area is set statically and never changed. This leads to a very rigid approach wherein the current network conditions are not taken into consideration for adjusting the proactive zone dynamically. Another issue is the periodicity of GW_ADV messages in the proactive zone. Adaptive gateway discovery approaches exist which addresses the above two issues separately. In this paper, we introduce a gateway discovery mechanism which addresses the above two issues. We propose a gateway discovery mechanism called Path Load Balanced-Fuzzy Logic based Adaptive Gateway Discovery algorithm which helps in balancing the load on different paths by directing traffic towards less congested paths and at the same time dynamically adjusting the range in which gateway advertisement messages are broadcast so as not to congest the ad hoc network with unnecessary control messages. To the best of our knowledge, this scheme is the first of its kind among adaptive gateway discovery approaches which incorporates path load balancing between gateways and mobile nodes, and adaptive gateway discovery. We propose a gateway discovery mechanism which adjusts the proactive zone dynamically by using an average gateway hop metric, a metric similar to the one first used in [4]. We propose a modified form of the fuzzy mechanism of [5] to dynamically adjust the periodicity of the GW_ADV messages.

The rest of the paper is organized as follows: In section II, we present the state of the art in adaptive gateway discovery mechanisms and load balancing approaches in Integrated Internet-MANET, and at the same time develop our motivation for the proposed approach. In section III, we present the proposed Path Load Balanced-Fuzzy Logic based Adaptive Gateway Discovery mechanism. In section IV, we present the simulation results. Finally, section V concludes the paper. II.

RELATED WORK

Various adaptive gateway discovery mechanisms exist which dynamically adjust the advertisement zone, or advertisement interval. Traditional Integrated Internet-MANET strategies [1] have focused on only the hop count metric while making a decision about gateway selection, that is, the gateway which is minimum number of hops away from a mobile node is selected as the default gateway. In few strategies like the proposal of Rakesh Kumar et al [6], a mobile node’s neighbor queue occupancy is taken into account apart from hop count for making the decision about routing. In [7] an optimization to Rakesh Kumar et al’s approach was proposed by including the packet load along a route in addition to neighbor queue occupancy and hop count. But in both the above two approaches, the proactive approach is used. The proactive approach periodically sends agent advertisements, the GW_ADV messages, which traverse a fixed area of the ad hoc network (called advertisement zone) without regard to the current load in the network. If a source moves out of the advertisement zone, then it instantly loses its internet connectivity. This is a drawback of the above two approaches. Apart from the above two approaches, several gateway load balancing strategies [8] exist which use various metrics to balance the gateway load without adapting their gateway discovery mechanisms. That is, they rigidly either use proactive, reactive or hybrid approaches without dynamically adjusting either the advertisement zone or the advertisement periodicity. Their drawback is the same as that of the two previously discussed strategies. The maximal source coverage [9] mechanism is the most well known adaptive gateway discovery approach. In this approach, the gateway dynamically adjusts the proactive zone by setting the TTL value of the next GW_ADV to the number of hops of the farthest mobile node source which is using this gateway. Periodicity of GW_ADV is static. Usman Javaid et al [10] proposed an adaptive distributed gateway discovery mechanism which redirects gateway advertisement messages towards mobile source wherever they are in the ad hoc network. Again, periodicity is set statically. In A.J.Yuste et al’s approach [5], a fuzzy system is used which predicts the advertisement interval periodicity based on three metrics, viz number of received MRS, link changes, and TTL changes. In [11], a prediction algorithm is used wherein the gateway predicts the number of GW_ADV messages that will be required in the next cycle based on network load in the current cycle. In [5] and [11] for determining the proactive zone, the maximal source coverage algorithm is used. In [12], the proactive zone is

dynamically adjusted based on the current load in the network. GW_ADV periodicity is not adjusted. An alternative to the maximal source coverage to determine the proactive zone was presented in [4]. From the above discussion, we observe that the work so far in adaptive gateway discovery mechanism has focused on either proactive zone adjustment or GW_ADV periodicity. Some of the authors who have proposed adjustments to both proactive zone and GW_ADV periodicity have used the maximal source coverage algorithm for proactive zone adjustment. In the next section, we propose our gateway discovery approach which uses an alternative metric to the maximal source coverage to adjust the proactive zone. This metric is a modified form of the metric discussed first introduced in [4]. Our approach also uses a modified form of the fuzzy logic based approach of [5], to adjust the GW_ADV message periodicity. III. PATH LOAD BALANCED-FUZZY LOGIC BASED ADAPTIVE GATEWAY DISCOVERY MECHANISM In the proposed approach, the path load balancing mechanism of [7] is used to route packets along less congested paths. Along with path load balancing, the proactive zone and the TTL periodicity both are adjusted. The proactive zone in a hybrid or adaptive approach is defined by the Time-To-Live (TTL) field value of the GW_ADV message. In the following three subsections, we briefly describe the three components of our adaptive gateway discovery mechanism. A. Path Load Balancing in Integrated Internet-MANET This mechanism was presented in [6]. The main problem that this gateway discovery algorithm addresses is that of balancing packet load across different paths between mobile nodes and gateway. The proactive approach of gateway discovery is used. Gateways periodically send agent advertisement messages. Mobile nodes which are within the transmission range of the gateway are said to be one hop away from the gateway. For how many hops these advertisements are forwarded, depends on the advertisement zone, which was set to 3. Nodes which are more than three hops away cannot access gateway connectivity. This is another drawback of proactive approach, which our proposed approach (or any hybrid approach) overcomes. In this algorithm, the minimum number of hops and the least load path to find a route between the mobile nodes and to select the gateway is used. Basically, the working is: when a selection is to be made between two routes, the one with the lesser path load is selected. Various metrics like gateway queue occupancy, path load and route queue length metric are used to calculate the load along a route, based on which an informed decision is made regarding route selection. B. Adjustment of the TTL value Adjustment of the TTL value results in the proactive zone adjustment. The proactive zone adjustment is according

to the following metric, which is a shortened version of the metric used in [4]:

∆

(1)

Where, TTLn+1 is the TTL value to be used in the next cycle, TTLn is the TTL value used in the last gateway advertisement cycle and ∆TTLSource is the difference between the averages of the distances in hops of the active sources from their respective sources. Here, active source means the mobile node which is using Internet connectivity to communicate with a fixed node. ∆TTLSource is given as:

∆

∑

∑

and TTL changes (TTLC) are the same as in [5]. We present a brief overview of these three metrics: • Number of received GW_SOL messages (NMRG): It is the ratio of the number of GW_SOL messages that the sources generate and the number of active sources. NMRS is represented as: NMRG •

(2)

Where Hopsi is the number of hops active source i is away from a gateway, Nn is the number of nodes registered with a gateway in the nth cycle, Nn+1 is the number of nodes registered with a gateway in the n+1th cycle. The working of dynamic TTL adjustment is depicted in figure 2. Mobile nodes MN1, MN2, MN3 and MN4 are registered with Internet gateway IGW1, whereas MN5, MN6, MN7, MN8 and MN9 are registered with Internet gateway IGW2. The active sources of IGW1 are MN2 and MN4, which are currently at a distance of 2 and 1 hop respectively. For IGW2, the active sources are MN9 and MN7, which are at a distance of 2 and 3 hops respectively. IGW1 and IGW2 calculate the TTL to be set in their respective next GW_ADV by using eq. (1).

LC •

(4)

TTL changes (TTLC): This metric represents the ratio of number of changes in the distance of the active sources to the number of active sources. It is represented as: TTLC

TTL

(5)

The values obtained from equations 3, 4 and 5 are fuzzified according to the rules given in [5] and the output is the periodicity of the GW_ADV messages in the next cycle. Our approach to determining the periodicity of GW_ADV message differs from [5]. Instead of determining the convenience of sending a GW_ADV message, we directly convert convenience into GW_ADV interval, which denotes the periodicity of the GW_ADV message. In the original proposal [5], the authors give five values to the convenience output viz. very low, low, moderate, high and very high. We map these convenience values to GW_ADV interval using the following table: GATEWAY ADVERTISEMENT INTERVAL VALUES Convenience Very Low Low Moderate High Very High

C. Adjustment of GW_ADV Periodicity To adjust the periodicity of the gateway advertisement messages, we make use of the fuzzy logic based approach discussed in [5]. In this approach, three metrics NMRS, LC and TTLC are used to determine the convenience of transmitting a GW_ADV message. We have used a modified version of NMRS metric which we call Number of received GW_SOL messages (NMRG). Link changes (LC)

(3)

Link Changes (LC): This metric is used to measure the mobility of nodes near the gateway. It represents the ratio of number of link changes a gateway detects and the number of active sources. LC is represented as:

TABLE I.

Figure 2. Adjustment of TTL value dynamically.

GW_SOL

GW_ADV Interval 2 3 4 5 6

We summarize the procedure that a gateway follows in our proposal to adjust TTL value and GW_ADV periodicity in the following algorithm: D. Algortihm for Path Load Balanced Fuzzy-Logic based Adaptive Gateway Discovery Step 0: Use the path load balancing mechanism for routing packets within the MANET. Step 1: Initialize the first TTL value. Let us call it TTL0. Step 2: Calculate the TTL value to be used in the next cycle, TTLn+1, using equation (1). Step 3: Calculate the gateway advertisement interval for

the next cycle, using table 1. Step 4: Repeat steps 2 and 3. IV.

PERFORMANCE EVALUATION

The proposed gateway discovery mechanism is implemented in ns-2.33 [13]. Its performance is compared with the maximal source coverage adaptive gateway discovery approach [6]. For the integrated Internet-MANET framework, we use the AODV+ routing protocol [14].

better packet delivery ratio while incurring lesser routing overhead. On the other hand, the end to end delay is slightly worse, as can be seen from figure 5. This is due to the path load balancing mechanism, which results in slight delay in the packet delivery.

A. Simulation Setup We have simulated two scenarios, one with 15 nodes and the other with 25 nodes. The common parameters for both these simulation scenarios are given in table 2. For each x value in the following figures, ten simulation runs were performed and their average taken. TABLE II.

SIMULATION PARAMETERS

Simulation Parameter Scenario

Number of Mobile Nodes Number of gateways Toplogy Mobile node radio range Simulation time Number of traffic sources Traffic Type Mobility Model Node Speed Number of destination nodes Pause Time Ad Hoc Routing Protocol

Value 1

2

25 15 3 2 1000X1000 800X500 250m 900 sec 5 CBR Random Waypoint 5-25 Mts/Sec 2 5 seconds AODV+

B. Performance Metrics In order to ascertain the performance of the proposed mechanism, we use the following metrics: Packet Delivery Ratio: The ratio of the total packets received to the number of packets sent. End-to-End Delay: The average delay that a packet undergoes while traversing from a source node to a destination node. Normalized Routing Load: The total number of control packets generated for every data packet delivered at the destination. C. Results Discussion As mentioned earlier, we have simulated two scenarios. The first scenario, which we call scenario – 1, consists of 25 mobile nodes whereas the second scenario, scenario – 2, consists of 15 mobile nodes. Figures 3 to 5 show the results comparison of the existing maximal source coverage mechanism and the proposed mechanism for scenario – 1. From figure 3, we observe that the proposed mechanism gives the better packet delivery ratio for different mobile node speeds ranging from 5 mts/sec to 25 mts/sec. The proposed mechanism provides the better packet delivery ratio while incurring lesser routing overhead, as can be seen from figure 4. Thus, the proposed mechanism provides

Figure 3. Packet Delivery Ratio Vs Node Speed

Figure 4. Normalized Routing Load Vs Node Speed

Figure 5. End to End Delay Vs Node Speed

significantly lesser routing overhead. The end-to-end delay is almost the same for both the mechanisms, from figure 8. V.

Figure 6. Packet Delivery Ratio Vs Node Speed

CONCLUSION AND FUTURE WORK

In this paper, we proposed a path load balanced fuzzy logic based adaptive gateway discovery mechanism which addresses three issues in gateway discovery in integrated Internet-MANET, namely, load balanced routing, proactive zone adjustment and gateway advertisement periodicity adjustment, dynamically. We briefly discussed the existing work by different authors. Based on the simulation results, we can conclude that our proposal gives improved packet delivery ratio and excellent improvements in normalized routing overhead. In the future, we propose to test our mechanism for higher traffic loads in dense scenarios, where we expect the path load balancing and adaptive gateway discovery mechanism to perform significantly better than existing approaches. REFERENCES [1]

Figure 7. Normalized Routing Load Vs Node Speed

Figure 8. End to End Delay Vs Node Speed

Figures 6 to 8 show the results comparison of the existing maximal source coverage mechanism and the proposed mechanism for scenario – 2. From figure 6, we observe that the proposed mechanism gives the better packet delivery ratio for node speeds 5, 15 and 20 mts/sec. In this scenario, the routing overhead incurred by the proposed mechanism is much lower than the existing approach, as can be seen from figure 7. Thus, we can say that, the proposed mechanism overall provides better packet delivery ratio while incurring

Khaleel Ur Rahman Khan, Rafi U Zaman, A.Venugopal Reddy, “Integrating Mobile Ad Hoc Networks and the Internet: challenges and a review of strategies”, Communication Systems Software and Middleware and Workshops, COMSWARE 2008. 3rd International Conference, 536 –543, (2008). [2] Dave Cavalcanti, Carlos Cordeiro, Dharma P. Agrawal, B. Xie, and Anup Kumar, “Issues in Integrating Cellular Networks, WLANs, and MANETs: A Futuristic Heterogeneous Wireless Network,” IEEE Wireless Communications Magazine, Special issue on Toward Seamless Internetworking of Wireless LAN and Cellular Networks, Vol. 12, No. 3, 30-41, (June 2005) [3] A. Hamidian, U. Korner, and A. Nilsson, "Performance of Internet Access Solutions in Mobile Ad Hoc Networks", Springer's Lecture Notes in Computer Science (LNCS), pp 189-201 ( 2004). [4] Fang Xie, Lei Du, Yong Bai, Lan Chen, “Adaptive Gateway Discovery Scheme for Mobile Ubiquitous Networks”, WCNC 2008, 2916-2020, (2008). [5] A.J. Yuste, Alicia Trivino, F.D. Trujillo, E. Casilari: “Using Fuzzy Logic in Hybrid Multihop Wireless Networks”. International Journal of Wireless & Mobile Networks Volume 2, Issue 3, 96-108, (2010). [6] Kumar, R., Misra, M. and Sarje, A.K., “An Efficient Gateway Discovery in Ad Hoc Networks for Internet Connectivity”, Proc. of the International Conference on Computational Intelligence and Multimedia Applications, pp 275-281, 2007 [7] Khan K. U. R., Reddy A. V., Zaman R. U., Kumar M. “An Effective Gateway Discovery Mechanism in an Integrated Internet-MANET (IIM),” Proc. of the International Conference on Advances in Computer Engineering, India, pp. 24-28, June 2010 [8] Rafi U Zaman, Khaleel Ur Rahman Khan and A. Venugopal Reddy, “A Review of Gateway Load Balancing Strategies in Integrated Internet-MANET”, International Workshop on Advances in Peer-Peer Technology IWAP2PT’09 co-located with IMSAA-09, IIIT Bangalore, 9-11 December 2009. Proceedings in IEEE Digital Library, pages: 141 – 146, 2009 [9] Pedro M. Ruiz, Antonio F. Gomez-Skarmeta: “Maximal Source Coverage Adaptive Gateway Discovery for Hybrid Ad Hoc Networks”, Lecture Notes in Computer Science, vol.3158, 28-41, (2004) . [10] Javaid, U., Rasheed, T.M., Meddour, D., Ahmed, T.: “Adaptive Distributed Gateway Discovery in Hybrid Wireless Networks”. WCNC – 2008, 2735-2740. (2008)

[11] Triviño-Cabrera, A., Ruiz-Villalobos, B., Casilari, E., Yuste-Delgado, A.J.: “Study on the need for adaptive gateway discovery in MANETs”, IWCMC – 2009, 1091-1095, (2009). [12] Bok-Nyong Park, Wonjun Lee, Choonwa Lee, “QoS-aware Internet access schemesfor wireless ad hoc networks”, Computer Communications 30 (2007), 369-384, (2007). [13] Ns 2 Home page : http://www.isi.edu/nsnam/ns/ index.html

[14] Hamidian A., “A Study of Internet Connectivity for Mobile Ad Hoc Networks in NS2”, Masters Thesis, Department of Communication Systems, Lund Institute of Technology, Lund University, January 2003