An Efficient and Minimum Cost Topology Construction for Rural Wireless Mesh Networks
IJCSIS is an open access publishing venue for research in general computer science and information security. Target Audience: IT academics, university IT faculties; industry IT departments; government departments; the mobile industry and computing industry. Coverage includes: security infrastructures, network security: Internet security, content protection, cryptography, steganography and formal methods in information security; computer science, computer applications, multimedia systems, software, information systems, intelligent systems, web services, data mining, wireless communication, networking and technologies, innovation technology and management. The average paper acceptance rate for IJCSIS issues is kept at 25-30% with an aim to provide selective research work of quality in the areas of computer science and engineering. Thanks for your contributions in September 2010 issue and we are grateful to the experienced team of reviewers for providing valuable comments.
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, September 2010 An Efficient and Minimum Cost Topology Construction for Rural Wireless Mesh Networks Prof. V. Anuratha & Dr. P. Sivaprakasam Abstract is rural and are on the wrong side of the digital divide. Many research efforts as well as deployments have chosen IEEE802.11 as a Bridging this divide necessitates, providing low-cost, long-distance access technology to internet connectivity to each and every bridge the digital divide. IEEE 802.11 Wi-Fi village. Providing the same by expanding equipment based wireless mesh networks the current telephone network to rural areas have recently been proposed as an is infeasible because of the huge initial inexpensive approach to connect far-flung infrastructure costs. Also, deployment of rural areas. To establish such network high- cellular wireless would not be sustainable gain directional antennas are used to achieve because of its business model, which long-distance wireless point-to-point links. demands more high-paying consumer Some nodes in the network are called density. gateway nodes and are directly connected to the wired internet, and the remaining nodes Emerging technologies like 802.16 connect to the gateway(s) using one or more WMAN,, have not yet reached the hops. scale of competitive mass production, hence the equipments are expensive. In this regard, In this paper the cost of constructing the the 802.11 Wi-Fi has shown tremendous antenna towers required is investigated. The growth and acceptance as a last hop access problem is NP hard is shown and that a solution, because of their low price. better than O(log n) approximation cannot Although 802.11 was primarily designed for be expected, where n is the number of indoor operation, but  has established the vertices in the graph. To minimize the possibility of using 802.11 in long-distance construction cost a new algorithm is networking. proposed called constant time approximation algorithm. The diverse requirements are in provisions of 1) Communication pattern which deals The results of proposed approximation with the mode of communication one-to- algorithm are compared with both the one, one to- many, many-to-one, and many- optimal solution, and a naive heuristic. to-many, 2) Delay (real-time, non-real-time, and delay-tolerant), 3) Service availability INTRODUCTION (centralized, distributed, and location-aware) that deals with the awareness of the There has been a huge proliferation of availability of different services, such as Internet and other communication based Internet access, real-time communications, services in the last two decades. However, content distribution, interactive gaming, this spread is confined to developed medical applications, and vehicular safety countries, and metropolitan pockets of applications. 4) Security and 5) Reliability. developing countries. This is really An essential requirement to establish long- unfortunate for developing countries like distance links is that line-of-sight is India, where around 74% of the population maintained between the radio antennas at the end-points. 202 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, September 2010 To ensure line-of-sight across such long on the most critical design issues in distances would require the antennas to be multihop wireless networks. Topology mounted on tall towers. The required height control has been investigated extensively in of the towers depends both on the length of the literature. Nevertheless, it is noted that the link, and the height of the obstructions most existing studies do not consider the along the link. requirements on upper layer applications or services. In this article the author address the The cost of the tower depends on its topology control issues on service-oriented height and the type of material used. For wireless mesh networks. In particular, the relatively short heights (10- 20 meters) author provides a comprehensive survey of antenna masts are sufficient. For greater existing works on topology control from a heights, sturdier and much more expensive service-oriented perspective. A general antenna towers are required. In this paper, framework for topology control in service- several important contributions are made oriented WMNs is proposed. To towards developing efficient algorithms  demonstrate the effectiveness of the to solve this problem. First, the requirements framework, a case study is conducted in to establish a point-to-point 802.11 link which the main objective is to maximize the between two nodes of a given network graph overall throughput in a network with random is described. Then the formal definition of unicast traffic. The performance of this the Topology Construction problem (denote topology control scheme is evaluated by by TC) is given. It’s proved that the problem numerical results. In addition, it is illustrated to be NP hard by a reduction from the set- that the generated topology can support cover problem.The approximation algorithm advanced technologies, including network is presented for this NP hard problem and coding and physical-layer network coding, the establishment cost of the tower in rural which can significantly improve the areas using constant time approximation throughput capacity of a network. The cost algorithm is presented. The rest of this paper of laying wire to rural areas is prohibitively is organized as the following. Section II expensive. gives the Related Works of this technique is presented. In Section III, the Methodology Also, traditional wireless technologies such of proposed approach is given. Section IV as cellular data networks (e.g., EV-DO) and has the Experiment Results and this paper is upcoming technologies like IEEE 802.16 concluded in Section V. WiMAX have prohibitively expensive equipment costs. As a result, there has been RELATED WORKS considerable recent interest , ,  in the design of rural mesh networks using 802.11-based long-distance networks have IEEE 802.11 (Wi-Fi) equipment. The cost of been proposed as a cost-effective option to an 802.11 radio (»$50/PCMCIA card) is provide Internet connectivity to rural areas orders of magnitude less than that of in developing regions, to enable Information cellular/WiMAX base stations. Thus, this and Communication Technology (ICT) approach is an attractive option for building services . low cost networks. D. S. Lun et.al, Rural areas (especially in developing presented a distributed random linear regions) have populations with very low network coding approach for transmission paying capacities. Hence, a major factor in and compression of information in general network deployment is the cost of the multisource multicast networks. infrastructure and the network equipment. In Network nodes independently and randomly this context, efficient algorithms are select linear mappings from inputs onto investigated for the minimum cost topology output links over some field. The author construction problem in rural wireless mesh shows that this achieves capacity with networks. R. Ramanathan et.al, discussed 203 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, September 2010 probability exponentially approaching 1 determined by an elliptical area between u with the code length. Random linear coding and v termed the first Fresnel zone. To is demonstrated which performs establish RF line-of-sight, a significant area compression when necessary in a network, of the Fresnel zone (> 60% of the radius of generalizing error exponents for linear the Fresnel zone at the location of the Slepian-Wolf coding in a natural way. obstruction ) should also clear all Benefits of this approach are decentralized obstructions between u and v. However, this operation and robustness to network changes can be simply modeled by extending the or link failures. The author shows that this height of the obstruction to include the approach can take advantage of redundant radius of the Fresnel zone that has to be in network capacity for improved success the clear. probability and robustness. Some potential advantages of random linear network coding are illustrated over routing in two examples of practical scenarios: distributed network operation and networks with dynamically varying connections. The derivation result also yields a new bound on required field size for centralized network coding on general multicast networks. METHODOLOGY In this paper a Novel Topology Control Scheme is used to identify a set of semi-permanent highways, such that the best Figure1: Computing the height of towers throughput capacity of the network can be at the end-points of a link obtained. Particularly, the wireless highways In reality, there can be multiple are predicted to be rather similar to the obstructions between u and v. As in Figure highway system in public transportation 4b, consider multiple obstructions, O1, O2, system, which can efficiently provide …,Ok between u and v. Now, let h(u) and connectivity in real application. h(v) represent the tower heights at the nodes A. Computing tower heights at the end- of u and v. Covering edge (u,v) requires a points of a link visual and RF line-of-sight connection Consider two nodes, u and v that are between the towers at its two terminal separated by a distance luv. The edge (u; v) nodes. This would imply that the straight is considered to be covered if an 802.11 line fuv joining the top of the two towers (of based point-to-point communication link can heights h(u) at u and h(v) at v) should clear be established between u and v. Assume that every obstruction in (u,v). Hence, it is noted the transmit powers  and the gains of the that given a particular pair of tower heights antennas at both ends are sufficient to over at u and v, deciding whether these heights come the free-space path loss between the covers edge (u,v) can be done in time linear two points. The first basic requirement to in the number of obstructions on that edge.2 cover the edge between u and v is that there B. Modeling tower costs is a clear visual line-of-sight between the An important component in this problem is antennas at the end-points (as shown in Fig. the nature of the cost function that maps 4a). In other words, the line joining the tower heights to the cost of building the antennas mounted on the towers should clear tower. any obstructions along the path. Secondly, it is also required that RF line-of-sight is maintained between the two points. This is 204 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, September 2010 There are two types of antenna towers that of the above properties. In the remainder of are used. For heights less than 20 meters, this paper, when the height function h is one can use the cheaper masts. For greater unambiguous, often the cost of the tower is heights, the more expensive steel towers is denoted at a node v as c(v) rather than used. c(h(v)). Further, there is an order of magnitude C. The Topology Control Scheme difference between the cost of the cheaper In this framework, it is considered that masts and that of the steel towers. Thus, the highways can be partitioned into two roughly speaking, the cost function is groups, horizontal and vertical. Highways in constant as long as the cheaper masts can be each group can operate simultaneously used and becomes linear in height once the because they are mutually parallel and can steel towers are needed, with a jump in cost be placed away enough to reduce when we switch from masts to steel towers. interference below a certain threshold. Consequently, horizontal and vertical Let us denote the height at which the highways will partition the whole material of the tower has to be switched as geographical area into grids, in which nodes hmin. will try to forward their traffic to the nodes on neighboring highways. The combination Further, there is a physical restriction on the of the following parameters can be maximum possible height of a tower, considered for the Topology Control denoted by hmax. Thus, the cost function c Scheme can be formally defined as 1) Transmission range: Transmission range of each node in the network is traditionally an important Where A, B and K are constants and design parameter in topology Ahmin + B >> K. Although, in practice, the control. In general, a smaller cost function can be modeled as discussed transmission range will improve above, our algorithm works with a much the channel reuse but may more general cost functions. Specifically, compromise the connectivity. A we only require the cost function c to satisfy larger transmission range will the following two natural properties C1 and improve the connectivity but C2. C1 reduce the channel reuse. Therefore, an appropriate range is Given the tower costs at two neighboring chosen as a trade-off between nodes u and v, it can be determined (in connectivity and channel reuse. polynomial time) whether the corresponding 2) Type of antenna: When directional tower heights cover the edge (u,v). This antenna or beam forming is used it simply requires that the corresponding tower may improve the capacity of the height can be computed (in polynomial network by reducing the time) given the tower cost. As mentioned interference and improve the earlier, determining whether the height of transmission quality. the towers is sufficient to cover an edge can 3) Traffic pattern: Traffic pattern is be done in polynomial time.C2 the cost very important parameter to the function is monotonically increasing with topology. In most studies height, i.e., h1≥ h2 and c(h1) ≥ c(h2) for previously done suggest that the any values of h1 and h2. traffic is broadcast. With such an assumption, the problem is It is easy to verify that the cost function formulated in a way such that the c defined earlier in this section satisfies both overall transmission for each 205 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, September 2010 message is minimized. However, For the first step of this study, only broadcast traffic may only be a omnidirectional antenna and purely special case in the future service- random unicast traffic pattern are oriented WMN, in which a variety considered. Moreover, the random of patterns may appear, from one- wireless network elaborated to gain to one to many-to-many . insights for the future investigation. 4) Quality of service (QoS): To achieve an efficient network a crucial issue is to enable services with certain QoS requirements, such as bandwidth, delay, security and reliability. Figure 2: Approximation algorithm for NP-hard problem 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. Mark all unlabeled vertices of with label else (i.e., 16. 17. 18. 19. 206 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 20. (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, September 2010 EXPERIMENTAL RESULTS Given a set of links to be covered, the height assignment problem is formulated as In this section, extensive numerical a simple LP, and compute the heights simulations are carried out to evaluate our required on every node. approximation algorithm with the optimal solution and also a naive heuristic. For this simulations, synthetic topologies are generated that aim to match the geographical C. Comparing with the naive heuristic structure of village clusters. Simulation setup is described in more detail as follows. Now the naive heuristic described above is A. Generating synthetic graph compared with the proposed approximation topologies. algorithm. Graphs considered with number of nodes n =10, 15, 20,…, 50. For each value of n we generate 50 graph instances. A circular plane is considered with a For each graph, Cnaive the cost of the radius of 25Kms. Nodes are placed at solution produced by the naive heuristic, and random locations on this plane. A link (u,v) Capprox cost of the approximation is considered between any two nodes, u and algorithm is computed. With this value it is v, and for these simulations, assume just one observed that the proposed approximation obstacle, ouv, located on the middle of this algorithm performs substantially better than link. the naive heuristic. The height of the obstruction (ho) is On average, the solutions returned by the selected randomly with a maximum value of naive heuristic range from 60% (for n = 10) 20 meters - the typical height of trees and to as much as 225% more expensive (for n = small houses in a rural setting. A weight 50) compared to the solution returned by the wuv is assigned to the link equal to twice the naive algorithm. effective height of the obstruction on this link. As described earlier, the effective D. Comparing with the optimal solution height of an obstruction, is the sum of the physical height (ho) and 60% of rf , the The optimal solution is computed by radius of the fresnel zone. solving an ILP that models the topology construction problem. The CPLEX LP- solver  to solve this ILP. This approach B. Naive heuristic is, however, computationally very expensive, and the LP-solver could return solutions for graphs with at most 11 nodes. In order to compare with the proposed approximation algorithms, a naive heuristic The solution returned by our is described for selecting connected approximation algorithm is compared with subgraphs and assigning heights to the the optimal solution for graphs with number nodes. As a first step, to select a connected of nodes n = 8, 9, 10, 11. For subgraph of an input graph G, the heuristic each value of n we generate 50 graphs. For computes the minimum spanning tree each graph Capprox, the cost of the solution (MST), T of G (using the link weights returned by proposed approximation computed as described above). Next, the algorithm, and Copt, the cost of the optimal heuristic has to assign heights to the nodes solution. in G, so as to cover all the edges in T while minimizing the total cost. 207 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, September 2010 Then the mean and standard deviation is open research problems in the topology computed over all graphs for different construction. values of n . One immediate problem is to consider the case of k ≥ 2 vertex or edge n Mean (std. dev.) of connectivity, similar to the power optimal Ropt network construction for k-connectivity. 8 0.58 (0.30) Another important research direction is the 9 0.57 (0.25) geometric version of this problem. In practice, all nodes within a certain distance 10 0.55 (0.23) of each other can establish a link. 11 0.52 (0.25) In this paper, the location of the Table1: New Approximation algorithm towers is assumed to be fixed (within a vs. Optimal Solution. village). A variant of the problem would make the location of the tower to be a The results presented in Table 7 show variable. that the incorporated approximation algorithm gives solutions that are 50 - 60% This method has added flexibility more expensive than the optimal solution than the previous method which would (for small values of n). Thus, the new result in reduced cost. The numerical approximation algorithm performs much experiments demonstrate that the proposed better than the worst case guarantee of O(log method with constant time approximation n) on the approximation factor. algorithm performs well within its worst case performance bounds, and outperforms While this gap between constant time the naive heuristic by a substantial margin. approximation algorithm and the optimal solution is not small, its expected computing the optimal solution (even if it has to be done only once) is practically infeasible for References real-life networks. Moreover, this approximation algorithm performs  I. Akyildiz and X. Wang, “A survey on substantially better in practice than the naive wireless mesh networks,” IEEE heuristic and previous approximation Communications Magazine, vol. 43, no. algorithms 9, pp. S23–S30, Sept. 2005.  M. Lee, J. Zheng, Y.-B. Ko, and D. CONCLUSION Shrestha, “Emerging standards for wireless mesh technology,” IEEE In this paper an overview of Wireless Communications, vol. 13, no. establishing a low cost wireless mesh 2, pp. 56–63, April 2006. network for rural areas is presented. In rural  Pravin Bhagwat, Bhaskaran Raman, and areas nodes are connected using long Dheeraj Sanghi,” Turning 802.11 Inside- distance 802.11 wireless links which are Out”, In HotNets-II, Nov 2003. established using high-gain directional  R. Ramanathan and R. Rosales-Hain, antennae. The main problem is with the “Topology control of multihop wireless topology construction for long distance networks using transmit wireless communication. poweradjustment,” vol. 2, 2000.  Eric Brewer, Michael Demmer, Bowei An efficient approximation Du, Kevin Fall, Melissa Ho, Matthew algorithm is proposed for the topology Kam, Sergiu Nedevschi, Joyojeet Pal, construction problem in rural mesh Rabin Patra, and Sonesh Surana. The networks. This work introduces a number of Case for Technology for Developing 208 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 6, September 2010 Regions. IEEE Computer, 38(6):25–38, AUTHOR BIOGRAPHIES June 2005. Mrs. V. Anuradha graduated with B.Sc  P. Dutta, S. Jaiswal, K. Naidu, D. Computer Science in the year 1995 and Panigrahi, R. Rastogi, and A. Todimala. completed M.C.A at Madras University in Villagenet: A low-cost, 802.11-based the Year 2000. Completed her M.Phil in the mesh network for rural regions. In year 2003 and also got Guide Approval for WIreless Systems: Advanced Research M.Phil in Bharathiar University, Peiyar and Development Workshop University, Bharathiar University and (WISARD), 2007. currently doing her Ph.d.Hear area of  R. Patra, S. Nedevschi, S. Surana, A. interest is Networks and data mining. Mrs. Sheth, L. Subramanian, and E. Brewer. V. Anuratha have guided 20 M.Phil scholors WiLDNet: Design and Implementation and she have participated and presented of high performance wifi based long many papers in the national and distance networks. In NSDI, 2007. international conferences and etc..  B. Raman and K. Chebrolu. Design and evaluation of a new MAC for long Currently she is working as a H.O.D – PG distance 802.11 mesh networks. In Department of Computer Science at Sree Mobicom, 2005. Saraswathi Thyagaraja College with a  D. S. Lun, N. Ratnakar, M. Medard, R. decade a teaching experience Koetter, D. R. Karger, T. Ho, E. Ahmed, and F. Zhao, “Minimum-cost multicast Co author biography……. over coded packet networks,” IEEE Dr. P. Sivaprakasam : Transaction on Information Theory, vol. 52, no. 6, pp. 2608–2623, June 2006. He have completed his M.Sc(c.s) in the year  K. Lu, Y. Qian, H.-H. Chen, and S. 1986, M.Phil in the year 1995 and Ph.d in Fu, “WiMAX Networks: From Access 2005 on the topic “An Analysis of Web To Service Platform,” IEEE Network, performance and caching”. He have 19 years 2008, accepted. of teaching experience. He have published 6  ilogcplex. papers at national level and 3 at international http://www.ilog.com/products/cplex/. level. He was also sanctioned with 2 UGC  M. T. Hajiaghayi, N. Immorlica, and research projects. V. S. Mirrokni. Power optimization in fault-tolerant topology control He is now currently working as a Associate algorithms for wireless multi-hop professor in Computer Science at Sri Vasavi networks. In MOBICOM, 2003. college of Arts and Science. His areas of  M. T. Hajiaghayi, G. Kortsarz, V. S. interest are Internet, Computer Networks, Mirrokni, and Z. Nutov. Power Service Oriented Architecture. optimization for connectivity problems. In IPCO, 2005  Z. Nutov. Approximating minimum power covers of intersecting families and directed connectivity problems. In APPROX-RANDOM, 2006.  S. Sen and B. Raman,”Long distance wireless mesh network planning: Problem formulation and solution”, In WWW, 2007.  P. N. Klein and R. Ravi. A nearly best-possible approximation algorithm for node-weighted steiner trees. J. Algorithms, 19(1), 1995. 209 http://sites.google.com/site/ijcsis/ ISSN 1947-5500