International Journal of Wireless Communications and Networking 3(1), 2011, pp. 21-25 CELL BREATHING FOR LOAD BALANCING IN WIRELESS LAN R. Latha and S. Radhakrishnan Rajalakshmi Engineering College, Thandalam, India E-mail: latha.rajendran1@gmail.com, radhakrishnan_s@hotmail.com Abstract: Wireless Local Area Networks (WLANs) are now commonly available on many academic and corporate Campuses [1]. Load balancing is necessary to improve the network performance. The load may be unevenly distributed across a small number of Access points in the WLAN. In this paper, the load of AP is balanced using load balancing algorithm utilizing Cell breathing technique. Cell breathing is a technique where an AP can adjust its cell boundary by changing the power of AP beacons. The goal is to balance the users by setting a threshold. The throughput and fairness of APs are also calculated. In future, the transmission power of AP data traffic will be controlled. Keywords: cell breathing; load balancing; wlan I. INTRODUCTION The widespread strategic reliance on networking among competitive businesses and the meteoric growth of the Internet and online services are strong testimonies to the benefits of shared data and shared resources. With wireless LANs, users can access shared information without looking for a place to plug in, and network managers can set up or augment networks without installing or moving wires. Wireless LANs offer the following productivity, service, convenience, and cost advantages over traditional wired networks. Recent studies [1], [4] on operational IEEE 802.11 wireless LANs (WLANs) have shown that traffic load is unevenly distributed among the access points (APs). Wireless LAN systems can provide LAN users with access to real-time information anywhere in their organization. This mobility supports productivity and service opportunities not possible with wired networks. Installing a wireless LAN system can be fast and easy and can eliminate the need to pull cable through walls and ceilings. Wireless technology allows the network to go where wire cannot go. Reduced Cost-of-Ownership. While the initial investment required for wireless LAN hardware can be higher than the cost of wired LAN hardware, overall installation expenses and lifecycle costs can be significantly lower. Long-term cost benefits are greatest in dynamic environments requiring frequent moves, adds, and changes. Wireless LAN systems can be configured in a variety of topologies to meet the needs of specific applications and installations. Configurations are easily changed and range from peer-to-peer networks suitable for a small number of users to full infrastructure networks of thousands of users that allows roaming over a broad area. In Infrastructure WLANs, multiple access points link the WLAN to the wired network and allow users to efficiently share network resources [6]. The access points not only provide communication with the wired network but also mediate wireless network traffic in the immediate neighborhood. Multiple access points can provide wireless coverage for an entire building or campus. Wireless communication is limited by how far signals carry for given power output. WLANs use cells, called microcells, similar to the cellular telephone system to extend the range of wireless connectivity. At any point in time, a mobile Personal Figure 1: Infrastructure WLAN
22 International Journal of Wireless Communications and Networking Computer (PC) equipped with a WLAN adapter is associated with a single access point and its microcell, or area of coverage. Individual microcells overlap to allow continuous communication within wired network. II. CELL BREATHING Transmission power of an AP is adjusted to dynamically modify its cell boundaries in order to balance the AP load. When the transmission power is increased, the cell will enlarge when the transmission power is reduced, the boundaries will shrink. Controlling the size of the cell is known as cell-breathing. The increase of the number of active users in a cell causes the increase of the total interference sensed at the base station. Therefore, in congested cells, users need to transmit with higher power to maintain a certain signal-tointerference ratio at the receiving base station. As the users in a congested cell increase their transmission power, they also increase their interference to the neighboring cells since all cells use the same frequency in CDMA networks. III. LOAD BALANCING Users are not evenly distributed, some APs tend to suffer from heavy load, while their adjacent APs may carry only light load. Such load imbalance among APs is undesirable as it hampers the network from fully utilizing its capacity and providing fair services to users. In [5], a mathematical foundation for distributed frequency allocation and user association for efficient resource sharing is provided. In [8] a unified mathematical framework for dynamic load balancing was provided, which leads to closed-form performance expressions for evaluating the performance of some of the most important dynamic load balancing strategies proposed in the literature. Users tend to remain in the same location for long periods. Users are generally associated with the closest AP. In order to avoid the congestion, congested APs reduce the size of their cells; alternatively, underutilized APs increase their cells to attract further stations. A novel load balancing scheme reduces the load of congested APs by forcing the users near the boundaries of congested cells to move to neighboring less congested cells. To overcome this deficiency, various load balancing schemes have been proposed. These methods commonly take the approach of directly controlling the user-ap association by deploying proprietary client software or hardware. For instance, vendors can incorporate certain load balancing features in their device drivers, AP firmware, and WLAN cards. In these proprietary solutions, APs broadcast their load levels to users via modified beacon messages and each user chooses the least-loaded AP. In [2] a parallel Constraint Programming (CP) method to solve Constraint Satisfaction Problems (CSP) was provided. Some attributes induced by the CP model of a CSP to improve the load balancing procedure embedded in the parallel tree-based search algorithm was used. For instance, in [10], a solution was presented which is based on the combination of cell breathing and bandwidth space partitioning. In [3] a distributed load balancing technique was proposed that utilizes a bobble oscillation algorithm. In [7], a method that coordinates the packet level scheduling with cell breathing techniques was proposed. In [9], an algorithm was proposed to reassociate users when the total load exceeds a certain threshold or the bandwidth allocated to users drops below a certain threshold. Generation of nodes and Access Points Figure 2: Identification of AP having larger number of clients and setting a threshold Flow Diagram IV. NETWORK MODEL Balancing the loads by considering the threshold in a particular Access Point Figure 2 Flow diagram From [11] Consider a WLAN with a set of APs, denoted by A. Mod(A) denotes the number of APs. All APs are directly attached to a wired infrastructure. Each AP can support several transmission power levels. Without loss of generality, each AP can use one of K+1 transmission levels, denoted by {P k k [0...k]} (1) where the minimal and maximal levels are denoted by P min = P 0 (2) and P max = P k, (3) respectively. A power level P k is identified by its power index k and its transmission power is γ times stronger than its predecessor P k-1, where k γ = ( P /) P (4) max min
Cell Breathing for Load Balancing in Wireless LAN 23 Because it follows that for every i.e. γ 1. (5) P k = γ P k-1 (6) P k = P min γ k (7) k [0...k] (8) This assumption is consistent with the transmission power level configurations supported by commercial AP products. The transmission power level of an AP a A as the power index of AP a and it is denoted by p a. The network coverage area is defined as the union of the transmission ranges of all APs in A. Initially, the AP deployment ensures a high degree of overlaps between the ranges of adjacent APs, so that every user is covered by at least one AP even when all APs are transmitting at the minimal power level P min. U denotes the set of all users in the network coverage area and mod(u) denotes their number. For optimal algorithms, the users have a quasi-static mobility pattern. In other words, users can move freely, but they tend to stay in the same locations for a long period. User mobility can be handled by online strategy. At any given time, each user is associated with a single AP. When a user enters a WLAN, it scans all channels (i.e. listening for the beacon messages) for identifying all APs in its reach. Then, based on the RSSI s of the beacon messages, the user associates itself with the AP that has the strongest RSSI. The RSSI depends on the transmission power of the beacon messages and the signal attenuation between the AP and user. Generally, the channel quality is time-varying. For the user-ap association decision, a user performs multiple samplings of the channel quality, and only the signal attenuation that results from long-term channel condition changes are utilized. Based on the RSSI, the network coverage area is divided into mod(a) disjoint cells. The transmission bit rate for a user-ap pair is determined by the signal-to-noise Ratio (SNR). Users associated with the same AP may transmit at different bit rates, and each user may contribute a different amount of load to its serving AP. The appropriate channel allocation should be made, so that adjacent cells do not interfere with each other. The load of an AP a, denoted by y a, is the sum of the load contributions of its associated users. The load contribution term of a user u on an AP a is denoted by l a,u. This contribution is constant, so that the load of each AP a A is Ya = U u a l (9) a, u where U a denotes the set of users associated with a. The APs that experience the maximal load are called the congested APs and their load, termed congestion load, is denoted by Y. Other APs with lower load are called noncongested APs. V. RESULTS The simulation software is developed on the NS2 software which runs on the top of Linux. TCL scripts work very well in the NS2 environment for simulating the generation of nodes, generation and controlling of access points, balancing the nodes, adjusting the transmission power dynamically, etc. By applying Min-Max Priority Load Balancing Algorithm [4], the Load of APs becomes balanced. Four APs and 21 nodes of users are generated by Network Simulator. Initially, the load of APs were Unbalanced i.e. AP1 has 5 clients, AP2 has 7 clients, AP3 has 4 clients and AP4 has 4 clients. The output of the simulation is shown below. Figure 3: Unbalanced Loads The bit rate of all APs is 250kbps. Since the clients are more at AP2, from time interval 9sec to 17sec, the rate of the second AP is increased by 50kbps. Then load of the AP2 is calculated. If the load is greater than the threshold of 2000kbps, then overload is calculated. Then it checks, whether the
24 International Journal of Wireless Communications and Networking addition of the overload and load of AP1 is lesser than the threshold, if so then the clients of AP2 are reduced by 1 and the clients of AP1 are increased by 1. So, the load of AP1 is increased by the overload amount and the load of AP2 is decreased by the overload amount. Otherwise, no change will occur. Since the clients are more at AP1, at intervals 17sec and 19sec, the clients of AP1 are reduced by 1 and the clients of AP3 are increased by 1. So, the load of AP3 is increased by 250kbps and the load of AP1 is decreased by 250kbps. After balancing, AP1 has 5 users, AP2 has 5 users, AP3 has 6 users and AP4 has 5 users. its corresponding bandwidth allocation is max-min fair. It is widely accepted that the primary approach for obtaining a fair service is balancing the load on the access points. However, for WLANs the notion of load is not well defined. Several recent studies have shown that neither the number of users associated with an AP nor its throughput reflect the APs load. This motivates the need for an appropriate definition. Intuitively, the load of an AP needs to reflect its inability to satisfy the requirements of its associated users and as such it should be inversely proportional to the average bandwidth that they experience. Consequently, we are able to extend existing load balancing techniques to balance the AP loads and obtain a fair service. Taking users in x axis and throughput (bit rate in Mbpsx10-3 ) in y axis, the values of balanced loads and unbalanced loads were compared and a graph is drawn to find the user s bit rates. Figure 4: Balanced Loads Simulation parameters are user s bit rate and load of AP. Intuitively, a system provides a fair service if all users have the same allocated bandwidth. Unfortunately, such a degree of fairness may cause significant reduction of the network throughput, since all users get the same bandwidth allocation as the bottleneck users. The common approach to address this issue of fair allocation that also maximizes the network throughput is to provide max-min fairness. Informally, a bandwidth allocation of a weighted system is called max-min fair if there is no way to increase the bandwidth of a user without decreasing the bandwidth of another user with the same or less normalized bandwidth. Consider the case that each AP provides a weighted fair bandwidth allocation to its associated users. Then, a user association is termed max-min fair if Figure 5: Users Vs Throughput Throughput is the average rate of successful message delivery over a communication channel. To get maximum throughput, the load must be minimum. Similarly, the fairness is computed by by taking users in x axis and fairness (bit rate in Mbpsx10-3 ) in y axis. Each user s bit rate is calculated for achieving fairness. Fairness measures are used to determine whether users or applications are receiving a fair share of system resources.
Cell Breathing for Load Balancing in Wireless LAN 25 Figure 6: Users Vs Fairness x n x i i (10) Fairness = ( ) 2 2 The above equation rates the fairness of a set of values where there are n users and x i is the throughput for the i th connection. The result ranges from 1/n (worst case) to 1 (best case), and it is maximum when all users receive the same allocation. This index is k/n when k users equally share the resource and the other n k users receive zero allocation. This metric identifies underutilized channels and is not unduly sensitive to a typical network flow patterns. The power of APs can be controlled by modeling a wireless LAN with APs, each having users in unbalanced state. A server can be established to maintain the transmissions in order. While balancing the load of APs, the power can be maintained and controlled. VI. CONCLUSION Only the load of Access points in wireless LAN is balanced by setting a threshold value in this paper. While balancing the load, the power of APs will be controlled in future. REFERENCES [1] M. Balazinska and P. Castro. Characterizing mobility and network usage in a corporate wireless local area network, Proc. USENIX Int l Conf. Mobile systems, Applications, and Services (Mobisys 03), 2003. [2] Y. Bejerano and S. J. Han Cell Breathing Techniques for Load Balancing in Wireless LANs, IEEE Transactions on Mobile Computing, Vol. 8, No. 6, 2009. [3] S. Boivin, B. Gendron and G. Pesant, A Load balancing procedure for parallel constraint programming, 2008. [4] L. Du, J. Bigham, and L. Cuthbert, A Bubble Oscillation Algorithm for Distributed Geographic Load Balancing in Mobile Networks, Proc. IEEE INFOCOM, 2004. [5] T. Henderson, D. Kotz and I. Abyzov, The changing usage of a mature campus-wide wireless network, Proc. ACM Mobicom, pp. 187-201, 2004. [6] B. Kauffmann, F. Baccelli, A. Chaintreau, K. Papagiannaki, and C. Diot, Self Organization of Interfering 802.11 Wireless Access Networks, INRIA Research Report RR-5649, 2005. [7] T. Ralli, National Strategies for Public WLAN Roaming, Helsinki Research Report, 2006. [8] A. Sang, X. Wang, M. Madihian and R. Gitlin, Coordinated Load Balancing, Handoff/Cell-site Selection, and Scheduling in Multi-cell Packet Data Systems In Proc. ACM Mobicom 2004, pages 302 314, Philadelphia, PA, USA, September 2004. [9] O. K. Tonguz and E. Yanmaz The mathematical theory of dynamic load balancing in cellular networks, IEEE Transactions on Mobile Computing, Vol. 7, No. 12, pp. 1504-1518, 2008. [10] T. C. Tsai and C. F. Lien, IEEE 802.11 Hot Spot Load Balance and QoS-Maintained Seamless Roaming, Proc. Nat l Computer Symp. 2003. [11] S. T. Yang and A. Ephremides, Resolving the CDMA cell breathing effect and near-far unfair access problem by bandwidth-space partitioning In Proc. IEEE VTC 2001 Spring, Vol. 2, pages 1037 - Philadelphia, PA, USA, September 2004.