An efficient hybrid algorithm for balancing the load of wireless local area network

Highlights

• An efficient and effective genetic algorithm for WLAN load balancing called MMM-GA is proposed.

• Simulations show that MMM-GA provides a higher bandwidth than GA.

• Simulations show that MMM-GA takes much less computation time than GA.

Abstract

Although the IEEE 802.11 wireless local area network (WLAN) provides users of mobile device a convenient method to connect to the Internet, the load of access points (APs) may have a strong impact on the performance of a WLAN, such as the congestion problem. To avoid this issue and to maximize the performance and scalability of a WLAN, an effective method for balancing the load of APs, called a modified multiple-search multi-start framework based micro-genetic algorithm, is presented in this paper. Simulation results show that the proposed algorithm gives a much better solution than the original genetic algorithm does in terms of both the quality and the computation time.

Introduction

In recent years, it has been witnessed that the most popular technique for connecting mobile devices, such as smart phones, tablet computers, and notebooks, to the Internet is certainly the so-called wireless local area network (WLAN). The main usage of a WLAN is to link devices in a limited area, such as a room, home, or building, to the Internet wirelessly. WLAN has been widely used and plays an important role in many applications—from remote device control, communication, game, to healthcare, home care, and even the internet of things (IoT) [1]. Because several successful results of IoT were found in recent studies [2], [3], [4], [5], [6], [7], how to make it more effective and how to make it more efficient are apparently two promising research issues. One important way to make the applications of IoT more effective is to integrate data mining or intelligent methods into it to provide more suitable services to the users. On the other hand, an intuitive way to make such kind of systems more efficient is to enhance the performance of its components, such as algorithms, mechanisms, protocols, devices, and systems. These are all possible solutions. Since WLAN is the infrastructure of IoT and many other applications, its market scale is also very large today. According to the report [8], the total revenue of enterprise WLAN vendors in 2015 was about 5,011 million U.S. dollars. The report [9] forecast that the market of WLAN in 2019 will be worth 12,101 million U.S. dollars. This means that the market of WLAN has a great potential in the future.

It is the access point (AP) of a WLAN that acts as a base station for connecting a device to the Internet. In other words, it is the AP that is used to transmit and receive signals between wireless enabled devices and the Internet. If a device is covered by two or more APs, there is a standard mechanism, which is based on the received signal strength indicator (RSSI), to decide to which AP the device is supposed to link. According to the RSSI, the device simply links to the AP that has the strongest signal. This mechanism will, however, cause a serious problem in a WLAN. Since the AP that has the strongest signal will be chosen by all the devices, some APs will become popular (i.e., the hot spot), meaning that the load of these APs will be much higher than that of the others. Furthermore, the bandwidth of each AP is limited. This implies that the higher the number of clients connected to an AP, the lower the bandwidth each client can share. This phenomenon may congest the network of some AP, thus degrading the quality of service (QoS) for the clients connected to it. On the other hand, the APs that have a weaker signal may waste their bandwidth because most of them are idle. Some researches use different methods to solve the load balance problem of a WLAN, such as 3G and LTE [10].

Fig. 1 gives an example of six clients (numbered 1 to 6) covered by three APs (named A, B, and C). Based on the RSSI [11], all the clients will connect to the AP with the strongest signal. The strength of a signal is influenced by the distance between the client and AP. Fig. 1 shows that clients 4 and 5 are covered by C only, but the other clients are covered by two or three APs. Since AP C is closest to these clients, according to the RSSI, these clients (i.e., those other than 4 and 5) will also connect to AP C. As a result, AP C will serve six clients at the same time while the other APs will serve no clients at all. Consequently, the load of the WLAN will be highly unbalanced. It is quite obvious that for the APs of a WLAN, some may be underloaded while some may be overloaded. This situation will cause clients to face the congestion problem, which includes the queuing delay, packet loss, and blocking of new client. Therefore, a WLAN can be made more scalable if the underloaded APs are used in a more efficient way, thus ensuring the QoS or maximizing the performance of each client.

According to our observations, the relevant technologies of WLAN have become more and more important owing to their usage in wireless network environments, such as IoT and wireless sensor networks (WSN). The rise of handheld devices has found more and more APs being deployed. As a result, how to find good solutions to enhance the performance of WLAN, especially the deployment of APs, has become an important issue because the result of the deployment may have a strong impact on the performance of these network environments. Some researches proposed to use either heterogeneous wireless networks (such as cellular, WiFi, Zigbee, and Bluetooth) or homogeneous networks (such as WLAN and WPAN) to balance the load of WSN and IoT [12]. All these researches show that a more balanced network load implies a less communication delay. From this perspective, it will be useful to have an effective method to balance the load of IoT and WSN. Motivated by these observations, this paper presents an efficient algorithm to balance the load of WLAN, by constructing a better distribution of the APs, which we expect can also be applied to the network environments of IoT and WSN. This means that to maximize the bandwidth and to ensure the QoS of each client at the same time, the algorithm should be able to decide a more suitable connection between AP and client and migrate clients when a new client comes in to the place that is also covered by some other APs. For such a dynamic network environment, the network management system should be able to find a better result to balance the load at real time.

Metaheuristics [13], [14] provide a promising way to solve complex optimization problems because they can find an approximate, or even the optimal, solution within a reasonable time using a limited amount of resource. They can be divided into two classes: single-solution-based and population-based [13]. Unlike single-solution-based metaheuristic algorithms that search one direction (i.e., solution) at a time for the possible solution, population-based metaheuristic algorithms search multiple directions at a time for the possible solutions. Because multiple search directions can increase the search diversity, the population-based metaheuristics can typically provide an efficient way for finding a high quality result for the load balancing problem of WLAN. However, the main disadvantage of population-based metaheuristics is in that they take a much longer time than rule-based algorithms, deterministic algorithms, and single-solution-based metaheuristics in most cases. Thus, our aim is to reduce the computation time of population-based metaheuristic algorithms while improving the quality of the results. Micro-GA [15] was presented to find a better state for WLAN in a reasonable time. It sets a limit on the population size to reduce the computation time. As a trade-off, it usually finds a result that is worse than the simple genetic algorithm (simple GA) because a smaller population means a fewer number of search directions, which will reduce the chance of finding better results. Inspired by this observation, this paper presents a modified multiple-search multi-start framework to improve the quality of the micro-GA to find an approximate solution in a reasonable time, which we refer to as the MMM-GA.

The main contributions of this paper can be summarized as follows:

• An effective heuristic algorithm is presented to construct a suitable connection between APs and clients so as to improve the QoS of a WLAN.

• This paper gives a detailed description of how the proposed algorithm leverages the strength of micro-GA and modified MSMS framework.

• This paper also gives a detailed description of how the proposed algorithm deals with the load balancing problem of WLAN and how it can be applied to real optimization problems.

The remainder of this paper is organized as follows. Section 2 presents the related work for the load balancing problem of WLAN. Section 4 describes in detail the proposed algorithm. Section 5 gives the experimental results to show the performance of the proposed algorithm and the influences that different parameter settings may have. The conclusions and future work are drawn in Section 6.