Article Preview
TopIntroduction
With the advance on cloud computing and wireless communication technology, using mobile devices to access cloud services becomes common nowadays. The provision of mobile cloud service calls for a wireless access infrastructure that offers high bandwidth to mobile users. Enabling technologies include WiFi, 3G, Worldwide Interoperability for Microwave Access (WiMAX), and Long Term Evolution (LTE). Among all these technologies, WiFi or IEEE 802.11 wireless network has the advantage of widely available user equipments. An IEEE 802.11 wireless network generally consists of multiple base stations (called access points) interconnected by a wired network (which can be Ethernet or optical fiber network). This type of network deployment incurs high cost and complexity especially in some inaccessible terrain. IEEE 802.11s is a proposal for wireless mesh network (WMN) that allows IEEE 802.11 access points to connect with each other by a wireless technology. Wireless mesh network can effectively extend the service range of a typical wireless local area network while reducing the deployment cost. To provide more bandwidth, access points in WMNs are commonly equipped with multiple wireless interfaces (radios) that can operate on multiple non-overlapping channels in parallel. This paper discusses the allocation of radio resource (radios and channels) to access points in WMNs.
In a WMN, two base stations must tune to the same radio channel before they can communicate with each other. If some neighboring base station uses an overlapping channel or the same channel, the receiver will suffer from co-channel interference. In the physical layer, co-channel interference may reduce signal quality of the receiver and increase the data error rate. In the data link layer, co-channel interference will increase transmission collisions and channel contention. These two factors lead to degraded throughput (Lee, Lee, Kim, Jo, Kwon, & Choi, 2009). IEEE 802.11 and other wireless communication techniques provide multiple non-overlapping channels for the devices to reduce co-channel interference. An efficient utilization of these non-overlapping channels in WMNs can potentially increase network capacity.
Many scholars have proposed approaches to an effective utilization of multiple non-overlapping channels. So and Vaidya (2004) presented channel switching that allows single radio to use multiple channels. The mechanism needs a complex scheduler to ensure that the sender and the receiver can switch to the same channel simultaneously. For a multi-hop transmission path, if channels allocated to the nodes on the path are too diverse, additional delays will be incurred on each link for channel switching, which may significantly increase the end-to-end packet delay. If nodes have multiple radios operating on different channels, sending and receiving can be done simultaneously and thus avoids channel switching. However, a link is operative only if the devices at both ends of this link are configured with a common channel (common-channel constraint) and the interference experienced by both devices does not exceed a threshold (interference constraint). The common channel constraint and the interference constraint are conflicting in nature. Therefore, how to maximize the number of operative links subject to these two constraints has become a research issue.
To deal with this issue, researches have proposed many solutions. Most solutions consider co-channel interference at the link layer while this study focuses on physical-layer co-channel interference. Furthermore, conventional approaches treat this problem with heuristic algorithms. Recently, game theory has been applied to this problem. Our work wants to study whether simulated annealing, a well-known optimization technique, perform better than these existing methods.
When analyzing co-channel interference, most studies consider protocol model while others take physical model (Raniwala, Gopalan, & Chiueh, 2004; Yen, Huang, & Leung, 2012). Protocol model assumes that link interference relation is binary and Boolean-valued. The interference relation is defined on pairs of links and two links either completely interfere with each other or not. Physical model considers intensity of interference, which takes different levels for different link pairs, and whether a link operating on a particular channel is severely interfered should consider overall interference intensity caused by all other links that use the same channel in the network. This study takes physical model and uses signal-to-interference ratio (SIR) as the interference intensity value.