Traffic load aware channel assignment method

Most of the channel assignment algorithms introduced above focus on how to maximize the throughput or minimize the interference of the network. They aim to build a high capacity high quality network. There are some other researchers who consider the channel assignment to satisfy the traffic load requirement of the individual wireless devices. Rozner et al [116] proposed a static or infrequent channel assignment which considers the traffic pattern of stations and APs. They assume the traffic load requirement of each AP is well known by the centralized controller and the AP with high individual demand is first assigned to non-overlapping channels. However, the network performance is no better than other algorithms with an incomplete traffic information or when the APs cannot communicate with each other to obtain the complete traffic information. LCA [117] proposed a load aware channel assignment algorithm where the AP will select the channel which has the lowest airtime cost which is an approximation of the average per-packet delay. AP and stations will continuously

scan the available channels to obtain the airtime cost value. With an increase in the number of channels or when the traffic load changes dramatically, the process overhead can be unacceptable. In the fifth chapter of this thesis, we propose an autonomous channel selection algorithm which autonomously selects the channel without a central controller in order to reduce the incidence of congestion.

3.5 Conclusion

This chapter introduces other research work in the area of available bandwidth estimation and channel assignment. There are two main methods used to estimate the available bandwidth: Passive methods and active methods. The active methods estimate the capacity of link using the concept of packet dispersion where the dispersion between the packets indicates the capacity of links of both sides of a path. Probe packets are transmitted with different packet sizes and PHY transmission rates to increase the accuracy of estimation.

The next topic presented was the use of channel assignment algorithms to increase the throughput of network by utilizing multiple channels. Depending on whether or not the MAC protocol was modified, multiple channels were used in different ways. Some researchers focus on developing a MAC protocol suited to utilizing multiple channels simultaneously. Some of the protocols separate the channel set into a control channel and several data channels. The control channel is used to maintain the network connectivity and the data channels are used to transmit data frames. Other novel MAC protocols divide the time into small intervals and also separate the control frames and data frames to increase the accuracy of network management. Because these novel MAC protocol methods cannot be directly implemented on the current IEEE 802.11 hardware, some other researchers pay more attention on developing channel assignment algorithms utilizing the standard IEEE 802.11 MAC protocol. These algorithms are divided into three main categories: Static channel assignment, Dynamic channel assignment and Hybrid channel assignment. Static channel assignment combines the channels and radios permanently or for a long time (in the unit of hours or days).

Dynamic channel assignment assigns the channels to the radios depending on the traffic load and channel status. When the current operating channel is highly contended or it cannot satisfy the traffic requirement, it will switch to another available channel. Hybrid channel assignment combines the static and dynamic features together and has been utilized in multi-radio networks. Some of the radios were statically assigned channels to keep the network connectivity and other radios were assigned channels dynamically to suit a dynamic traffic load.

There are also other channel assignment algorithms which combine metrics from other layers such as the transmit rate from PHY layer and routing from the network layer.

Game theory based channel assignment algorithms are usually utilized in cognitive networks. All the radios were considered as players and the action of channel selection is defined as the strategies. Minimizing the interference or maximizing the channel utilization is the utility function in these algorithms. However, these algorithms still have many constraints that prevent their implementation in the real networks.

All the previous channel assignment algorithms focus on how to arrange the channel for the wireless radio to maximize the throughput or minimize the interference. Once any of the condition changes (such as a change in the traffic load or a new station joins) the algorithm will be triggered to reassign the channel. The centralized channel assignment has a low efficiency in a large scale network. The distributed channel assignment always changes to the channel with lowest-interference or the channel with largest available bandwidth without considering this channel will satisfy the traffic load or not.

In next chapter, we will introduce a novel available bandwidth estimation which based on passive bandwidth estimation. It collects packet information for all the neighbours and calculates the access efficiency and traffic load of these stations. A dynamic channel selection algorithm is used to select the channel with the help of bandwidth estimation algorithm. This channel selection algorithm operates in distributed manner and all the stations maintain this channel until this channel cannot satisfy the traffic load. If there is no channel that can satisfy the traffic load, it utilizes a “neighbour forcing” method to rearrange the channel of the neighbour stations in order to satisfy the traffic load of all

the stations.

Chapter 4 Passive bandwidth estimation based on