Wireless communications have completely revolutionized the way we connect with people and access information. With the advent of Wi-Fi and 3G/4G cellular technologies, we can access the Internet from nearly everywhere around the world without the need for cables, and more importantly, while on the move. Increasingly powerful portable devices, such as tablets and smartphones, are further increasing the traffic demand on wireless infrastructures.
Increasing the capacity of a wireless network can happen in two ways: either through the authorized use of additional or wider frequency bands, or through the densification of the underlying wireless infrastructure through the deployment of an increasing number of Wi-Fi Access Points (APs) or cellular Base Stations (BTSs). The former solution is typically a very lengthy, expensive, and heavily regulated process. The latter bears a very significant investment cost on the part of the operator, currently undertaken by a number of cellular providers in their deployment of femtocell and picocells. When it comes to Wi-Fi networks, enterprises tend to spend a very significant amount of money deploying APs at a density that may be as high as one AP every four meters (commonly found in enterprise Wi-Fi deployments).
However, the pure increase of the density of Wi-Fi APs or cellular BTSs is not a solution by itself. Additional complexity arises from the management of a larger number of devices in the network, and more importantly its configuration to reap the desired gains. Benefits in terms of capacity are only delivered if the operator is able to appropriately allocate frequencies/channels to the different devices in its network, so as to minimize areas of overlap between APs that operate on the same frequency. The fundamental problem addressed in frequency selection is that given the wireless medium is a shared medium, the more devices you have operating in the same frequency band, the lower the effective throughput for each individual device (something that leads to even lower performance if a device is located in the area of frequency-overlapping base stations). Power control, combined with intelligent frequency selection algorithms, aims to increase frequency reusethe ability to reuse the same frequency often across space, however, without creating areas of frequency overlap.
The complexity of such a task is not for the fainthearted and has essentially led to the creation of an entire industry around Wi-Fi centralized architectures, pioneered by companies such as Aruba, Meru Networks, and Cisco, among others.
The following paper says that having multiple APs with an overlapping coverage area operating on the same frequency may not be a problem anymore. The authors describe a solution that can allow a wireless LAN to scale its throughput by continually adding more APs on the same channel! The target deployment scenario is that of a conference room or an auditorium, where APs are connected to each other through a high-speed wired network, and where dense AP deployments are absolutely necessary to accommodate traffic demand, while channels are too limited in number to prevent overlap.
The authors describe a solution that can allow a wireless LAN to scale its throughput by continually adding more APs on the same channel!
The authors borrow the fundamental working principle in today's Multiple Input Multiple Output (MIMO) transmittersthat of beamformingand make it work across a number of independent transmitters. They call their scheme Joint Multi-User Beamforming (JMB). The challenge that must be addressed is the JMB transmitters need to control the relative phases of their transmitted signals to enable effective beamforming, by which the signals to unintended recipients cancel out. Given that independent transmitters have independent oscillators, such a requirement is not naturally met.
The authors address this challenge by designating one AP in the wireless LAN as the lead AP. The solution works in two phases. During the measurement phase, each AP measures its channel to each receiver, as well as the channel from the lead AP to the slave APs. During the data transmission phase, each slave AP corrects its frequency offset with respect to the lead AP, and all APs jointly transmit to concurrently deliver packets to multiple receivers. They show that such a mechanism can be easily accommodated within the context of 802.11n.
An actual implementation on a 10-node software-radio Wi-Fi testbed demonstrates a linear increase in network throughput with a median gain of 8.1 to 9.4x. Further experiments on unmodified 802.11n cards highlight the tremendous potential of the proposed solution.
The work discussed in this paper could completely change the philosophy underlying the design of dense enterprise wireless LAN deployments.
To view the accompanying paper, visit doi.acm.org/10.1145/2618413
The Digital Library is published by the Association for Computing Machinery. Copyright © 2014 ACM, Inc.