SON vs Wi-Fi

Wireless LAN technology is an indispensable part of the enterprise network. As the network increases, managing tens of thousands of Access Points (APs) becomes an issue. This leads to the requirement of self-configuring, self-managing and self healing networks. SON (Self optimizing networks) is the answer to this.

A self optimizing network of wireless nodes represents a collection or network of co-located network nodes which can jointly set their own operating configuration so as to maximize network performance (measured in whatever mode makes sense). SON applications as part of wireless LAN APs have been launched by Cisco, Aruba and others. The CISCO suite of Wireless LAN applications includes Cisco Unified Wireless Network which incorporates various SON features. In the LTE domain, SON has been included into the 3gPP specifications and is being actively pursued by Nokia Networks, Ericsson, AT&T and others.

In a typical self-optimizing network, we are interested in optimizing some global metric in performance, subject to constraints in another global metric, by setting the local policy for specific operational parameters at each individual node in the system. Thus, a SON can try to minimize handovers in the network as a whole, subject to average call drop probability being above a certain threshold; both of these are global, user-visible matrices. The optimization is achieved by setting certain policy parameters for each and every network node; for example the SINR threshold at which a handover is triggered. In most of the currently deployed self-optimizing networks, the SON function resides in a controller or set of controllers, which receive feedback from the network nodes (base-stations, access points, or node B, depending on the technology) and in turn sends out policy directives to them. We call this the central control model.

We at HSC have been studying this technology and have found that SON solutions with central controls are limited by information exchange and future networks in turn would need to deal with 100s of access points. We feel static planning will be supplanted by dynamic† management (Proliferation of WiFi mini-APs, power management and† energy conservation).In the last quarter we have conceptualized (and executed a proof of concept ) of a fully distributed SON solution (can operate with or without a central node). The solution created is highly scalable and is integrated with Linux hostapd. Power control, coverage optimization, QoS management and energy optimization are the key highlights of this solution.

We are now planning to build further on this base initial solution where idea is to provide the better user interface, revamped data visualization and modeling, enhanced stability. We are working on a new algorithm which can scale to much larger networks and to provide the support for QoS differentiated networks.