There is an emerging trend to apply SDN (Software Defined Network) to Wi-Fi. Much of this discussion today is geared towards applying SDN constructs like OpenFlow and Network Function Virtualization (NFV) to Wi-Fi access points (APs). For example, OpenFlow interface on the AP to configure its settings from the SDN controller or to configure traffic handling on the AP such as packet classifiers and QoS from the SDN controller. Another discussion point is around tunneling AP traffic to NFV node for service provisioning. While these are good ideas, they do not cut to the essence of Wi-Fi, which is in the wireless. That is to say, these concepts are applied to APs in the same way they apply to other wired network elements like switches.
I have mused about how SDN may apply to the wireless side of the network. I did not find the answer in Wi-Fi itself as it exists today. However, I had an “aha” moment recently while looking into LTE. Of late, I have been looking rather deeply into LTE, as a result of a Wi-Fi offload project we are undertaking with a telco carrier, as well as all the recent talk about LTE-U coming to the Wi-Fi spectrum. LTE has architected something called SON (Self-Organizing Network), which has SDN-like concepts – that are applied to the wireless side of the network.
What SDN does in principle is to isolate network intelligence from network functions. The network intelligence is centralized in an SDN controller. The SDN controller can reside in the cloud. The controller orchestrates network functions over the standard protocol (such as OpenFlow) to achieve desired application performance. Network intelligence can be provided separate from the network hardware and potentially by an entity not tied to the network hardware. The SON does something similar for the wireless side of the LTE network.
What is a Self-Organizing Network (SON)?
LTE isn’t immune to the fundamental challenges of wireless. For example, channel reuse is an important concept in wireless networks to build spatial capacity. However, during channel reuse, it’s important to avoid channel overlap among different cells. In LTE, a given telco carrier may have only a few (and sometimes only one) LTE channels for all its deployments region-wide, or even nation-wide, and these channels have to support a mix of overlapping large (macro) cells and small cells (femto cells, micro cells, pico cells). Another important concept in wireless is spectral efficiency, that is, using the available spectrum to carry as much useful data as possible. LTE spectrum is licensed and very expensive, thus there is high incentive to maximize spectral efficiency. User mobility is another factor that needs to be effectively managed in the wireless network. To address these problems at the scale of the LTE network, a large number of complex and dynamic radio parameters need to be configured, monitored, and optimized on an ongoing basis. 3GPP has come up with SON to address this problem.
Some of the key concepts of SON include:
- SON server: The SON server runs algorithms to continuously analyze and optimize the radio parameters in the network. Inputs from the radio network elements (mobile devices, base stations, etc.) are communicated to the SON server over northbound interface and the decisions on radio parameters are sent down from the SON server to the radio network elements.
- Separation of radio network intelligence from the radio network elements: The SON server is in charge of radio network intelligence and the radio network elements are responsible for radio parameters measurements and enforcement of the SON server decisions. This is significantly different from the pre-LTE networks where a large part of radio intelligence was integrated into the radio network.
- Standard interfaces: The interface between the SON server and the radio network elements is standardized by 3GPP. This facilitates interoperability between the radio network intelligence and the radio network elements.
- Automation of radio configurations and optimizations: On a high level, SON algorithms are expected to hit objectives of self-configuration, self-optimization, and self-healing of the radio network. High emphasis is placed on automation of radio parameter configuration and optimization. How exactly this is done depends on the specific SON server/algorithm implementation.
Going into details of many radio parameters addressed in SON (such as cell coverage and capacity, dynamic neighbor relations for handovers, mobility, load balancing, random access, and self-healing) will not fit in this blog and requires a deep dive into the 3GPP specification. A good introduction to SON is also found in this paper by 4G Americas: Self-Optimizing Networks in 3GPP Release 11: The Benefits of SON in LTE.
Wireless SDN for Wi-Fi
Wi-Fi is no stranger to the complexities of radio resource management (RRM). The 802.11 standard offers little to solve the radio management problem. Fair enough if the 802.11 standard authors did not feel this pain point initially, when there were ample channels available in Wi-Fi and they did not anticipate today's rabid Wi-Fi penetration. This resulted into emergence of various proprietary RRM and RF optimization features in the APs to improvise on the omission in the standard and reliance on vagaries of client implementations. However, as Wi-Fi gets integrated into LTE architecture for wireless data offload, it is conceivable that various aspects of Wi-Fi RRM may also get folded into LTE SON. In other words, we could see standardized interfaces between the AP and the SON server and intelligent algorithms in the SON server for RRM purposes. And that is how Wi-Fi may get its first wireless SDN!
Is there a path for wireless SDN to come to enterprise Wi-Fi? To make these types of projections, we can take clues from the history of SDN on the wired side.
SDN on the wired side has traditionally seen great traction where large network workloads are involved, such as data centers and back-end networks of web powerhouses. It is trickling into the large enterprise vertical slowly but surely.
At this time, wireless SDN concepts seem to be taking shape where large radio management workloads are involved. But who knows; it may trickle to enterprise Wi-Fi in the future.