Mobile wireless networks today are faced with two conflicting trends.

  • The demand side. Traffic is growing exponentially and becoming increasingly diverse. Several industry reports estimate that traffic will double approximately every nine months, leading to a several orders of magnitude increase in load over the next few years.
  • The supply side. Available spectrum and spectral efficiency (i.e. the maximum throughput achievable per Hz of spectrum) are flattening out. In fact, the spectral efficiency of 4G LTE PHY is fairly close to (within 20% of) the Shannon capacity limit, and further improvements are likely very expensive to implement and will only provide limited gains.

To cope, future wireless network deployments have to get dense. This is motivated by a simple observation: The surest way to increase per-user capacity is to make cells small and bring the basestation closer to the mobile client, since this improves per link SNR and reduces the number of users per cell. Some industry estimates suggest basestation density will have to increase by 20× to handle the exponential traffic growth.

However, current network architectures are ill-equipped to support a dense and dynamic wireless infrastructure. First, since it will be impossible to obtain regularly placed cell sites for an infrastructure with 20× higher density, basestations will be deployed wherever possible in a chaotic fashion. However, a chaotic and dense wireless deployment will be very complex to manage, since it will experience highly variable loads and unpredictable inter-cell interference among other things. Finally, a dense infrastructure is very expensive to deploy and operate. Current deployments are unaffordable except to the largest operators, so a deployment with 20× more basestations will likely be infeasible even for the largest operators.

Our larger research goal is to design a novel network architecture called OpenRadio for future dense and adaptive wireless infrastructure. At its core, OpenRadio aims to apply the software defined networking (SDN) approach to design systems and abstractions that simplify the deployment and management of dense cellular wireless networks. The agenda is ambitious and involves several components, including developing the programmable basestations and network infrastructure (akin to programmable switches for wired networks), developing a wireless control plane (akin to NOX for wired) and capabilities to slice the wireless network. In the next 18 months, as part of this research agenda, we plan to focus on the first aspect:

Programmable Wireless Data Plane

A basic requirement for a SDN based network architecture is a programmable wireless dataplane. Like programmable switches for wired SDN architectures, we plan to design a high performance, programmable wireless basestation (BS). The same programmable BS can be customized to support LTE, WiMAX or WiFi. It will provide further the following properties:

  • Real-time high performance. Managing wireless networks requires making modifications at all three layer, the PHY, MAC and network layers. For example, to manage inter-cell interference in dense networks, operators may need fine-grained control over subcarrier spectrum allocations. Or to adapt for a different traffic QoS requirement (e.g. provide low latency access for DNS requests), the MAC scheduling algorithm may need to be changed. While programmable wired devices stay above the link layer, a wireless one must be able to implement and evaluate modifications across the entire stack in the context of modern wireless protocols that operate at high rates (100Mbps) and under strict real- time deadlines (e.g. LTE's 1ms scheduling granularity). Doing so requires very tight timing guarantees as well as tremendous computational performance (e.g., 100GFlops for WiFi).
  • Modular Abstractions. To simplify management, a net- work operator must be able to program the data plane using high-level modular interfaces (like the rule/ac- tion abstraction for SDN switches). As discussed above, programming a wireless network involves not just the network, but also the PHY and MAC layers. Further, the changes to the network typically touch a small aspect (such as how scheduling is handled for video traffic, how the backhaul is managed for a specific basestation etc). Operators should be able to use high-level interfaces to in- tegrate these changes, without having to understand how the protocol is actually implemented on the hardware.
  • Scalable. To be deployable at scale, the programmable base stations have to be reasonably affordable while operating at modern speeds as well as accommodate multiple operator specified protocol enhancements. Our goal is to build such a programmable basestation for under $3000, which is at least an order of magnitude cheaper than commercially available basestation.?We have made initial progress in our research on architecting such a programmable basestation. We are collaborating with Texas Instruments and currently have a prototype basestation that currently runs a full 802.11n WiFi stack with data rates up to 108Mbps. We are currently architecting the LTE basestation, and researching the modular interfaces that the programmable BS should expose to allow operators to easily modify the network behavior across the PHY, MAC and network layers.

For more info, please contact: Manu Bansal