Policy Driven Multi-band Spectrum Aggregation for Ultra-broadband Wireless Networks

Sustaining the impressive mobile internet revolution that is transforming our lives requires addressing the key challenge of designing ultra-broadband wireless networks of the future that can keep up with the ongoing impressive data traffic growth. One of the key techniques to increase per-user as well as system capacity is to increase amount of spectrum used in the network. However, given cost-effective radios that can radiate in 100s MHz of spectrum are still not practical, alternate solutions that concurrently exploit multiple radios and multi-band support available in current devices and networks need to be explored. The carrier aggregation technology implemented LTE-Advanced networks represents a step in this direction. However, such an intra-network, single standard, layer-2 approach at baseband does not scale well as it does not effectively exploit multi-network. multi-band and multi-operator connectivity and therefore, to rectify these shortcomings we envision an aggregation solution at transport or network-layer. We propose an end-to-end architecture that uses Multi-path TCP (MPTCP) as the key building block and supports concept of multi-path policies that expose to end-users and networks unprecedented control on how network resources on various spectrum bands and paths get used. We identify that uncontrolled widespread MPTCP use can lead to unfair resource allocation and reduced spectral efficiency. We solve this problem using a centralized network-resident service for managing multipath usage throughout an administrative domain. This service makes central decisions based on network-wide information and operator policies. We describe our end-to-end prototype that uses a (1) modified MPTCP implementation, (2) a transparent, high capacity, network resident aggregation proxy in the data path, (3) a network resident policy controller with our subflow control algorithm and (4) client devices with support for receiving policy commands and controlling MPTCP subflows. We report on experiments that demonstrate optimized aggregate network capacity in a variety of interesting scenarios.