Cloud-Enhanced Open Software Defined Mobile Wireless Testbed for City-Scale Deployment

Project Objectives

Awarded by NSF under the Platforms for Advanced Wireless Research (PAWR) program, COSMOS (“Cloud-Enhanced Open Software-Defined Mobile-Wireless Testbed for City-Scale Deployment”) is an advanced wireless testbed which enables researchers to explore the technology “sweet spot” of ultra-high bandwidth and ultra-low latency, a capability that will enable a broad new class of real-time applications including augmented/virtual reality and cloud-based autonomous vehicles, in the most demanding real world environments such as a city.

Technology Rationale

A large-scale advanced wireless testbed like COSMOS is a complex engineered system that requires careful architectural design and technology selection in order to realize the full set of requirements. In particular, the functional requirements of the testbed as articulated in the PAWR RFP include advanced wireless capabilities along with qualitative features such as remote access for experimenters, dynamic sharing and/or virtualization of resources across multiple experimental users, open interfaces and programmability right through the protocol stack (including radio, network and compute layers), control over network topologies and node mobility, diversity of radio environments, reproducibility of experimental results, and ease of use in terms of running an experiment, collecting measurement data and storing/processing the data, extensibility with new technology components, and interoperability with other testbeds/networks. Quantitative requirements for the testbed include scale (in terms of # nodes, geographic coverage area, user density), radio bandwidth and speed, network latencies, cloud computing capacity, etc.

Achieving these multi-dimensional design goals requires careful consideration of testbed architecture and technology building blocks. The architecture of COSMOS is informed by best-of-breed networking and wireless testbeds such as GENI, Emulab, PlanetLab, OneLab, CloudLab, CIAN TOAN, and most notably ORBIT, which is a large-scale indoor emulation testbed with somewhat similar design requirements and scale. Some of the tried-and-tested design principles adopted from ORBIT include centralized testbed control and management, programmable sub 6 GHz radio nodes accessed through a separate control plane, SDR including open-source realizations for WiFi and LTE standards, software-defined switching infrastructure for experimenter control of network topology, virtualization of resources for multi-user support, the option for thin clients and provisioning of cloud resources for radio signal processing, providing smaller scale “sandboxes” to simplify, debug and offload early stage experiments, and an extensive control and management software suite for experiment orchestration and data collection.

The COSMOS testbed poses additional design challenges due to new requirements including wideband radio signal processing (with bandwidths of ~500 MHz or more), support for mmWave communication (28 or 60 GHz), technical challenges of effectively virtualizing radio resources, low latency front- and back-haul, tightly coupled edge cloud and real-world deployment. Many of these issues can be addressed architecturally by building COSMOS as a multi-layered computing system (see Fig. 1) with an RF thin client that can flexibly partition signal processing and network function virtualization (NFV) between a local SDR (with FPGA assist) and a remote cloud radio access network (CRAN) with massive CPU/GPU and FPGA assist. Further, these two SDR computing layers are backed up by a third layer of general purpose cloud computing useful for network and application level functions associated with an experiment.

In terms of technology, radio nodes in COSMOS will provide a mix of fully programmable SDRs for flexible wireless experimentation as well as commercial hardware capable of supporting networking and applications research with currently available end-user devices. Following the ORBIT’s open API philosophy, COSMOS will be built in a bottom-up manner with commodity components and customized open-source hardware and software modules (thus, largely avoiding the use of advanced white box or black box commercial solutions for base stations, access points, routers, mobile core networks, etc. which are very difficult to open up for general purpose experimentation). The developed wireless platforms will cover the full range of spectrum including the sub 6 GHz bands used for today’s services as well as emerging 28 GHz and 60 GHz mmWave bands. SDRs to be used will utilize a new design which achieves an order-of-magnitude performance headroom over current technology (such as X310 or USRP Rio), achieving real-time processing of wide bandwidths ~500 MHz via novel acceleration techniques. In addition to fully programmable radios, COSMOS will incorporate state-of-the-art COTS equipment (e.g., the best available 5G or 802.11ad) in order to enable a base service layer that can be used for networking or applications research and for general control and communications needs.

The COSMOS system will also incorporate fast programmable core network technology to keep pace with significant increases wireless link bandwidth and to effectively integrate emerging radio access networks with edge cloud computing. The proposed design includes novel 100 Gbps+ fiber, free space optical, and microwave backhaul technologies interconnected with a software-defined network (SDN) switching fabric for minimum latency and flexibility in setting up experimental network topologies. Sub-microsecond speed optical switching technology will offer the option of passive WDM switch fabrics and radio over fiber interfaces for the purpose of achieving ultra-low latency connections to edge computing services, which will be built in as an integral part of the system. Edge cloud technology to be integrated into the COSMOS system includes commodity CPU + GPU hardware along with FPGA-based co-processors in order to achieve processing speeds necessary to support CRAN, virtualized network functions, and low-latency cloud application scenarios. Another key technology for a successful testbed is the control and management framework which supports experiment setup, resource assignment, orchestration and measurements. The ORBIT Management Framework (OMF) along with selected control elements from GENI and CloudLab provide a proven control software baseline which can be leveraged and extended for use in this project.

Multi-layered SDR computing architecture of COSMOS

Technical Approach

COSMOS advanced wireless platform is to be deployed in New York City (NYC). Rutgers, Columbia, and NYU have formed a multi-university team to address the technical and deployment challenges of PAWR in collaboration with our municipal partner, the City of New York, our community partner, Silicon Harlem, and our testbed partner, The City College of New York (CCNY). The team has extensive experience in the design, development and operation of large-scale wireless, networking, and cloud testbeds (ORBIT, GENI, CloudLab, and CIAN), along with specialized expertise in key enabling technologies necessary for PAWR including software-defined radio (SDR), dynamic spectrum, massive MIMO, mmWave, software-defined fiber and free-space optical x-haul, and edge cloud.

The COSMOS testbed will be deployed in upper Manhattan covering an area of about 1 sq. mile which is adjacent to the Columbia University and CCNY campuses and includes much of West Harlem. The City of New York, under the leadership of the Chief Technology Officer, will facilitate access to City assets, infrastructure, buildings and fiber to enable the installation of node sites to support the PAWR platform. We believe that it is essential for the testbed to be located in a dense urban area where wireless solutions can be evaluated for emerging mega-cities which will soon account for over half the world’s population. The COSMOS proposal offers a testbed and research platform in one of the country’s most populated urban centers, New York City, offering considerable diversity in radio signal propagation, vital real-world application cases and a potentially large community of 20,000+ end-users. The deployment will incorporate 40 advanced wireless base stations and 100’s of mobile nodes along and around Broadway, Amsterdam avenue, and W 125 St. These corridors, which will be thoroughly characterized, will become real world reference standards in the research community not available today. Further, COSMOS will be positioned as a central pillar within a tech venture eco-system of over 60,000 professionals in which meets and other events typically draw thousands of participants. Harlem and NYC community engagement, education, and workforce development programs will be facilitated through the NYC gigabit centers and a wide array of programs with partners Silicon Harlem, Apollo Theatre, Community Boards 9 and 10, Community School Districts 3 and 5, the Schomburg Center for Research in Black Culture, NYCHA Housing developments, the Grant Houses, the Manhattanville Houses, the St. Nicholas Houses, Digital Divide Partnership, the Zahn Center at the City College of New York and the Morningside Area Alliance.

Results To Date and Future Work Plan


COSMOS is currently in deployment stage and following core technology research challenges can be addressed with the COSMOS platform:

  1. Advanced PHY techniques such as massive/cooperative MIMO and adaptive beamforming, full duplex, multi-connectivity and coordinated scheduling
  2. Conclusively verifying mmWave feasibility for mobility services
  3. Heterogeneity, multi-homing and densification in future cellular networks
  4. Dynamic spectrum access including distributed protocols & directional sharing mmWave
  5. Latency reduction in mobile network MAC/PHY and network layers
  6. Integration of optical x-haul technology with wireless technologies
  7. Clean slate architectures for mobile networks
  8. Adaptive multicast for crowded venues
  9. Edge cloud integration with wireless networks


Prof. Dipankar Raychaudhuri
ray (AT) winlab (DOT) rutgers (DOT) edu
COSMOS Website