Heterogeneous Resource Allocation for Hierarchical Software

Heterogeneous Resource Allocation
for Hierarchical Software-Defined Radio
Access Networks at the Edge
03.03.2015 - 31.12.2016
Prof. Zhu Han
Dr. Xianfu Chen
Prof. Guoliang Xue
Dr. Mehdi Bennis
..Where do we stand..
5G and the1000x bps/Hz/km2
mmWave + HetNets
Bandwidth (10x more Hz)
millimeter Wave
mmWave + massive
Smaller antennas for
mmWave, seems promising
But competition for the
DoF offered by antennas
Improved SINR via
mmWave with high gain
antennas, interference
becomes negligible?
Effective Density (20x
More Loaded BSs/km2):
Spectral efficiency (5x more bps/Hz)
Densifying mmWave cells
yields huge gains (SNR plus
cell splitting)
Can possibly do selfbackhauling!
More dimensions (massive MIMO?)
Interference suppression? (must fight
through log)
Efficient HetNets,
more small cells sand
WiFi and D2D
HetNets + massive MIMO
Small cells probably not be
able to utilize massive
Cost a key challenge
• Exponential traffic growth, limited capacity gains, low ARPU, etc.
• Solution: small cells, more antennas (massive MIMO), more spectrum
• Bottleneck: Backhaul issues, limited spectrum, scalability, complexity
• Trend: from sparse to ultra-dense networks
 More cells means more cell edges
 Coordination schemes are needed (CRE, ABSF, etc.) -- 40ms for CRE (latency!)
 LTE control plane designed for sparse (not dense) deployments
 Paradigm shift (Phantom-cell concept, cloud-RAN, FOG/EDGE networking.)
This Proposal
• Software-defined networking (SDN) to the rescue
• Old concept in IT, new paradigm in wireless communication (wireless SDN)
• Decoupling control plane and data plane; Abstraction and virtualization are key!
• SDN will be instrumental in 5G architecture
 In this project focus on SDN at the edge
Research Problems
• How to design a scalable and optimized SDN-atthe-Edge type of architecture for large-scale radio
access networks?
• How to adaptively and elastically distribute
resources* over space and time, and across
various network elements as a function of:
– Users’ heterogeneous traffic requirements
– Flow urgency, latency, congestion
– Fronthaul/backhaul requirements, etc
– Ensuring Incentive compatibility when
• How to optimally decouple control and data
planes based on resource usage and latency ?
• What are the benefits-costs of SDN-aware RANs?
 Research methodologies: learning, game theory,
stochastic optimization, matching theory, GibbsMarkov.
*Resources: spectrum, storage, computing, etc.
Orchestration & Management
Spectrum sharing
Allocate Resources
to Small Cells
Small cell cluster shared
among operators
Work Packages
Thrust 1: Hierarchical Software Defined RAN At the Edge
• Task 1.1: Architecture Design and Abstraction of BS Clusters
• Task 1.2: Protocol Design for the Control Network
• Task 1.3: Locality and Optimization
• Task 1.4: Survivability
Thrust 2: Large Scale Elastic Resource Adaptation
• Task 2.1: Optimal Resource Adaptation Strategy.
• Task 2.2: Enhanced Protocol Design
Thrust 3: Learning Techniques for Resource Allocation
• Task 3.1: HSDRAN-aware Energy efficiency
• Task 3.2: Enhanced Mobility Management
• Task 3.3: Dynamic DL/UL TDD Optimization
• Task 3.4: Stability, Efficiency, and Optimality
Thrust 4: Resource Allocation via Matching Theory
• Task 4.1: Extend to Heterogeneous Types of HSDRAN at the edge, such as Vehicular networks
• Task 4.2: Extend Resource Allocation to Other Matching Theory Approaches, and Game Theoretic
Approaches such as Coalitional Graph Games
• Task 4.3: Extend to other Types of Applications such as Popular Content Distribution
Steering Group Committee
Thus far..
• Mobile data offloading from
cellular networks to alternate
wireless technologies.
• Software defined network (SDN)
at the edge: Dynamically route
the traffic in a mobile network.
• Iterative gather-scatter scheme
(Map-reduce) via ADMM.
L. Liu, X. Chen, M. Bennis, G. Xue and Z. Han, "A
Distributed ADMM Approach for Mobile
Data Offloading in Software Defined
Network, " IEEE WCNC, New Orleans, LA, Mar.
FI-FI collaboration
FI-USA collaboration
Inbound and outbound mobility
Collaborate with other projects due to overlap
L. Liu, X. Chen, M. Bennis, G. Xue and Z. Han, "A Distributed
ADMM Approach for Mobile Data Offloading in Software
Defined Network, " IEEE WCNC, New Orleans, LA, Mar. 2015.
My book 