5G mmWave Communication: Network Architectures and

PlatformsDr. Geng Wu

Chief Scientist, Head of Intel 5G Research and Standards

geng.wu@intel.com

The evolution towards 5G: compute and communications

2

Market demands lead to 5G:

1G-4G were focused on improving

communications,

Communications & processing are

diffused across networks and

mobile devices

Scalability, Versatility, Energy

Efficiency

Capacity, Intelligence and User

Experience

5G = compute & communications

5G requirements on network and device platforms

3

~20x

~25x

1x – 2x

3x – 5x

40x-50x

Last Decade Next Decade

Air Interface

Spectrum

ICT Platforms for

Network and

Devices

• Interference mitigation, beam-forming,

hyper-Tx/Rx, new waveforms/ coding

» Massive computing5x – 10x

• Licensed, unlicensed and shared spectrum.

• Air interface for high frequency bands

» New architecture/ new components

• Cell densification

• Overlay network integrated Multi-RAT

• Underlay network device-to-device, relay

• Intelligent network/edge cloud

» Communication and compute convergence

“5G Network Capacity – Key Elements and Technologies,” IEEE Vehicular Technology Magazine, March 2014

Wireless technology is at a turning point. Future performance is measured in bit/s/Hz/m2/joule.

5G network architecture trend 1:

Overlay networks for high frequency bands and energy saving

Control plane anchored at macro-cell, user plane from small cellsAvoid excessive handovers in traditional heterogeneous networks

Turn off selected cell sites at light traffic hoursIncremental small cell deployment according to traffic growth

Cellular band

mmWave

User experience for now, new RAT @ high frequency bands for future

Coverage for now, network energy saving for future

On OffOff

LTEWi-Fi

4

mmWave

backhaul

5G network architecture trend 2:

Underlay networks for things/wearables and proximity services

One big intelligent and information network

Compute, storage, networking

StorageContext-based

Processing

Smart Apps

D2D Coordinator

D2D UnicastLink

D2D Coordinator

D2D Broadcast

Link

Pico/HomeeNB

MacroeNB

Pico/HomeeNB

D2D UnicastLink

D2D Cluster #1

D2D Cluster #2

D2D Cluster #3

D2D Coordinator

D2D UnicastLink

D2D Cluster #0

Spatial Reuse in One Cell Area

D2D Broadcasting

Multi-hop

D2D Cluster #4

Mesh Data on WiFi DirectDiscovery and control on LTE

5

Many devices, types of devices, connections

Many moving underlay network clusters

5G network architecture trend 3:

Cloud expansion to network edge and to devices

Remote Cloud / Basic Terminal CDN / Cooperative Radio Edge Cloud / Underlay Net Clusters

6Logos shown represent categories of common Internet content and are used for conceptual illustration only.

5G network platforms: cloud evolution of 5G networks

7General Performance disclaimer: For more complete information about performance and benchmark results, visit Performance Test Disclosure

453.85

11.26

265.38

10.2

Interrupt MAX Latency on Intel

Processors Trends (µs)

Xeon Sandy Bridge Xeon Ivy Bridge

OFF the Shelf, KVM

Wind River, OVP KVM

10x

100x

1,000x

10,000x

5G network capacity scaling

• Edge cloud shortens the distance between an user

and his contents/services, matches 5G air interface

capability.

• Data, information, and intelligence intensify around

user. Edge cloud efficiency becomes crucial.

• Cloud RAN architecture plays a pivot role in network

densification and cooperative networking.

• Real time workload for air interface related signal processing

requires complete hardware + software platform solutions.

5G device platforms: virtualization of 5G device and access

8

5G

Device + Access

Virtualization

Time

Co

mp

ute

Pla

tfo

rm

Pe

rfo

rma

nce

Feature

Phone

Smart

Phone

Portable

Wearable

5G

• Future applications require intensified compute and

communication but often smaller device form factor

• 5G high data rate + low latency radio links enable mobile

device + access virtualization across the air interfaces

• 5G services are immersive. Sensing, intelligence and contents

require edge cloud and device + access virtualization

• Breaking computing barrier through communication may

transform consumers’ relationship with network

5G d+a virtualization for scalability, versatility, energy efficiency

mmWave channel modeling

9

reflectors

NLOS Cluster #1

LOS Cluster

NLOS Cluster #2

“Channel Model for millimeter-Wave Communications Based on Geometry Statistics”, IEEE GLOBECOM 2014

0 50 100 150 200-200

-180

-160

-140

-120

-100

-80

Tx-Rx distance (m)

Pat

h l

oss

(dB

)

Reflector Mean Distance = 20 m

Proposed Model

NYU NLOS PL

Lack of data in NYU field measurement for Tx-Rx distance >150

Hi-K Metal Gate

Strained Silicon

3D Transistors

65 nm 45 nm 32 nm 22 nm 14 nm 10 nm 7nm90 nm

Intel 5G research and technology development

10

P5: Efficient Hardware/Software and Platforms

for 5G Network Elements and Devices

Intel Strategic

Research

Alliance on 5G