Resource Management in 5G Transport NetworksDomenico SiracusaFuture Networks (FUN)

Need for a better transport

• New service-level KPIs: demand for massive increase of network capacity

• Operators need to reduce CAPEX/OPEX in a scenario with reduced ARPU and increased needs in term of infrastructure

• Transport resources are vastly heterogeneous

• Need for adaptive, sharable, cost-efficient and cloud-based 5G transport – Integrating fronthaul and backhaul resources

– Cooperating with the access and core networks

11/7/2015 2CTTE Conference

Source: Ericsson mobility report, June 2014

Project

• Develop novel physical and link layer technologies meeting 5G requirements

• Develop a unified data plane for backhaul and fronthaul– Supporting all RAN functional splits

– Supporting required synchronization

– With one versatile frame format

– Requires new packet switch architectures

• Develop a unified control plane for backhaul and fronthaul based on SDN– Common network model

– Common set of API functions

– Resulting in abstraction layer for BH/FH

• Develop SDN apps on top of abstraction– Monitoring and prediction framework

– Backhaul/fronthaul infrastructure planning and dimensioning

– Context-aware resource management (policies, routing, function placement)

– Network-aware media distribution framework

• Evaluate the developed Xhaul technologies integrated on a 5G testbed

11/7/2015 CTTE Conference 3

Xhaul EU projectxhaul.eu

@Xhaul_eu

Physical Infrastructure of Xhaul

• Interconnection Plane– Xhaul Packet Forwarding Elements (XFE) interconnecting a broad set of novel technologies to

create a packet-based network

• General Processing Plane– Xhaul Processing Units (XPUs) that carry out the bulk of the operations

• End-Point Plane– Network Functions and their interconnections with the other planes

11/7/2015 4CTTE Conference

Functional Architecture of Xhaul

• Technology abstraction and integration layer:– Integration of the different technologies (including fronthaul and backhaul) in a common

packet network

– based on technology abstraction, unified framing and common data, control and management planes

• Xhaul Services Layer (e.g. Multi-tenancy)

11/7/2015 5CTTE Conference

Xhaul Control Infrastructure (XCI)

• The Xhaul Control Infrastructure is in charge of:

– Controlling the forwarding in the network

– Assigning RRHs to virtual BBUs based on network capacity, multi-tenancy contracts and functional splits

– Allocating VNFs, such as BBUs, CDN, etc.

– Orchestrating the networking and IT resources

– Optimizing the energy consumption of the network

– …

• It is the brain of the Xhaul architecture

11/7/2015 6CTTE Conference

High-level vision of the XCI

• Leveraging SDN and NFV standards

11/7/2015 7CTTE Conference

XCI SDN architecture

11/7/2015 8

Net core services

SB plugins

XFE XFE XFENode 1 Node 2 Node N

Agent Agent Agent Agent Agent Agent

Driver X Driver Y Driver Z

Proprietary i/f Proprietary i/f Proprietary i/fProprietary i/f

Open, standard i/f (e.g. OF + extensions)

Open, standard i/f (e.g. OF + extensions)

Open, standard i/f (e.g. OF + extensions)

Inventory Flows actions Statistics

SBI (D-CPI)

Net Apps....Analytics for

monitoring

APIs

REST APIs

Multi-tenancy net virtualization

TopologyTopologyTopology

VIM (Virt InfrMan)

XC

I

....TV Broadcasting/Multicasting

Energy Mngt & Monitoring

Xhaul Resource Manager

Content Delivery Network (CDN)

XH

AU

L A

pp

s

Path Computation Engine

XHAUL ABSTRACTION

Xhaul Resource manager

11/7/2015 9

RAN CORE

Link- Bandwidth- Available resources- Delay (average, var)- SINR or SNR (pack loss)

Node- Index + Location- Type- Computation/Storage capacity- Switching Matrix- Functions

Demand- Features (uni/multicast,

survivability, …)- Endpoints- Expected load- Tolerated delay- Max Pack loss rate- User mobility

CTTE Conference

• Dynamic resource allocation– Selection of the most appropriate technology

– Function placement

– Traffic offloading

• Network adaptation– Modification of physical layer parameters

– Handling of link failure or degradation

But…

11/7/2015 10

What about the other side of the core?

Is the optimization of the Xhaul enough?

Transport Network shall be able to satisfy end-to-end service-level KPIs

Most of the traffic passes through the Internet

Applications are the driver

• Two network applications:– Content Delivery Network (CDN): per-user distribution of video contents,

typically on-demand

– TV broadcasting: broadcasting of video contents, typically live events

11/7/2015 CTTE Conference 11

Source: Ericsson mobility report, 2014

Case study: applications and requirements

• CDN– Unicast connections

• Origin to replica server– Very High BW (massive contents)– No strict latency requirements

• Replica server to customer– Moderate BW– No strict latency requirements

• TV broadcasting– Multicast connections

• Site to production (over IP)– Very High BW (e.g. 800 Mbit/s)– No packet loss, low jitter

• Production to customer– High BW (40 Mbit/s)– No packet loss, low jitter

11/7/2015 12

Need for end-to-end application-centric service in the transport

CTTE Conference

The approach

• Applied to our case:– CDN:

• Shared optical connection

• Optical restoration

– TV broadcasting:• Dedicated optical connection (low latency)

• IP protection

11/7/2015 13

Today’s Transport Networks

Application layer

Grooming layer

Fixed transport layer

App specific requirements

Aggregate requirements

Low bandwidth

High bandwidth

ACINO

Application-aware transport layer

App specific requirements

Grooming layer

Application layer

Low bandwidth

Medium / High bandwidth

App-class requirements

Multi-layer multi-technology (IP/Optical) network orchestration

ACINO EU projectwww.acino.eu@acinoH2020

CTTE Conference

Transport network: end-to-end vision

• Multi-layer optimization including all transport (L0 to L3) layers

• Hierarchy of controllers11/7/2015 14

CONTENT

INTERNETIP/OPTICAL

REPLICA

ORIGIN

MOBILEXHAUL + RAN

MOBILEXHAUL + RAN

APPLICATIONS (CDN/TV BROADCASTING/…)

INTENT PROVIDER INTENT PROVIDER INTENT PROVIDER INTENT PROVIDER

SERVICE LEVEL ORCHESTRATION

ACINO network orchestratorXHAUL XCIDC CTRL

RAN CTRL TECH1

CTRLIP CTRL

TECH1 CTRL

OPTICAL CTRL

CTTE Conference

Conclusions

• As of today, fronthaul and backhaul are separate and incompatible network segments

• project addresses this issue by proposing integrated control and data planes

– Enabler for dynamic resource management in 5G transport networks

• Applications require end-to-end perspective

– Interoperation with the core/metro transport is needed

– Synergy with project (application-centric transport)

11/7/2015 15CTTE Conference

Thank you for your kind attention!!domenico.siracusa@create-net.org@custoz