radio access network (ran) virtualization
TRANSCRIPT
Mobile Cloud Networking FP7 European
Project: Radio Access Network as a Service
Dominique Pichon (Orange)
4th Workshop on Mobile Cloud Networking
19.06.2014, Lisboa, Portugal
Optical
switch
BBU-pool
RAT 1
BBU-pool
RAT 2
BBU-pool
RAT N
WC-Pool
(in a data centre)
© 2012-2015 MCN. All rights reserved. / Page 2
Outline
Motivations and Objectives
MCN Quick Overview
Challenges for an on demand RAN
A PaaS Approach for RANaaS
Conclusion
© 2012-2015 MCN. All rights reserved. / Page 3
Outline
Motivations and Objectives
MCN Quick Overview
Challenges for an on demand RAN
A PaaS Approach for RANaaS
Conclusion
© 2012-2015 MCN. All rights reserved. / Page 4
Motivations for Virtualizing
Networks
Flexible Network Operations
Flexible resource allocation
analytics tools for dimensioning the network
Automated Network operation
auto-scaling
less trouble shooting by use of automatic configurations and isolation between tenants
Faster Speed of Time to Market
Cost efficiency
sharing hardware => reduced power and space consumption
non proprietary hardware
installation, maintenance and removal costs reduction
flexible and automated network operations
multi-tenancy
© 2012-2015 MCN. All rights reserved. / Page 5
Motivations: Focus on RAN
Today’s Radio Access Networks (RANs) have a large number of Base
Stations (BSs), of multiple Radio Access Technologies (RATs), of high
power consumption and
cost (CAPEX/OPEX).
Explosive capacity needs vs. falling revenues per user.
RAN is typically dimensioned for the busy hour;
still, offered load varies drastically, with large periods of low utilisation.
BSs
BSs
BSs
Backhaul
Network
Core
Network
© 2012-2015 MCN. All rights reserved. / Page 6
Outline
Motivations and Objectives
MCN Quick Overview
Challenges for an on demand RAN
A PaaS Approach for RANaaS
Conclusion
© 2012-2015 MCN. All rights reserved. / Page 7
The EU FP7 European Mobile Cloud
Networking Project
European collaborative project FP7.
Started in Nov. 2012. End: Oct. 2015
http://www.mobile-cloud-networking.eu/site/
Who?
Objective: to offer on demand a mobile network
© 2012-2015 MCN. All rights reserved. / Page 8
MCN FP7 Objective: RAN as a
Service
To offer a cloud-based RAN as a Service (RANaaS): Heterogeneous, virtualised and multi-tenant RAN, following cloud
principles (infrastructure sharing, elasticity, on-demand, pay-as-you-go).
Dynamically adapted to geographic and temporal load variations and traffic type.
Centralised RAN processing architecture, based on virtualised pools of Base Band Units (BBU-pools) on datacentres.
Flexible relationship between Remote Radio Heads (RRHs) and BBUs, linked by a high bandwidth and low-latency optical fronthaul network.
Fronthaul
Transport
Network
Central Office
Datacentre
BBU-pool
RAT 2
BBU-pool
RAT 1
BBUs
...
RRHs
RRHs
RRHs
Backhaul
NetworkCore
Network
© 2012-2015 MCN. All rights reserved. / Page 9
Outline
Motivations and Objectives
MCN Quick Overview
Challenges for an on demand RAN
A PaaS Approach for RANaaS
Conclusion
© 2012-2015 MCN. All rights reserved. / Page 10
Challenges for RANaaS
Fronthaul transport network (between BBUs and RRHs)
High bit-rate CPRI links:– Site with 3 RRHs (LTE, 20MHz) requires 7.4 Gbit/s link.
Low latency:
– Maximum round trip delay of 150µs (~15 km optical fibre).
Jitter and synchronisation:– Stringent requirements for frequency and phase synchronisation
Real-time performance
base stations requires strict real time performance
– e.g., air interface LTE L2 packets to be handled in 3.66 ms
© 2012-2015 MCN. All rights reserved. / Page 11
Challenges for RANaaS
Offer cloud based multi-tenant on demand RANaaS:
Offer SLA-guaranteed connectivity to multiple tenants (MVNOs)
through an elastic on-demand allocation of resources.
on demand => dynamic spectrum management
– Detect and predict resources usage to optimize the offered services.
Datacentre
D-RoF
Optical
switch
RRHs
RRHs
RRHs
BBU-pool
RAT 1
BBU-pool
RAT 2
BBU-pool
RAT N
WC-Pool
(in a data centre)
Fronthaul
Transport
NetworkRadio
Interface
Backhaul
Transport
Network
Core Network
Radio Access Network
Radio
resources
Fibre optic
resources
Computational
resources
BBU-pool
RAT 2
BBU-pool
RAT 1
BBUs
...
Virtual radio
resources
MVNO
Gold
2 Mbps to
all users
MVNO
Cheap
Best effort
for all users
RRHs
RRHs
Remote
Radio
Heads
(RRHs)
© 2012-2015 MCN. All rights reserved. / Page 12
Worker
Worker
Challenges for virtualized RAN
Elasticity
by means of horizontal scalability and load balancing
LTE air interface protocols are not RESTful
Load
BalancerWorker
=>so you cannot simply add an
HTTP load balancer and add VM
hosting http servers for handling
LTE workload
=>need for redesigning the base
station software to make it
horizontally scale
© 2012-2015 MCN. All rights reserved. / Page 13
Outline
Motivations and Objectives
MCN Quick Overview
Challenges for an on demand RAN
A PaaS Approach for RANaaS
Conclusion
© 2012-2015 MCN. All rights reserved. / Page 14
A Platform as a Service for Mobile
Network
PaaS allows service providers to host and execute web applications using
third-party managed servers
e.g., Google App Engine, OpenShift
A Mobile Network using PaaS allows mobile network service providers to
host and execute network functions using a third party infrastructure
© 2012-2015 MCN. All rights reserved. / Page 15
Platform as a Service by Google
[Octo.com]
A user uses the service via the web
--e.g., a web app to manage post-its
with your wife/husband/child(ren)
- hosts the application software,
runs it, provides means to improve it
and log s on the service delivery
- and charges you
the developer
-designs the service using the
Google SDK and pushes the related
code to Google App Engine
© 2012-2015 MCN. All rights reserved. / Page 16
Platform as a Service for RANaaS
© 2012-2015 MCN. All rights reserved. / Page 17
RANaaS: Architecture
Design cloud-based RANaaS architecture, supported by a set of functional elements:
based on MCN architecture reference model– Service manager (SM): interface to client,
– Service Orchestrator (SO): In charge of the RANaaS lifecycle
– Cloud Controller (CC): Providing physical resources for RANaaS
using support services– Monitoring, charging, SLA management, analytics
Plus RAN virtualized network functions– LTE Base station protocol stacks broken down to enable
– fine mapping between layers processing requirements and physical resources
– elasticity by balancing the load from UEs on different servers
© 2012-2015 MCN. All rights reserved. / Page 18
RANaaS: Architecture
© 2012-2015 MCN. All rights reserved. / Page 19
RANaaS: Service lifecycle
Design
Deployment
(load components for an SLA chosen by a client)
Provisioning
(customise)
Runtime management
(scale up/down)
Disposal
(destroy)
© 2012-2015 MCN. All rights reserved. / Page 20
MCN architecture under development
RANaaS components
designed by RANP to
instantiate RAN on demand
Support services
used by RANP
VNFs deployed by
RANP for its EEU
Managing lifecycle
of base stations
Use to configure the
base stations
© 2012-2015 MCN. All rights reserved. / Page 21
Outline
Motivations and Objectives
MCN Quick Overview
Challenges for an on demand RAN
A PaaS Approach for RANaaS
Conclusion
© 2012-2015 MCN. All rights reserved. / Page 22
Cloud-based RAN is a novel architecture that aims at
bringing cost and efficiency benefits from the cloud
computing model.
MCN aims at offering elastic, scalable, and on-demand
RANaaS, dynamically adapted to geographic and
temporal load variations.
It faces several challenges, regarding:
Fronthaul
Radio resource management
Real-time performance
Scalability
The approach is being developed for further evaluation
Conclusion
Thank You
© 2012-2015 MCN. All rights reserved. / Page 24
Fronthaul: Scenarios and Solutions
Fibre availability is already a requirement
in Orange France LTE backhaul:
90% of sites in dense areas have fibre.
96% of links are shorter than 10km.
In urban areas, up to 28 cell sites are
connected to one Central Office.
Several solutions can be
considered for the fronthaul:
One fibre per RRH.
One fibre per site, shared by RRHs with
Wavelength Division Multiplexing (WDM).
Alternatively, microwave links can be used for small radio sites.
© 2012-2015 MCN. All rights reserved. / Page 25
Fronthaul: Challenges
Fronthaul transport network (between BBUs and RRHs):
Digital Radio over Fibre (D-RoF).
Using typically the Common Public Radio Interface (CPRI) standard.
Digitation requires high bit-rate CPRI links:
Site with 3 RRHs (LTE, 20MHz) requires 7.4 Gbit/s link.
Site with 15 RRHs (LTE-A (2 bands), 3G (2 bands), 2G (1 band))
requires up to 20Gbit/s link.
Low latency:
Maximum round trip delay of 150µs
(~15km optical fibre).
Jitter and synchronisation:
Stringent requirements for frequency
and phase synchronisation.
D-RoF
D-RoF
Optical
switch
BBU-pool
RAT 1
BBU-pool
RAT 2
WC-Pool
(in a data centre)
Optical
switch/router
Fronthaul
Transport
Network
BBU-pool
RAT 2
BBU-pool
RAT 1
BBUs
...
RRHs
RRHs
RRHs
Centra Office
(datacentre)
© 2012-2015 MCN. All rights reserved. / Page 26
Fronthaul: Research
Impact of Mobile Cloud Networking on Fronthaul.
Dimensioning and optimisation of the
optical link between the BBU-pools and RRHs:
Based on realistic network configurations.
To optimise the architecture and have an end-to-end picture.
Reduce required bit-rates on CPRI links:
Several propositions for function splitting between
RRH and BBU will be taken into account.
Support interoperability, multi-technology (2/3/4G, WiFi)
and RAN sharing between different operators.
© 2012-2015 MCN. All rights reserved. / Page 27
With few users, only 3 RRH-BBU are sufficient to cover
the service area and provide the requested capacity.
Load Balancing: Motivation Scenario
BBU Plane
End-user Plane
RRH Plane
BBUpool 1
RRH1
RRH2
RRH3
ControllerControl Plane
BBUpool 1
BBU1 (20% load)
BBU2 (30% load)
BBU3 (20% load)
© 2012-2015 MCN. All rights reserved. / Page 28
When the number of users increases, more RRHs are
activated, and associated BBUs instantiated on a single
BBU-Pool, to provide the requested capacity.
Load Balancing: Motivation Scenario
BBUPool 1
BBU1 (10% load)
BBU2 (10% load)
BBU7 (10% load)
BBU6 (10% load)
BBU5 (10% load)
BBU4 (10% load)
BBU3 (10% load)
RRH1RRH5
RRH7RRH2
RRH3RRH4
RRH6
BBU
Plane BBUpool 1
End-user
Plane
RRH
Plane
Controller
Control
Plane
© 2012-2015 MCN. All rights reserved. / Page 29
Load Balancing: Motivation Scenario
With a highly loaded network, some BBUs must be
instantiated on a different BBU-Pool, as the load exceeds
the BBU-Pool capacity.
End-user
Plane
RRH1RRH5
RRH7RRH2
RRH3RRH4
RRH6
BBU
PlaneBBUpool 1
BBUpool 2
RRH
Plane
ControllerControl
Plane
BBUPool 1
BBU7 (70% load)
BBU6 (60% load)
BBU5 (80% load)
BBUPool 2
BBU1 (60% load)
BBU2 (80% load)
BBU4 (80% load)
BBU3 (70% load)
© 2012-2015 MCN. All rights reserved. / Page 30
Load Balancing: Challenges
Virtualisation of a sub-set of BBU functions, which can
be scaled up and down.
Quantification of the relation between load and
processing needs.
Map user load into required processing of virtualised BBU functions.
Balance RAN processing load between BBU-pools:
Dimension the number of cell sites one BBU-pool can manage.
Account for temporal and geographical variations of load.
Take into consideration radio, fronthaul and datacentre capabilities and
constraints, as well as SLAs (and associated QoS requirements).
Efficiently allocate processing resources (locally or remotely).
Balance load between multiple RATs.