Network Virtualization as an enabler

for Cloud Computing:

a telco perspective
p p

Frédéric DANG TRAN

Orange Labs

20th ITC Specialist Seminar

18-20 May 2009

research & development

France
a ce Telecom
e eco / O
Orange
a ge
„ Orange: unified brand of France Telecom for Internet,
t l i i and
television d mobile
bil services
i
„ France Telecom
„ Number #1 # broadband Internet service in Europe: 13
millions (ADSL) customers
„ Number #3 mobile operator: 122 millions mobile
customers
„ Orange
g Labs:
„ Innovation network of France Telecom
„ 3,800 researchers in 18 laboratories
• 4 labs located in Asia: China, Japan, South Korea
„ Participation to collaborative projects
• Future Internet
Internet, Grid Computing
Computing, Autonomics
research & development
Out e
Outline

„ Context:
„ Telecommunications service landscape
„ IT and Network infrastructucture

„ Network virtualization meets Cloud Computing

„ Service-oriented virtualized resource management
framework
„ Conclusion
C l i

research & development

Telco service landscape
Residential users Business users
Broadband Internet
Access
VPN
VoIP
Online video games VoD
Virtual Server Hosting

IPTV
DVR teleconferencing
email

Network-Storage of user-generated content

(photos, videos etc. IT backup service

„ Different QoS requirements:

„ Latency, throughput, loss
„ Different communication patterns
„ Unicast streaming, multicast streaming
„ New usages and traffic patterns difficult to anticipate

research & development

Andd ttraffic
a c tetelcos
cos a
are
e less
ess happy
appy
with…
„ P2P traffic largest bandwidth consumer on the Internet
„ Huge strain on the network infrastructure

BitTorrent
eDonkey

non P2P traffic

research & development

C a e ges for
Challenges o a Se
Service
ce Provider
o de &
Network Operator
„ Reduce time-to-market for new telco services
„ Better Quality of Experience for end-users
„ Cost reductions
„ CAPEX + OPEX
„ Green
G IT

„ Flexible network architecture

„ Should adapt itself to the service requirements
„ Automation: self
self-* infrastructures

research & development

End-to-end IT&N infrastructure
A/V servers A/V servers
Mail servers Mail servers

Datacenter/ Datacenter/
Service platform Intelligence Service platform
Movingg
towards
The
Service
consumers
Core Network

POP POP

Access network
copper
FTTH

research & development

Out e
Outline

„ Context:
„ Telecommunications service landscape
„ IT and Network infrastructucture

„ Network virtualization meets Cloud Computing

„ Service-oriented virtualized resource management
framework
„ Conclusion
C l i

research & development

Element
e e t virtualization
tua at o
„ Server virtualization VM

„ Several virtual machines each hosting a full application

VM
stack co-executing on a single physical server VM
„ Server virtualization technologies (hypervisors) are now a
commodity

„ Link virtualization
„ Mature techniques: VLAN, MPLS, Lambda…

„ Network router virtualization

„ Commercial Virtual Routers

„ Software Virtual routers on commodity hardware, viable

and cost-effective solution

research & development

Virtualization
tua at o p process
ocess
Virtual end-to-end infrastructure

physical substrate

research & development

 Cloud
C oud Co
Computing,
put g, XaaS
aaS
„ Infrastructure-a-service
„ On-demand
O d d provisioning
i i i off IT resources
„ Programmatic access through “cloud API”

„ Public/commercial cloud vs private cloud

„ Usage-based pricing model

Network
N„ k support if any lilimited
i d to the
h provisioning
i i i off
VPNs
„ Towards
To ards a Net
Network-as-a-Service
ork as a Ser ice model
„ Provisioning of virtual networks along with IT resources
„ Dynamic reconfiguration of the virtual network

research & development

 NV & C
Cloud
oud Co
Computing:
put g p players
aye s

Service Provider/
Broker/mediator
Virtual Network operator

API

Network Network
Cloud IT Provider
Infrastructure Provider Infrastructure Provider

research & development

Service
Se ce provider
p o de use cases

„ Content Delivery Network revisited

„ Custom virtual network optimized for live streaming
„ Processing capacity at the edge (POPs)

„ Massively Multiplayer online game service

„ Intelligent application-specific routing
„ Game server CPU capacity
p y spread
p across multiple
p
service platforms

research & development

Out e
Outline

„ Context:
„ Telecommunications service landscape
„ IT and Network infrastructucture

„ Network virtualization meets Cloud Computing

„ Service-oriented virtualized resource management
framework
„ Conclusion
C l i

research & development

Virtual
tua Resource
esou ce Management
a age e t
„ A generalized architectural & service-oriented
approach
h ffor virtual
i t l resource managementt
„ Requirements:
„ Clear policy/mechanism
C / separation
„ Technology dependence encapsulated

„ Virtual
Vi t l resources as fi
first-class
t l entities
titi
„ Focus of services with end-to-end QoS

„ Provisioning of end
end-to-end
to end (virtual) resources:
• Network
• Computational
• Storage
St

research & development

 Framework building blocks
Service Service Service

M di ti
Mediation Information
Marketplace
Service

Resource
Management Admission Virtual Resource Virtual-to-physical
Control Provisioning resource mapping

Virtual Resource
Factory

Resources Virtual Cluster Virtual Machine Virtual Router Virtual Switch

research & development

 Utility-based
Ut ty based VR p
provisioning
o so g
„ Dynamic
y a c resource
esou ce ((re)allocation
e)a ocat o acacross
oss se
services
ces
according to actual demand
„ Service information:
„ Business-level Service Level Objectives (SLO), e.g.
• Max time to download a video file + nb of downloads /s
„ Performance
P f model
d l provides
id ththe amountt off resources
(net+cpu) required to meet a given SLO
„ Service-specific utility function:

U=f(service demand, resource capacity)

research & development

 m
U global = max ∑ (w .U i i − cos t i )
i= 1

VR Provision
Decision Module

Give/take
Resource
Utility Ua Utility Ub

Perf goals Perf goals

Decision Module Decision Module

Resource Resource
Resource allocation Resource allocation
Usage reconfiguration Usage reconfiguration

Service A SLAa SLAb Service B

research & development

Resource
esou ce marketplace
a etp ace
„ Open resource marketplace supporting
„ multiple
lti l providers
id and
dbbuyers
„ multiple market mechanisms

Pluggable market mechanisms

Publish Bids
Combinatorial Auction
Auction Server
Publish Offers ….

Discover active auctions Market Information

R ti
Retrieve aggregatet market
k t Service
information (price, demand)

research & development

 Co b ato a auct
Combinatorial auctions
o s for
o ttrading
ad g
virtual network resources
„ Allow buyers to place bids on a package of items:
„ Multiple virtual link bandwidth and nodecapacity
„ Combination of network, CPU and storage capacity

„ Bidding language
„ Buyers bids: XOR statements:
• { P1 XOR P2 XOR P3 … XOR Pj }
• Resources = {p
{price,LINKa-b(bw=100Mbps)
( p ) } AND {price,NODE
{p a((cpu=40%)}…
p )}
„ Sellers bids
• Price, capacity of {links, nodes}

research & development

Out e
Outline

„ Context:
„ Telecommunications service landscape
„ IT and Network infrastructucture

„ Network virtualization meets Cloud Computing

„ Service-oriented virtualized resource management
framework
„ Conclusion
C l i

research & development

Co c us o
Conclusion
„ Towards a telco cloud:
„ Integrated virtualization of service plaforms & networks
„ Self-provisioning of services along with their customized
IT&N architecture
„ Cost reduction for an operator internal infrastructure

„ New business opportunities & revenue streams

„ Key issues:
„ Right level of exposition of the physical infrastructure
(Infrastructure Provider API)
„ Reliability,
R li bili security,
i iisolation
l i off virtual
i l services
i

research & development

Thank you

research & development