1

1

Network Virtualization together Network Virtualization together

with IT Resource Virtualization with IT Resource Virtualization ––

How We Approach It?How We Approach It?

Tuesday 26th October 2010 @ NII, Tokyo

Lecture #4

Dr. Kun Yang

University of Essex, Colchester, UK

2

Agenda

Network Virtualization

PlanetLab

Some EU Projects on Network Virtualization

Test-bed @ UEssex

Q&A

2

3

Virtualization

Virtualization exists practically in every aspect of computing, for instance, � an operating system is a virtualization of bare computer hardware

� a data centre is a virtualization of a group of physical resources such as storage, CPU and the software toolkits operating on top of these resources.

Cloud computing is Internet-based computing, whereby shared resources, software, and information are provided to computers and other devices on demand, like the electricity grid. [wikipedia.org]

Cloud computing also represents a brand-new business model where virtual providers come into existence.

4

Cloud computing – a new paradigm

3

5

Hypervisor

Hypervisor (Virtual Machine Monitor)

Hypervisor Management API

• Create Guest

• Destroy Guest

• Stop/Start/Pause Guest

• Add/Remove Devices

Virtual Hardware API

Memory Disk Network CPU

Virtual Machine 1 Virtual Machine 2

Virtual

Memory 1

Virtual

Disk 1

Virtual

NIC 1

Virtual

CPU 1

Physical Hardware

6

IaaS

PaaS

SaaS

Application

Middleware

App. Dev. & Exe Platform

(e.g., Oracle WebLogic)

…App 1 App 2

skype

player

3rd party apps

Web browser or app-

specific front-end

XaaS for Cloud Apps

CPU, memory, NAS, Networks

Hardware Resource Mngt.

Virtualization & Mngt.

App. Program I/F (API)

XCP

OS, database, servers

Develop Env. simple Complex

Market Scale: small big

(datacentre)(Domain apps)

4

7

Network Virtualization

Network Virtualization is the logical next step after cloud computing (which includes storage/server virtualization).

Network virtualization provides an abstraction layer that decouples physical network devices from business services delivered over the network to create a more agile and efficient infrastructure.

It allows multiple applications to run side-by-side over the same physical network.

� Each virtual network has its own business or service oriented policies while providing the security, availability and performance required for each service.

Virtual networks optimize the manageability and control of physical networks that are shared between multiple applications. � Thus resulting in a quickly deployable, more reliable service.

Enable Agile Business: Accelerate the roll-out of new services and advanced capabilities through automated multi-vendor provisioning.

8

Slicing

5

9

Types of Network Virtualization

Internal network virtualization: network in a box on a single system.

External network virtualization: virtualization real network devices such as switches, routers, etc.

Hybrid network virtualization: e.g., VMware hypervisor + VMware Infrastructure software� The latter is to connect and combine networks in multiple boxes into

an external virtualization scenario.

Node virtualization vs. link virtualization

10

Node Virtualization

Virtual node is the virtual version representing the partition of a single physical node or the aggregation of multiple physical nodes (e.g. L3 router, L2 switch or L1/L0 optical cross connects - OXC).

aggregatingpartitioning

6

11

Link Virtualization

Virtual link is a connection (e.g. a cable between a pair of routers, optical light path, wavelength, sub-wavelength) between one port of a virtual network element to a port of another virtual network element.

partitioning

aggregating

40G

40G

40G

40G

10G30G

20G20G

80G

Wavelength with

fixed bandwidth

Virtual link with

variable bandwidth

12

Network Virtualization Initiatives

USA: GENI (Global Environment for Network Innovations)

EU: FIRE (Future Internet Research & Experimentation)

EU: FEDERICA (Federated E-infrastructure Dedicated to European Researchers Innovating in Computing network Architectures)

Japan: AKARI

GENI is to enhance experimental research in networking and distributed systems, and to accelerate the transition of this research into products and services that will improve the economic competitiveness of the United States. [wikipedia]

7

13

Agenda

Network Virtualization

PlanetLab

Some EU Projects on Network Virtualization

Test-bed @ UEssex

Q&A

Some materials courtesy to PlanetLab website:

http://www.planet-lab.org/

14

Overview

Experimentally-driven research is key to success in exploring the

possible futures of the Internet.

PlanetLab is a global research network that supports the

development of new network services such as distributed storage,

network mapping, peer-to-peer systems, distributed hash tables,

and query processing.

8

15

Sites

Started at 2003, it currently consists of 1133 nodes at 515 sites.

16

A Global View of PlanetLab

Per-node

View

9

17

Long-running Services

Content Distribution

Storage & Large File Transfer

Information Plane

DHT (Distributed Hash Table)

Routing / Mobile Access

DNS

Multicast

Anycast / Location Service

Internet Measurement

Pub-Sub

Management Services

Courtesy to L. Peterson’s PlanetLab Slides

18

Design Goals (1)

It must provide a global platform that supports both short-

term experiments and long-running services.

� services must be isolated from each other

� multiple services must run concurrently

� must support real client workloads

It must be available now, even though no one knows for

sure what “it” is.

� deploy what we have today, and evolve over time

� make the system as familiar as possible (e.g., Linux)

� accommodate third-party management services

Courtesy to L. Peterson’s PlanetLab Slides

10

19

Design Goals (2)

We must convince sites to host nodes running code written

by unknown researchers from other organizations.

� protect the Internet from PlanetLab traffic

� must get the trust relationships right

Sustaining growth depends on support for site autonomy

and decentralized control.

� sites have final say over the nodes they host

� must minimize (eliminate) centralized control

It must scale to support many users with minimal resources available.� expect under-provisioned state to be the norm� shortage of logical resources too (e.g., IP addresses)

20

PlanetLab in Europe (PLE)

Is also called OneLab as funded by EU FP7.

OneLab is an open networking laboratory integrating, testing,

validating and demonstrating new fixed and wireless networking

technologies in real world settings and production environments.

Knits together the best of today's networking testbeds

PanLab: Pan European Laboratory Infrastructure Implementation

NetLab (Use Cases for Interconnected Testbeds and Living Labs): A “user centric design” paradigm will be followed to capture

the stakeholders' (users, operators, industry, SMEs) feedback.

11

21

PlanetLab in Japan (PLJ)

A third one to be federated with PLC and PLE.

Still under construction, led by Professor Aki Nakao at the

University of Tokyo

22

Testbed Federation

Federation refers to a connection between testbeds which allows

the mutual sharing of their resources.

� Develop and promote your testbed

� Delegate user administration

� Offer new facilities to your users

� Join a community of testbed developers

12

23

PlanetLab Software

OneLab's flagship testbed PlanetLab Europe’s federation is

currently either PlanetLab-based or OMF-based.

MyPLC is the software package necessary to install and run a

PlanetLab system. It was created by the PlanetLab team at

Princeton University, as part of the global PlanetLab project.

OneLab in Europe is also involved in the further development

of MyPLC.

24

OMF (cOntrol and Managment Framework)

It is extremely challenging to reproduce experiments in the

networking field -> use OMF to conduct repeatable experiments.

Testbed owners use OMF to make the resources in their testbed

discoverable, to control access to them, to optimize their

utilization through virtualization, and to federate with other

testbeds.

OMF support testbeds with many different types of resources,

such as wired networks, wireless Wi-Fi or WiMAX networks,

servers and mobile devices, as well as wireless sensor networks.

OMF has been deployed and used by researchers on many testbeds

in Australia, Europe, and the USA, with many of them in active

use 24/7.

13

25

Agenda

Network Virtualization

PlanetLab

Some EU Projects on Network Virtualization

Test-bed @ UEssex

Q&A

26

OFELIA Project (EU FP7)

OpenFlow in Europe: Linking Infrastructure and Applications

10 partners (inc. Stanford U), Oct 2010-Sept 2013

Aim: to create a unique experimental facility that allows researchers to not only experiment on a test network but to control the network itself precisely and dynamically.

Strength of OFELIA is its concept of federated or interconnected islands. � A set of five islands creates a diverse OpenFlow infrastructure that allows

experimentation on multi-layer and multi-technology networks provided by the different islands.

It will incorporate the feedback of the user community and extending its reach to other test facilities.

14

27

OpenFlow Concept (1)

OpenFlow is a communications protocol that gives access to the forwarding plane of a switch or router over the network.

Essence: separation of routing decision making from forwarding

� Major Benefit: enabling more sophisticated traffic management than feasible today

Routing

Forwarding

Traditional switch

Routing

ForwardingForwarding(flow

mapping)

Openflow switch

Traffic Traffic Engineering

OpenFlow controller (e.g., NOX)

OpenFlow protocol

28

OpenFlow Concept (2)

In simpler terms, OpenFlow allows the path of network packets through the network of switches to be determined by a software (usually called switch controller) running on a separatemachine.

OpenFlow is considered an enabler of Software Defined Networking.� E.g., enables experimenters to change the behavior of the network as

part of the experiment rather than, if at all, as part of the experiment setup.

Originated from Stanford University, currently has been implemented by Cisco, Juniper, NEC routers/switches.

The current version of the OpenFlow protocol is 1.0 which was released on December 31, 2009.

15

29

Flow Definition

OpenFlow allows the definition of a flow to be:� any combination of L2- L4 packet headers for packet flows

� L0-L1 circuit parameters for circuit flows

Switch

Port

MAC

src

MAC

dst

Eth

type

VLAN

ID

IP

Src

IP

Dst

IP

Prot

TCP

sport

TCP

dportPacket

Flow

In

Port

In

Lambda

Starting

Time-Slot

Signal

TypeL1/L0 Circuit Flows

VCG

30

Virtualization in OpenFlow: FlowVisor

VN 1

VN 2

VN 3

FlowVisor

Controller for VN1

Controller for VN2 Controller

for VN3

Physical Network Infrastructures

λ1, λ2

λ3

λ4

16

31

MANTYCHORE Project (EU FP7)

IP Network as a Service, Oct. 2010-Sept. 2013

MANTYCHORE provides a software implementation and tools forproviding and managing routers and IP networks as services.

� Router as a Service

� IP Network as a Service

By means of these services:

� Infrastructure Provisioning: Physical/logical routers and IP networks

� Connectivity Provisioning: IP networks already configured

� Creation and configuration of IP networks

MANTYCHORE will be deployed over the infrastructure of 3NRENS: HEAnet, RedIRIS , NORDUnet

MANTYCHORE will be used and evaluated by a user communityformed by several research user groups.

32

MANTYCHORE VisionPhysical Router

• Users share the same physical

infrastructure.

• Each Router Instance can be

temporarily owned by a

different User (Router Instances

offered as IaaS).

Physical Link

Router Instance (may be a

physical router)

User site

Logical Link (may be a

full physical link)

Other user’s IP

Network or the

Internet

•IP networks can be made of

Router Instances from different

NRENs

Each User’s IP Network is

represented by a different color

With Prof. Dimitra Simeonidou (dsimeo@essex.ac.uk)

17

33

GEYSERS Project (EU FP7)

GEYSER (Generalised Architecture for Dynamic Infrastructure Services), starting on 1st Jan. 2010 for 3 years

to support ‘Network + Any-IT' resource provisioning; green issues

With Prof. Dimitra Simeonidou (dsimeo@essex.ac.uk)

34

EVANS Project (EU FP7)

End-to-end Virtual Resource Management across Heterogeneous Networks and Services (EVANS) – under negotiation

Mobile operator Mobile operatorBackbone provider

(IP)

Backbone provider

(Optical/GMPLS)

SP 1

SP 2

“vertical” resource management and control

by network infrastructure providers

“Horizontal” resource management

and control by service providers

Virtual network #1

Virtual network #2

18

35

Agenda

Network Virtualization

PlanetLab

Some EU Projects on Network Virtualization

Test-bed @ UEssex

Q&A

36

A Cloud App: Friends on Campus

19

37

Cloud Architecture

38

Inside a Xen-enabled Physical PC

20

39

Testbed under Construction:

XCP+OpenFlow

40

Contact，Q&A

Dr Kun Yang, Reader

School of Computer Science & Elec. Eng. (CSEE),

University of Essex, Wivenhoe Park, Colchester,

CO4 3SQ, UK

Email: kunyang@essex.ac.uk

http://privatewww.essex.ac.uk/~kunyang/