y-comm – exploring a new architecture for heterogeneous networking · 2015. 7. 31. · wednesday...

Wednesday 13th August Y-Comm: An Architecture for Heterogeneous Networking

Glenford Mapp

Y-Comm – Exploring a New Architecture for Heterogeneous

Networking

Glenford Mapp – Middlesex/CambridgeJon Crowcroft - Cambridge

David Cottingham- Cambridge Fatema Shaikh-Middlesex


Glenford Mapp

A Complete System for Heterogeneous Networking

• In order to build a complete system that – Does seamless vertical handover– Is extensible – seamlessly adds new technology– Is easy to develop new applications

• Requires a lot of work– Can’t do this from scratch– Need to also look at what other people are doing

• Ambient networks, etc• IEEE 802.21, etc


Glenford Mapp

We Need Two Not One!• A framework for the Peripheral network

– Represented by software running on the mobile node, supports:

• Applications, QoS, Vertical Handover, support for several interfaces

• A framework for the Core network– Represented by software running in the

network, supports• Programmable infrastructure, network

management, QoS, Service Platform


Glenford Mapp

The Y-Comm Framework

HARDWARE PLATFORM (MOBILE NODE)

HARDWARE PLATFORM (BASE STATION)

NETWORK ABSTRACTION (MOBILE NODE)

NETWORK ABSTRACTION (BASE STATION)

VERTICAL HANDOVER

POLICY MANAGEMENT

END SYSTEM TRANSPORT

QOS LAYER

APPLICATION ENVIRONMENTS

(RE)CONFIGURATION LAYER

NETWORK MANAGEMENT

CORE TRANSPORT

NETWORK QOS LAYER

SERVICE PLATFORM

CORE NETWORKPERIPHERAL NETWORK


Glenford Mapp

The Peripheral Framework

APPLICATION ENVIRONMENTS LAYER

QOS LAYER

END TRANSPORT SYSTEM

POLICY MANAGEMENT LAYER

VERTICAL HANDOVER LAYER

NETWORK ABSTRACTION LAYER

HARDWARE PLATFORM LAYER


Glenford Mapp

Do we really need the Core Framework?

• Yes– To support the Peripheral networks you need

to change• For client-based handover we need to have access

to network resources– Need a more open architecture

– But we also need an open architecture for other reasons


Glenford Mapp

Open Management Needed• More diversified networks needed

– Present mobile networks are built on a national or international level: -

• just like the old state-owned telecom companies – such as BT

– Difficult to really produce more tailored networks such as regional networks or city-wide networks

• Value added services– Customers will pay for these value-added services

» E.g. weather or traffic news about London


Glenford Mapp

Open Management Needed• So the question is who is really making

money from mobile systems at the moment??

• The answer:– Very few people

• Vodafone, T-Mobile – Spent loads of money on spectrum, they need to grow

bigger to recoup the money spent• These companies are highly vertical institutions

– Same company does hardware, manage the network and also run or tailor most applications


Glenford Mapp

Open Management Needed

• More niche players• Let different players provide different

components of the network but everyone needs to agree on a new framework and related standards

• Compare this to what happened with the PC industry


Glenford Mapp

Open Management Needed• 1960’s – 1980’s

– Main Frame or Minicomputer– Dominated by IBM and DEC

• 1978– Altair the first micro-processor

• Bill Gates and Paul Allen wrote Basic compiler

• 1982– The IBM PC released and that changed the

world


Glenford Mapp

Why did the IBM PC changed the world

• It allowed third parties to get involved and therefore allowing a mature industry to develop with many niche/specialist players – Visicalc, Eudora, etc

• Because of DOS– A broken but open operating system

• But it was the standard– Allowed new people to write applications– You didn’t have to be IBM or DEC


Glenford Mapp

How far are we from an open architecture?

• Far: but there have been developments to try and get us there– Programmable networks

• xBind

• The real problem is that a lot of this never made it to the wireless network infrastructure


Glenford Mapp

Also need a complete Framework

• If you really want to do this properly– you need a new framework

• Key issues– Network management and control– Things you want to support from the

Peripheral Framework• Vertical Handover

– Client-based Handover• QoS, Support for Applications and Services


Glenford Mapp

The Core Framework

SERVICE PLATFORM LAYER

NETWORK QOS LAYER

NETWORK TRANSPORT SYSTEM

NETWORK MANAGEMENT LAYER


NETWORK ABSTRACTION LAYER

HARDWARE PLATFORM LAYER


Glenford Mapp

Hardware Platform Layer

• Similar to Peripheral network – Supports a number of wireless networks– Includes Base-station technologies– Different technological settings

• Technologies in the same box• Technologies in different boxes


Glenford Mapp

Hardware Platform Layer Represented as Vertical Components – different base stations

3G WLAN 802.11

WiMax802.16 UltraWideBand


Glenford Mapp

Cognitive Radio will change this as it will all be done in the same box

WIDE SPECTRUM REECIVER

SOFTWARE DEFINED RADIO

SPECTRUM MANAGEMENT

3G WLAN WiMax

INTELLIGENT SIGNAL PROCESSING

UltrawideBand


Glenford Mapp

Network Abstraction Layer

• Again similar to Peripheral Network– Abstraction that abstracts a number of

different wireless technologies– Also includes base-station controller functions

• Extension of 802.21– Exportable interface so that the base-station

can be remotely controlled– Monitoring and reporting functions


Glenford Mapp

(Re)configurable Layer

• Controls core infrastructure including routers and switches– Programmable hardware

• Routelets and Switchlets• Already used internally in network equipment but

not exported to external systems especially end-devices

• It is needed to support client-based handover


Glenford Mapp

Programmable Networks (PNs)A brief history

• The idea started with the Xunet program– Program that built a research communications

network throughout the US• Sponsored by AT&T• Laboratory without walls

– Xunet 1 used DataKit switches– Xunet 2 used ATM

• Pushed understanding of ATM networks• Looked at management and control issues


Glenford Mapp

Programmable NetworksA brief history

• Columbia University joins Xunet 2 – Led by Aurel Lazar

• Interested in network management issues, etc

• Aurel proposes XBind– Hardware represented by software abstractions– Use that abstraction to build real applications– Similar to an operating system

• Xbind was called a broadband kernel• Instead of devices, it was controlling network hardware


Glenford Mapp

Xbind• So the big idea of Xbind is that we could use it

as a kernel and virtualize network infrastructure through the broadband Kernel

• Layers of XBind– Applications (WWW Server)– Services (e.g. Multimedia services – video conf)– Broadband Kernel – Management Plane– Binding Interface – Abstraction of real Hardware– Real Hardware


Glenford Mapp

XBind – The Layers

Hardware (Real Devices and Real Network)Camera, Computers, Links, Adapters, Displays

Binding Interface Base (BIB)Virtual Switch, Virtual Link, Camera device driver, Display device driver

BROADBAND KERNEL (High Level Abstraction and Service)Device Management, Connection Manager, Router

MULTIMEDIA SERVICESVideo Conf Manager

APPLICATIONSMultimedia Web Services


Glenford Mapp

Extensions of the X-Bind Key Ideas

• Since we can control the hardware virtually using software, the hardware and the software need not be in the same box– Make the hardware box only understand the

interface– Talk to the box over a serial link using a

computer– Control software runs on the computer


Glenford Mapp

X-BIND: The extensions

MANAGEMENTCONTROL OF CONNECTIONS

RAW ATM SWITCHVCI OPERATION

SERIAL LINE


Glenford Mapp

OpenArch and OpenSig

• You need a management protocol to manage the switch

• You also need a signalling protocol – You need to agree on the format of the

commands that go across your serial line• Two forums were established to do this

– OpenArch – management plane– OpenSig – for signalling


Glenford Mapp

Clever stuff by Ipsilon• This method offers us a way of using ATM

transparently in an IP Network• Let’s suppose the control boxes were IP routers,

so they understand IP. • Use IP to do the routing but the actual data

transfer was done using ATM • Proposed by Ipsilon

– Founded by Tom Lyon and Peter Newman– Proposed GSMP for ATM Switch– Made a lot of money


Glenford Mapp

ATM Substrate

IP Router(Edge)

IP Router(Core)

IP Router(Edge)

ATM SWITCH ATM SWITCH ATM SWITCH

IP Connections IP Connections

Data PathSetup Path


Glenford Mapp

Other Key Idea from Xbind

• Why not virtualize not just the interface but virtualize the box as well.

• Concept similar to virtual machines – Vmware and Xen

• So a physical switch can in effect support a number of virtual switches that do not know that they are running on the same hardware


Glenford Mapp

Switch Virtualization

ROUTER CAN ASSOCIATE EACH VIRTUAL SWITCH (VS) WITH A DIFFERENT IP

NETWORK(IP1) (IP2) (IP3) (IP4) (IP5)

VS1 VS2 VS3 VS4 VS5

Switchlets


Glenford Mapp

Switch Virtualization

• First suggested by the people at Cambridge– Ian Leslie, Kobus Van Der Merwe, Sean

Rooney– Developed a switch interface called Ariel– Suggested extensions to GSMP

• CPlane – a company to exploit this idea


Glenford Mapp

What happened to the Programmable Network work

• Got overtaken by active network research– More funding from the US military for ANs– Public research money dried up for PNs

• Was too closely attached to ATM– MPLS, same idea as ATM but based on IP ,

wasn’t keen on open architectures• Related companies died in the dotcom

bust


Glenford Mapp

Why do we need it now?• Because we need resources to do vertical

handover, especially for client-based handover

• We have to get those resources from the network– Channels on base-stations, QoS, etc

• We have never had open interfaces for mobile systems– Base-stations, BSC, MSO, etc


Glenford Mapp

Vertical Handover

POLICY MANAGEMENT LAYERDECISION HANDOVER

(BASE-STATION, 3G, QOS, TBVH)

GPSLocation,

Speed, direction

Connections (QoS)

VERTICAL HANDOVER LAYERACQUIRE RESOURCES

( 3GCHAN, BASE-STATION, QOS)

NETWORK MANAGEMENT LAYERSend to Mobile

TOPOLOGY, RESOURCES, QoS

TBVH New QoSNew IP

RECONFIGURABLE LAYERACQUIRE CHANNEL

(3G, BASE-STATION, QOS)

NETWORK ABSTRACTION LAYERBASE-STATION

CHANNEL ACQUIRED

DO IT

NETWORK ABSTRACTION LAYERDATA CHANNNEL = 3G

3G=ACTIVE WLAN=PASSIVE WiMAX= PASSIVE

3G WLAN WiMax 3G WLAN WiMax

Done

DO IT

L2 eventsMedia Info

CORE NETWORK


Glenford Mapp

The Network Management Layer

• Defines and controls several networks• Each network is controlled by a network

operator• Support for network virtualization and

partitioning– Several networks managed by different

network operators but on the same extended hardware platform

• Will facilitate regional or city-wide operators


Glenford Mapp

Diverse Networks on one Extended Hardware Platform

REGIONALNETWORK 1

REGIONAL NETWORK 4

REGIONAL NETWORK 2

REGIONALNETWORK 3

CITYNetwork

Extended Hardware Platform

CITYNetwork

CITYNetwork

CITYNetwork

N ETWORK CORE


Glenford Mapp

Status

• Very few open management tools– OpenNMS

• Most tools are about network monitoring– Using SNMP– Unable to manage a very large network

• Can’t manage global/national network– Can’t manage several networks in an

integrated way


Glenford Mapp

What we want• Networks to be managed in a flexible way• An interface that defines and manages an

entire network in terms of programmable components, including switchlets, routelets, etc. – i.e. network virtualization

• using software abstraction to control the network• Create new networks (spawning)• Merge and partition networks


Glenford Mapp

NETWORK MANAGEMENT


HIGH-LEVEL FUNCTIONS

NETWORK 1 NETWORK 2 NETWORK 3 NETWORK 4

CREATE MODIFY MERGE JOIN Network TopologiesNetwork Resources

Network Components


Glenford Mapp

Anything Similar??• GENESIS Project

– Columbia University in late 1990’s• Programmable Virtual Network

– Making networks more programmable– Key operations

• Spawning, Profiling and Management• Virtual Network Controller • Virtual Network Manager

– Didn’t get very far – funding dried up


Glenford Mapp

Core Transport System• Concerned about moving data between

points in the core network• TCP/IP is the Transport and Network

protocols for the Internet• Migration of IPv4 to IPv6

– Important for the new frameworK– Need enhancements

• Easier mechanisms to support security– VPNs, Tunnelling, etc


Glenford Mapp

Core Transport in Y-Comm

PERIPHERALWIRELESSNETWORK

CORE NETWORK

PERIPHERALWIRELESSNETWORK

Core Endpoints

QoS, Secure Connection


Glenford Mapp

Core Transport: Making Connections between Core

Endpoints

CORE TRANSPORT LAYER

NETWORK MANAGEMENT LAYER

Make connectionBetween WirelessNetworks

ENDP1, ENDP2, QoS Routing, Security


Glenford Mapp

Network QoS Layer

• Current QoS models– IntServ

• Largely abandoned– DiffServ

• Slow deployment

• Need to explore network QoS models– Negotiation structure between Core Network

and Peripheral Networks


Glenford Mapp

Enhance QoS Functionality

CORE NETWORK QOS

PER FLOW (IntServ), PER CLASS (Diffserv)

Negotiating withPeripheral Networks

ENDPT1 ENDPT2 ENDPT3


Glenford Mapp

Service Platform

• Services developed by third parties• Should be able to configure services on

several networks at the same time• Support for installing services for specific

types of networks– London travel service available to networks

around London


Glenford Mapp

SERVICE PLATFORM LAYER• Installation

– Install over many networks simultaneously– Specifying QoS (minimum SLA)

• Server Advertisement and Subscription• Session management

– Interaction between server and application running on the mobile node

• Supporting Mobility– How is QoS maintained even though the mobile is

moving• Server replication, proxy server support


Glenford Mapp

Service Platform and Core Network

CORE NETWORK QOS

PER FLOW (IntServ), PER CLASS (Diffserv)

Negotiating withPeripheral Networks

ENDPT1 ENDPT2 ENDPT3

Server AdvertisementSubscriptionApplications

SERVICE PLATFORM LAYER

SLA1 SLA2 SLA3 SLA 4

…….


Glenford Mapp

Need to Address Security

• Cannot be a separate Layer in the architecture as there are different kinds of security issues

• Y-Comm defines a 3-layer security architecture

• Comprehensive Design– Security system is integrated into the

structure of Y-Comm


Glenford Mapp

The Y-Comm Framework showing its Security Levels





VERTICAL HANDOVER

POLICY MANAGEMENT


QOS LAYER



NETWORK MANAGEMENT

CORE TRANSPORT

NETWORK QOS LAYER

SERVICE PLATFORM


SAS

NTS

NAS


Glenford Mapp

Security in Y-Comm• Network Architecture Security (NAS)

• Security dealing with the deployment and management of different wireless technology

• Managed from the Policy and Management layers• Network Transport Security (NTS)

– Security dealing with end-to-end transport through Y-Comm

• Done at Layer 5, NAT, IPSec, etc• Service and Application Security (SAS)

– Security dealing with running applications and deploying services

• AAAC, ACLs, User-based security


Glenford Mapp

Security Model: Another layer• Security Layer at QoS Layer • Qos Security (QoSS)

– Security should be viewed as part of QoS– Changes in security should be viewed as

changes in QoS• Other issues:

– Prevent overloading of endpoints• Mobile changing points of attachments due to

vertical handover• Migration of proxy servers to different networks


Glenford Mapp

Another Layer- QoS-based security (QBS)

• Key Issues– You might compromise the security of a

network by trying to satisfy SLAs• Install a server or proxy server which has a lower

security profile than the network that you want to use to migrate the service.

– Monitor SLAs, so can be used to detect Denial-of-Service (DoS) attacks


Glenford Mapp

The Y-Comm Framework showing its Security Levels- New Security Level





VERTICAL HANDOVER

POLICY MANAGEMENT


QOS LAYER



NETWORK MANAGEMENT

CORE TRANSPORT

NETWORK QOS LAYER

SERVICE PLATFORM


SAS

NTS

NAS

QBS

y-comm – exploring a new architecture for heterogeneous networking · 2015. 7. 31. · wednesday...

Documents