Optical Networks

CS294-3: Distributed Service Architectures in Converged Networks

George Porter Tal Lavian

Feb. 5, 2002

EECS - UC Berkeley

Overview
Physical technology, devices How are optical networks currently deployed? Customer-empowered networks

New applications, ways of doing business How does this change the big picture? How do we do it? What are the challenges? Payoffs?
EECS - UC Berkeley

Feb. 5, 2002

Overview
Physical technology, devices How are optical networks currently deployed? Customer-empowered networks

New applications, ways of doing business How does this change the big picture? How do we do it? What are the challenges? Payoffs?
EECS - UC Berkeley

Feb. 5, 2002

Why optical?
Handle increase in IP traffic
Moores law doesnt apply here 1984: 50Mbps, 2001: 6.4Tbps

Reduce cost of transmitting a bit

Cost/bit down by 99% in last 5 years

Enable new applications and services by pushing optics towards the edges
Feb. 5, 2002 EECS - UC Berkeley

Fiber capabilities/WDM
(Timeslots) (OC12,48,192)

Wavelengths (Multi Tbps)

Wavebands

Fibers (100+) Cable

Feb. 5, 2002

Wavelengths can be time-division multiplexed into a series of aggregated connections Sets of wavelengths can be spaced into wavebands Switching can be done by wavebands or wavelengths 1 Cable can do multi terabits/sec

EECS - UC Berkeley

Internet Reality
Data Center
SONET

SONET

DWD M

DWD M

SONET SONET

Access
Feb. 5, 2002

Metro

Long Haul
EECS - UC Berkeley

Metro

Access

Devices
Add/Drop multiplexer Optical Cross Connect (OXC)
Tunable: no need to keep the same wavelength end-to-end Switches lambdas from input to output port

For transparent optical network, wavelengths treated as opaque objects, with routing control brought out-of-band
Feb. 5, 2002 EECS - UC Berkeley

Overview
Physical technology, devices How are optical networks currently deployed? Customer-empowered networks

New applications, ways of doing business How does this change the big picture? How do we do it? What are the challenges? Payoffs?
EECS - UC Berkeley

Feb. 5, 2002

Overview of SONET
Synchronous Optical Network Good for aggregating small flows into a fat pipe Electric endpoints, strong protection, troubleshooting functionality
Feb. 5, 2002

OC3

OC48 OC48

SONET

EECS - UC Berkeley

Todays provisioning
Anywhere between months to minutes
Semi-automatic schemes Much like old-style telephone operator

The fact is there are tons of fibers underground, but they are not organized in a way where you can utilize their full potential
Feb. 5, 2002 EECS - UC Berkeley

 Make the network intelligent On-demand bandwidth to the edge of the network New applications
Disaster Recovery Distributed SAN Data warehousing

Drive to autoswitched network

Big Pipes on Demand

Optical VPN Feb. 5, 2002 Grid Computing

Backup Bunkers (no more tapes) Download movies to movie theaters Site replication
EECS - UC Berkeley

Overview
Physical technology, devices How are optical networks currently deployed? Customer-empowered networks

New applications, ways of doing business How does this change the big picture? How do we do it? What are the challenges? Payoffs?
EECS - UC Berkeley

Feb. 5, 2002

Customer empowered nets

Huge bandwidth to the enterprise
The curb The house The desktop

End hosts can submit requirements to the network, which can then configure itself to provide that service Issues of APIs, costs, QoS
Feb. 5, 2002 EECS - UC Berkeley

Changing the big picture

Now the converged network looks different Dial-up bandwidth has huge implications Pushing bandwidth to the edges of the network
Affects service placement, for example

Feb. 5, 2002

EECS - UC Berkeley

Bandwidth at the edges

Services placed there (ServicePoP) Need to connect services to customers and other services Metro networks
Use of Ethernet as low cost/flexible mechanism

Eventually fibers to pcmcia?!

Feb. 5, 2002 EECS - UC Berkeley

Protocol and Services on Edge Devices

Handle Protocol New Services

Internet

Access

Access

Feb. 5, 2002

EECS - UC Berkeley

ServicePoPs
ServicePoPs act as intermediary between service provider and customer Connectivity between ServicePoP and customer more important than provider to customer Feature is very fast infrastructure
Feb. 5, 2002 EECS - UC Berkeley

Metro networks
Interim step: services in servicePoPs Tap into fast connections here for enterprises Use of Ethernet as protocol to connect the enterprise to the MAN Avoid need for last mile for certain applications/services
Feb. 5, 2002 EECS - UC Berkeley

Amazon.comvs-Amazon.co.uk
Amazon.com

One site wants to do a software upgrade Reserve 100Gbps for outage time Send entire database over at outage time, reroute all customer requests to other site When outage is over, transfer all data back to original site
EECS - UC Berkeley

Amazon.co.uk
Feb. 5, 2002

Movie Distribution
Each movie theater in a large area (SF, New York, Houston) requests 1 hour of bandwidth a week (OC192) All movies transferred during this time Efficient use of expensive but necessary fat pipe
Feb. 5, 2002 EECS - UC Berkeley

New type of businesses

Data warehousing: no more mailing tapes Have tape vaults with gigabit connectivity Data is sent optically to destination, where it is written to magnetic tape

Feb. 5, 2002

EECS - UC Berkeley

How to do it
Generalized Multiprotocol Label Switching (GMPLS) UNI: user-to-network interface as API to specify requirements, service requests NNI: network-to-network interface acts as API between entities for service composition/path formation
Feb. 5, 2002 EECS - UC Berkeley

How to do it
Interdomain? Wavelength selection/routing Exchange info
Connectivity Wavelengths Qos, bandwidth requirements Switching instructions
EECS - UC Berkeley

Feb. 5, 2002

Canaries approach
OBGP (Optical BGP) Routers spawn virtual BGP processes that peers can connect to By modifying BGP messages, lightpath information can be traded between ASes

Feb. 5, 2002

EECS - UC Berkeley

1)
BGP OPEN

BGP

OPEN message sent to router with information about optical capabilities

A virtual

BGP process is

AS 123

spawned
OXC AS 456
A BGP

2)
Virtual Router BGP OPEN AS 123 OXC AS 456

session is initiated independently with new BGP process

The

virtual process (running on the router) configures the OXC to switch the proper optical wavelengths

What is ASON?
The Automatic Switched Optical Network (ASON) is both a framework and a technology capability. As a framework that describes a control and management architecture for an automatic switched optical transport network. As a technology, it refers to routing and signalling protocols applied to an optical network which enable dynamic path setup. Recently changed names to Automatic Switched Transport Network (G.ASTN)
Feb. 5, 2002 EECS - UC Berkeley

Optical Network: Today vs. Tomorrow

Applications Protection Topology Management DS3 STS-n STS-nc OC-48T, (OC-192T) 1GE (134Mb/s) 140Mb/s VC-4 VC-4-nc NUT Extra Traffic Broadcast VC-4-nv 10GE Flexible i/f Billing method (distance, time, bw, QoS) Asymitric bw connections Point-to-multipoint - sequential 2F/4F BLSR Matched Nodes Head end ring prot. NUT (non-preemptive unprotected traffic mixed with protected in ring/linear) Unprotected (extra traffic) Protection SW time Clear P =60ms With ET=160ms MN = 250ms

Today

2F/4F BLSR Linear 1+1 1:n Path protection

Provisioned path connection Trail management across multiple rings Multiple product

Tomorrow

Mesh Port connectivity - unconstrained - arbitrary

Wider range of SLA capability Path diversity verifiable Scalable to large NW size

Auto discovery of NW configuration Connection provisioning of paths over unconstrained line topology No pre-provisioning of connections? User signaling i/f for connection provisioning Scalable to very large NW Fast connection establishment <2s Resource (bw) management and monitoring

Optimized IP application - current driver for transparent NW

Additional SLA capability

Mesh network

Auto connection & resource mgnt

ASON value added

EECS - UC Berkeley

Feb. 5, 2002

ASON Network Architecture

Integrated Management

ASON control plane

OCC

OCC UNI

NNI

OCC

OCC IrDI_NNI

User signaling

CCI

Clients e.g. IP, ATM, TDM

GHCT NE

GHCT NE Transport Network

GHCT NE

Clients e.g. IP, ATM, TDM IrDI

Legacy Network
GHAT NE: Global High Capacity transport NE ASON: Automatic Switched Optical Network OCC: Optical Connection Controller IrDI: Inter Domain Interface

Interfaces: UNI: User Network Interface CCI: Connection Control Interface NNI: ASON control Node Node Interface

Feb. 5, 2002

EECS - UC Berkeley

ASON Layer Hierarchy

Domain B

Network Layer
Domain A Domain C

Domain E

Domain D

Domain

Domain/Region Layer

Fibers Conduit 1 Conduit 2

Conduit Layer Fiber Layer

l1

Feb. 5, 2002

EECS - UC Berkeley

ln

l Layer

Resilient packet ring (802.17)

Put lan on top of man 50ms protection

Feb. 5, 2002

EECS - UC Berkeley

The Metro Bottleneck

Other Sites

Access

Metro

End User
Ethernet LAN
IP/DATA 1GigE
Feb. 5, 2002

Access
T1 DS1 DS3

Metro
OC-12 OC-48

Core
OC-192 DWDM n x l
10GigE+

LL/FR/ATM 1Gig+ EECS - UC 1-40Meg Berkeley

RPR - Expanding the LAN to the MAN/WAN

MAN/WAN

LAN
LAN in the MAN Paradigm
Distributed Switch

Low Cost Simplicity Universality

Feb. 5, 2002

Low Cost Simplicity Universality EECS - UC Berkeley

Scalability Reach Robustness

What is RPR?
Ethernet networking on Optics (STS-Nc)

Ethernet Frame

Ethernet Frame

Ethernet Frame Ethernet Frame

Ethernet Ethernet Ethernet Frame Frame Frame

STS-N Envelope

Feb. 5, 2002

EECS - UC Berkeley

Scalable Bandwidth and Services

STS-N
TDM
VTs VTs VTs VTs 1000M

OC-3 / 12 / 48 / 192 STS-Nc

Ethernet

10M 300M 500M 1M 80M

Feb. 5, 2002

EECS - UC Berkeley

Network & Customer Management

Customer Ethernet Ports

Feb. 5, 2002

Customer Privacy through managed Virtual LANs (802.1Q tags) Customer Agreements through flow attributes (802.1p prioritized queues and EECS - UC traffic policing)Berkeley

Move to optical
The key is to find a way to use the infrastructure that we have available in an efficient manner What services are available? What can we do? Challenges?

Feb. 5, 2002

EECS - UC Berkeley

The Future is Bright

Feb. 5, 2002

EECS - UC Berkeley