1

Routing and Resilience in Future Optical Broadband Telecommunications Networks

21st January 2004

Andrew S. T. LeeSupervisor: Dr. David Harle

Broadband and Optical Networks Research GroupDept. of Electronic and Electrical Engineering

University of Strathclyde, Glasgow, UK

2

Introduction

All-optical physical layer using Optical Packet Switching

Synchronous operation at 100 Gbit/s

Higher layer Ethernet frames or IP packets are mapped onto multiple optical cells

IP / Ethernet IP / Ethernet

Transmitter Receiver

ConvergenceSublayer

ConvergenceSublayer

Segmentationand Reassembly(SAR) Sublayer

Segmentationand Reassembly(SAR) Sublayer

ELECTRONIC CLIENT LAYERS

AdaptationLayer

OPTICAL PACKET-SWITCHED LAYER

3

2x2 Optical Buffered Switch

Cells are queued using optical buffer Series of 2x2 optical switches and fibre delay lines Logarithmic scalability – discrete buffer lengths Emulates a 2x2 switch with non-optimal delay

2x2SW

n/221

.............

4

1log 2 n switches & delay linesB(n): chain ofmn 2,...,8,4,2,1

U

U

U

L

L

L

upperinput

lowerinput

upperoutput

loweroutput

lowerqueue

upperqueue

NL,2L, 2L

2U

UU,L2U, UU,L NL,U,U,L L,2LU,L

NU,2U,

2n

1n2n12

n10

NL,2L, NL,2L,

NU,2U, NU,2U,

NL,U,U,L

2U2U2U

2L2L2L

LU,L

4

Physical Implementation

Buffer Control

nintegratio SiSiO usingon wafer linedelay spiral 2

MMI couplers

Amplifiers

Modulators

Q Q Q Q Q QQ Q

5

Synchronization, Control and Header Modification

Synchronizer

LocalClock

Routing ProcessorBuffer Scheduler

SynchronizerControl

BufferControl

2x2 SharedBuffer SwitchTap

hop_count_UI/LIdefl_count_UI/LIdest_UI/LI

hop_count_UO/LOdefl_count_UO/LO

UI

LILO

UO

HeaderRecovery

O/E

HeaderModification

E/O

HeaderUpdate

Electronic Control

6

Self-Routing Networks

Each switch makes a fixed routing decision based on packet destination and other header information

Queue contention is resolved using deflection routing (different arbitration heuristics)

Ring(BS)1 2 3 4

1I

2I 2O

1OGB1 GB2Header Payload

1ns100 bit

0.32ns32 bit

1ns100 bit

3.45ns344 bit

7

Self-Healing Ring Architecture

Protection against node and link failures using additional switches

8

4-Switch Cyclic Node Design

Drop(BS)

RingUpper(BS)

RingLower(BS)

Add(A3/4)

1I

2I2O

1O

1I

2I2O

1O1I

2I2O

1O

1I

2I 2O

1O

Intra-nodal system and diverse routing reduces network congestion

Improved network scalability and operation for higher loads

9

Traffic Studies

Metrics – buffer depth, packet loss probability, end-to-end delay, ring size, etc.

Bernoulli traffic (results shown) Used to contrast different network topologies

Bursty traffic models Ethernet/IP frames are carried over the network Impact on packet reordering and sequence

integrity Interconnected rings

10

Packet Loss Probability

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

1.0E-03

1.0E-02

1.0E-01

1.0E+00

0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90

Source Load

Pack

et L

oss

Prob

abili

ty

1fiber-1sw 1fiber-2-linear 1fiber-3-linear1fiber-3-cycle 2fiber-2-ardr 2fiber-2-alt2fiber-3sw -uni 2fiber-3sw -bi 2fiber-4sw -type1-uni2fiber-4sw -type1-bi 2fiber-4sw -type2

11

End-To-End Delay

0

20

40

60

80

100

120

0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60

Source Load

End-

To-E

nd D

elay

1fiber-1sw1fiber-2-linear1fiber-3-linear1fiber-3-cycle2fiber-2-ardr2fiber-2-alt2fiber-3sw -uni2fiber-3sw -bi2fiber-4sw -type1-uni2fiber-4sw -type1-bi2fiber-4sw -type2 tw o-fiber, unidirectional

tw o-fiber, bidirectionalsingle-f iber, unidirectional

12

Conclusions

Multi-switch, bidirectional ring architectures offer best performance at modest buffering

Practical node implementation feasible with current technologies High-speed local area and metropolitan area

networks High performance computing backbone

Possible extensions Multi-wavelength networks using additional

componentry, i.e. aggregate of > 1 Tbps Mesh topologies – control and routing issues