1 CHAPTER 8 Transport Networks: Advanced concepts M. Pickavet and C. Develder

1

CHAPTER 8 Transport Networks:

Advanced concepts

M. Pickavet and C. Develder

Transport networks:Advanced concepts 2

Outline

1. Network management and control1.1 Network management1.2 Network control1.3 Interaction

2. Network recovery3. Optical packet/burst switching


Monitor, account and control

the network activities and resources

TMN or Telecommunications Management Network

FCAPS : Fault managementConfiguration managementAccounting managementPerformance managementSecurity management

Network Management


Data plane

XC

XC

XC XC

XC

Mgmt plane

MIB

MIBMIB

MIB

MIB

Network Mgmt System

Network ElementMgmt Agent

TelecommunicationMgmt Network

OXC or DXC

Customer Premise Equipment

Circuit (e.g. lightpath)

Network Management


NMS

Via signalling channels(SDH overhead)

DXCLTE

LTE

LTE

LTE

matrix

hard

ware

AGENTPROCESS

MANAGINGPROCESS

managedobject

managedobject

managedobject

softw

are

MIB

MIB

Conceptual architecture


Network Management: Pros and contras

Well-standardised Mature Widely used in practice (legacy) Centralised:

Efficient Scalable ? Vulnerable

Reconfiguration: slow (e.g. months) Semi-permanent connections Dynamic IP traffic ?


Outline




Network Control

Data plane

XC

XC

XC XC

XC

Circuit

Control plane

User-NetworkInterface

Network-NetworkInterface

ConnectionController

Connection ControlInterface

Request Agent


ASTN and GMPLS

Automatically Switched Transport Network (ASTN) E.g. ASON (Aut. Sw. Optical Netw.) Based on distributed control plane

architecture Enables fast reconfiguration

Protocol: GMPLS (Generalised MPLS) Standardisation by IETF Idea: MPLS concepts transport network

layer (OTN, SDH, SONET, …)


5

7A

B

C

D

IP Payload

IP Header

MPLS Label

IN IF IN LABEL OUT IF OUT LABELA 2 D 3B 5 C 7B 9 D 7

GMPLS exampleIP/MPLS router

MPLS:

OXC

A

B

C

D

IN

OUT

IN -->

OUT

GMPLS-capable OXCGMPLS:


OXC

A

B

C

D

IN

OUT

IN -->

OUT

GMPLS in general

GMPLS-capable OXC

GMPLS also applicable to other scales/technologies OTN label = fiber, waveband, wavelength, … SDH label = time slot, …

ASTN/GMPLS = efficient BW utilisation ? Longer time scale variations (s, min, …): reconfiguration

possible Shorter time scale (ms, s, …):

still circuit switching


Outline




Management and Control

Control plane

Data plane

OXC

OXC

OXC OXC

OXC

Mgmt plane

MIB

MIBMIB

MIB

MIB


Outline

1. Network management and control2. Network recovery

2.1 Network failures2.2 General (single-layer) recovery

concepts2.3 SDH & OTN examples2.4 Multi-layer recovery

3. Optical packet/burst switching


1988 : fire in small switch in Hinsdale (Illinois) 35000 residential lines disconnected

37000 trunk lines disconnected118000 long-distance lines failed(500000 residential and business users affected, normally calling 3.5 million times a day, O’Hare airport closed)

1988 : two fuses of 600A blown (Massachusetts) 35000 users disconnected for whole day

banks closed for security reasons1990 : failure in signaling network (SS7) 65 million connections lost over US1991 : 3 wrong lines of software (on a total of 2.1 million lines) 1 week no telephone connections between Washington, Los Angeles

and Pittsburg

Examples of network failures


Examples of network failures1995 : Hanshin Earthquake in Kobe (Japan) 7.2 on Richter Scale, 5379 people died, 34626 people injured


Illustration of the effects of a failure on the telephone network

affected infrastructure:• 193.000 circuits• 3.500 leased lines (14%)• 3.600 poles• 330 km aerial cable• 20 km buried cable• 2.600 manholes• 210 km cable conduit

number of call attempts per call after the disaster

Examples of network failures1995 : Hanshin Earthquake in Kobe (Japan)


Typical failure rates

Web server 104-106 1IP interface card 104-105 2IP router itself 104-106 2ATM switch 105-106 1SDH DXC 105-106 4SDH ADM 105-106 4OTN OXC 105-106 4OTN OADM 105-106 41 km cable 106-107 48

Equipment type MTBF (hours) MTTR (hours)

(1 year 104 hours)

time

TBF

TTR TTR TTR

TBFTBF TBF

TTR = time to repairTBF = time between failures

Statistics:MTTR = mean TTRMTBF = mean TBF

Avail. MTBF-MTTR MTBF


Example: pan-European network

Oslo

Stockholm

Copenhagen

Amsterdam

Dublin

London

Brussels

Paris

Madrid

Zurich

Milan

Berlin

Athens

BudapestVienna

Prague

Warsaw

Munich

Rome

Hamburg

Barcelona

BordeauxLyon

Frankfurt

Glasgow

Belgrade

Strasbourg

28 nodes 20000 km cable

Cable break: every 4 days (!)Node failure: every month


Need for recovery ?

Plain Old Telephone service 5Voice over IP 5Video telephony 5Video-conferencing 5Tele-working 4TV broadcast 4Distance learning 5Movies on demand 3News on demand 2Internet access 2Tele-shopping 2

Application Need for recovery

(5 = crucial, 1 = not needed)

streaming

traffic

elastictraffic


Network Operator

Defect Duration

Compensation

Ameritech >1 min/ month 1 month credit

AT&T 1-60 min

>9 hours

5% month tariff pay back

50% month tariff pay back

BellSouth >2.5 sec 1 month credit

Nynex >1 min/ month 1 month tariff pay back

Pacifi c Bell >2 hours 1 month credit

Examples of availability guarantee


Outline






Improving network availability (e.g.

99.999 %)

working path

More reliable equipment (safer design, more testing, …)

Duplicate vulnerable network elements Network recovery mechanisms:

RHE

RTE

recovered segment

recovery path


Essential failure scenarios

single link failure

single node failure

(traffic terminated in affected node can not be recovered)


Recovery mechanism: goals

Stability Scope of failure coverage Recovery time Backup capacity requirements Guaranteed bandwidth ? Additional delay and jitter Packet reordering or duplication ? State overhead Signaling requirements Scalability Recovery classes ?


Steps in recovery process

time

fault detection timehold-off time

fault notification timerecovery operation time

traffic recovery time

failurefault detected

recovery time

operationaloperational


Backup capacity: dedicated

working paths

recovery paths

channel 1

channel 2

A B C

D E F

G H I


Backup capacity: shared

one commonchannel

A B C

D E F

G H I

working paths

recovery paths


Backup capacity: dedicated shared

Dedicated:• More simple• Backup capacity usage less efficient

Shared:• More complex: check whether resource available higher recovery time• Backup capacity used more efficiently


Recovery paths: preplanned dynamic

Preplanned: for all accounted failure scenarios, path of recovery flow is calculated in advance

• Allows fast recovery• No flexibility for unaccounted failure scenarios• Can be shared or dedicated backup capacity

Dynamic: path is computed on the fly once the failure is detected

• Additional time needed to identify suitable recovery path• Can search for recovery of unaccounted failure scenarios

too• Leads typically to shared backup capacity


Protection vs. restoration

Protection: all signaling occurs before failure• No time needed for after-failure signaling fast

Restoration: part of signaling occurs after failure

• Typically shared backup capacity requires less capacity


Protection variants

1+1 protection: dedicated protection• Traffic is permanently duplicated• Signal selection at RTE

1:1 protection: dedicated protection with extra traffic

• Traffic only on one path• Other path if available: other traffic

1:N protection: shared protection with extra traffic• One backup entity for N working entities

M:N protection• M backup entities for N working entities


Recovered segment: local recovery

working path

recovery path

RTERTE RHERHE

working path

recovery path

working path

recovery path

RTERTE RHERHE


Recovered segment: global recovery

working path

recovery path

RTERTE RHERHE

working path

recovery path

RTERTE RHERHE


working path

Recovered segment: local global

Choose possible recovery paths if• local recovery ?• global recovery ?

Exercise


Recovered segment: local global

RTERHE RTERHE

Local recovery: nodes close to failure faster recovery

Local recovery: inefficient total paths

consuming more capacity

Slightly different failure coverage

Other state and signaling requirementsRTERHE RTERHE


Control of recovery mechanism

Centralisede.g. TMN-based

Distributed e.g. control plane based (IP, GMPLS)

Pros and contras:• Centralised: good overall network view• Centralised: typically less complex• Centralised: typically more efficient capacity usage• Centralised: central point vulnerable point on its own• Distributed: more scalable


Network topology: mesh ring

ring 1ring 2ring 3

working path

recovery paths


Outline



concepts2.3 SDH & OTN examples

2.3.1 Ring protection2.3.2 Mesh protection2.3.3 Mesh restoration

2.4 Multi-layer recovery



(O)MS-SPRing

(O)ADM

Connection

Working Capacity

Protection/backup capacity

A B C D

H G F E

(Optical) Multiplex Section Shared Protection Ring


(O)MS-SPRing

A B C D

H G F E

(O)ADM

Connection

Connectionlooped back

Working Capacity


B C

B C+


(O)MS-DPRing

A B C D

H G F E

Situation without failure

Connection Working Capacity


ADM

(Optical) Multiplex Section Dedicated Protection Ring


(O)MS-DPRing

Connection

Connectionlooped back

Working Capacity


ADM

A B C D

H G F E

Situation in case of a link failure


SNCP Ring

BridgeA B C

DF E

SubNetwork Connection Protection Ring

Switch/Selector

Bridge

Switch/Selector


Ring interconnection

A

B

C

D

H

E

G

F

J

I

Single point of failure


Ring interconnection

A

B

C

D

H

E

G

F

J

I

Drop & continu

e

Drop & continu

e

Drop & continu

e

Drop & continu

e


Outline








1+1 Multiplex Section Protection

Cable W: working STM-N signalCable B: backup STM-N signal

Bridge

Selector Bridge

Selector


1+1 Mesh SNCP

A

B C

D

EF


Outline








Path restoration

Working path 2

Working path 1 Working path 1

Working path 2

Recoverypath 1

Recoverypath 1

Recoverypath 2

Recoverypath 2


Comparison of capacity requirements

0

5000

10000

15000

20000

25000

no protection 1+1 path

protection, link

disjoint

1+1 path

protection, node

disjoint

path restoration link restoration

# o

f re

quir

ed w

avel

engt

hs

backup path

working path


Outline






Why multilayer recovery: Example 1

client layer

server layer

a

b

c

d

A

B

C

D

E working path

recovery path


Why multilayer recovery: Example 2

client layer

server layer


Why multilayer recovery: general

Trade-offs:• Lower layers will not notice failures of higher layer equipment

recovery at higher layer needed • Higher layer equipment can get isolated by lower layer

equipment failure (see example 1) recovery at higher layer needed

• Escalation of root failure (see example 2) recovery at lower layer preferred

• Native traffic injected in lower layer recovery at lower layer needed

• Multilayer recovery is complex (design, monitoring, operation)

Key questions:• In which layer(s) recovery ?• Which recovery mechanisms ?• How to coordinate the recovery mechanisms in multiple layers

?


Multilayer recovery: uncoordinated

A

D

B

C

E

client layer

server layer

a

d

b

c

Client Layer primary path

Client Layer recovery path Server Layer recovery path


Multilayer recovery: bottom-up

A

D

B

C

E

client layer

server layer

ad

b

c

Server layer recovery failed

A

D

B

C

E

client layer

server layer

ad

b

c

Client Layer primary path

Client Layer recovery path

Server Layer recovery path

Phase 1: recovery action in server layer

Phase 2: recovery action in client layer

• Hold-off timer• Recovery token


Outline

1. Network management and control


3.1 Introduction3.2 Node architectures3.3 Contention resolution


Optical switching:• direct light from an

input port to an output port• possibly wavelength conversion

circuit-switching:• continuous bit-stream• pre-established light-paths• set-up: “manual” or dynamic

packet/burst switching• chunks of bits, encapsulated in packets• packet header determines forwarding• e.g. label switching, GMPLS based

Optical packet/burst switching

f f

cc b

a

d

b

e

c

f

OPS/OBS:packet/burst (at least payload) stays in optical domain


Packet format

fixed/variable duration: pro variable = no

fragmentation/reassembly, no padding, less header overhead

contra = long packets can block many short ones

slotted/unslotted operation: pro slotted = easier packet

scheduling (synchronous switching)

contra = cost of synchronisation components

1

2

unslotted, variable length

1

2

slotted, fixed length

1

2

slotted, variable lengthpadding

single packet

OPS:

OBS:


Header format

1

2

3

4

1

2

3

4

1

2

3

4

phase

intensity

out-of-band: orthogonal channel (e.g. DPSK)

out-of-band: dedicated wavelength;

position of header: in-band: header and payload are sent sequentially,

separated in time


Typical operation of OPS fixed-length packets, slotted operation header accompanies payload

• contains necessary information to make forwarding decision

each timeslot:• inspect packets at input ports• decide which packets can be forwarded without

collisions

switch is “memory-less”• no knowledge of packets scheduled in past is necessary

OPSnode


Typical operation of OBS variable packet lengths, unslotted operation header is sent Toffset before payload

• contains necessary information to make forwarding decision• functions as one-way reservation (allows timely configuration

of switch fabric)• offset decreases by header processing time per hop• priority mechanism possible

on arrival of header:• decide whether burst can be forwarded without collisions• make necessary resource reservations if burst is accepted

switch needs “memory”:• keep track of reservations made in past

OBSnode

Toffset -

Toffset


Outline





Functionality of an OPS/OBS node input interface:

• header extraction (straightforward if out-of-band)• synchronisation: detect beginning of packet/burst• in OPS: align packets

switching matrix: fast reconfig. (s or ns) crucial• MEMS too slow• SOAs or fast TWCs possible

output interface• e.g. regeneration of optical signal; header re-writing…

synchr.control

switchcontrol

headerrewriting

inputinterface

switching matrix

outputinterface

payload

header


Header processing

Electronical header processing• Optics for capacity & switching (payload)• Electronics for routing & forwarding (header)

Optical header processing• Avoids O-E-O conversions for headers• Limited optical processing functionalities

synchr.control

switchcontrol

headerrewriting

inputinterface

switching matrix

outputinterface

optical processing

Electronic or opticalprocessing


Outline





Problem and possible solutions

Problem:two or more packets contend for same resource: destined

for same outgoing port at the same time

Solutions:

contention

deflection routing wavelength conversion optical buffer (Fiber Delay Lines)


Problem and possible solutions

contention

deflection routing wavelength conversion optical buffer (Fiber Delay Lines)

Try to predict the performance:• network throughput• packet loss rate as total network load increases

(cf. traffic jams)

Exercise


What solution to choose?

Deflection:• packets are “stored” in network:• increases load, increases delay• only works for low loads

Wavelength conversion:• no packet storage• allows high network throughput, no

increased delay

Buffering:• local packet storage at nodes• small delay penalty

Conclusion: Use combination of wavelength conversion and buffers figures © Yao et al., Opticomm’00

netw

ork

th

rou

gh

pu

tpack

et

loss r

ate

load (packet arrival rate, pkt/s)

load (packet arrival rate, pkt/s)


Buffer architectures

feed-forward vs feed-back• feed-forward: input or

output buffering• feed-back: shared,

recirculating FDLs

single stage vs multiple stage

• multiple stages separated by switching elements

• e.g.: each stage different delay resolution (“units”, “tens”, “hundreds”…)

choice of FDL lengths in stage

D-1

10

...

feed-forward vs feed-back

single vs multiple stages

D-1

10

D-1

10

...

...


OPS/OBS: fixed vs increasing FDLs

1.E-07

1.E-05

1.E-03

1.E-01

0 8 16 24 32 40 48 56 64nr. buffer ports (B)

ParetoOnOff, incr ParetoOnOff, fix

GeoOnoff, incr GeoOnoff, fix

Poisson, incr Poisson, fix

sample results for fixed-length, slotted OPS

Increasing FDL lengths give far lower PLRs (order of magnitude or more)

“penalty”: reordering of packets, higher delays

1

B

…1

1

1

B

…

…

1

B

Pack

et

loss

rate


“Network Recovery: Protection and Restoration of Optical, SONET-SDH, IP, and MPLS” by JP Vasseur, M. Pickavet, P. Demeester (July 2004, Morgan Kaufmann, ISBN 0-12-715051)

“Node Architectures for Optical Packet and Burst Switching” by C. Develder, J. Cheyns, E. Van Breusegem, E. Baert, A. Ackaert, M. Pickavet, P. Demeester, Invited - Technical Digest of PS 2002, the 2002 International Topical Meeting on Photonics in Switching, ISBN 89-5519-085-9, 21-25 July 2002, Cheju Island, Korea , pp. 104-106)

References

Documents

1 CHAPTER 8 Transport Networks: Advanced concepts M. Pickavet and C. Develder