• Aucun résultat trouvé

OpenFlow protocol extension for optical networks

N/A
N/A
Protected

Academic year: 2021

Partager "OpenFlow protocol extension for optical networks"

Copied!
108
0
0

Texte intégral

(1)

OPENFLOW PROTOCOL EXTENS

I

ON FOR OPTICAL

NETWOR I

<:S

THE

IS

PRESENTED

AS A PARTIAL REQUIREJ\IENT

FOR T

H

E 1\

I

ASTER

l

i'J'

ELECTRICAL ENG

INEERING

BY

1\

IAH

l

OUD

1\IOHA lED

BAHNASY

(2)

Service des bibliothèques

Avertissement

L

a diffusion de ce mémoire se fait dans le respect des droits de son auteur, qui a signé

le

formulaire Autorisation de reproduire et de diffuser un travail de recherche de cycles

supérieurs

(SDU-522- Rév.01-2006)

.

Cette autorisation stipule que

«conformément

à

l'

article 11 du Règlement no 8 des études de cycles supérieurs

, [l

'

auteur] concède

à

l

'

Université du Québec

à Montréal une licence non exclusive d

'

utilisation et de

publication

de la totalité ou d'une partie importante de [son] travail de recherche pour

des

fins pédagogiques et non commerciales.

Plus précisément, [l

'

auteur] autorise

l'

Université du Québec à

Montréal

à

reproduire, diffuser, prêter, distribuer ou vendre des

c

opies de [son] travail de recherche

à

des fins non commerciales sur quelque support

que ce soit, y compris l

'

Internet. Cette licence et cette autorisation n'entraînent pas une

renonciation de [la] part [de

l

'

auteur]

à

[ses] droits moraux ni

à

[ses] droits de propriété

intellectuelle.

Sauf entente contraire, [l'auteur] conserve la liberté de diffuser et de

commercialiser ou non ce travail

dont [il] possède un exemplaire

(3)

EXTE SIO

DU PROTO

COLE

OPENFLOW POUR LES RÉSEAUX OPTIQUES

t

!

ÉMOIRE

PR

ÉSE TÉ

COtl

lME

EX

IG

E:-JCE

PARTIELLE

DE LA

MAÎTRISE

EN

GÉNIE

ÉLECTR

I

QUE

PAR

MAH

·

IO

UD

MOHAMED BAHNASY

(4)

It

i

s

my p

l

eas

ur

e

to

thank a

U

those

peop

l

e

who m

a

d

e

thi

s

t

h

s

i

s possible.

At

first

,

I wo

ul

d

li

k

e

to t

h

ank

my

advisor

Prof

esso

r Ha

li

ma E

l

b

i

aze.

Wit

h

h

e

r

e

n

co

ur

-age

m

e

n

t,

in pir

at

i

o

n

a

nd h

e

r

great e

ffort

of ex

p

l

a

i

n

i

n

g

things

cl

ea

rl

y and s

imp

l

y

;

s

b

e made

t

h

e

r

esea

r

c

h p

r

ocess as

s

i

mp

l

e as

po

ss

i

b

l

e throug

h

o

u

t

m

y

f

aster

p

e

r

i

od.

S

h

e provided

u

sef

u

l

a

dv

i

ce,

good

g

u

i

ding

,

p

e

rf

ect co

mpan

y

,

a

nd

s

h

e

a

l

wa

ys gave

m

e

great freedom to pursue

i

nd

e

p

e

nd

e

nt worlc

I

wou

l

d l

i

k

e to thank

m

y co

ll

eag

u

es

fo

r

provid

i

ng

an activat

i

n

g a

nd

exc

i

t

in

g e

nvironm

e

nt

i

n

w

hi

c

h

I

cou

l

d

l

ea

rn

and

e

n

h

a

nce m

y co

mp

e

te

n

ce

.

I

wish to thank

m

y e

ntir

e exte

nd

e

d f

a

m

il

y

f

or a

ll

the

i

r

l

ove

a

nd

e

n

co

urag

e

m

e

nt p

ec

iall

y

m

y

b

e

l

oved

moth

e

r

.

And mo

st

i

mport

a

nt

l

y

,

I

w

i

s

h

to

thank m

y

l

ove

l

y w

i

f

e,

L

a

il

a, as t

h

ey

a

l

ways

s

upport m

e

i

n

th

i

s

road

and they

w

e

r

e e

n

co

u

r

ag

i

ng me

a

l

ong

t

hi

s

w

ay.

Fina

ll

y

,

I

wou

l

d lik

e

to

thank a

ll

th

e staff

m

e

mb

e

r

s of the Comp

u

ter

S

c

i

e

n

ce

d

e

p

a

rtm

e

n

t

a

t

UQA

1 for t

h

e

ir

d

i

r

ec

t

and

in

dir

ect

h

e

l

p dur

i

ng my

s

tud

i

es

at UQAM.

(5)

Software

D fin

e

d

Networking

(

SDN

)

offre

l

a

po

ss

ibilit

é

de

co

n

t

r

ô

l

e

r l

es

r

ésea

ux

e

n

util-i

ant

un

log

i

c

i

e

l fon

ct

ionnant

s

ur un

ystè

m

e

d

'

expl

oitat

i

on

d

a

n

s

un

co

ntr

ôle

ur

exte

rn

e, ce

qui offre un

maximum de

fl

ex

ibi

l

it

é et

d

e si

mpli

c

i

té.

Op

e

nFlow

(

OF

),

un

e

d

es

impl

éme

n

ta

-t

i

o

n

SDI

l

es

plu

uti

l

i

ée

,

e

t présenté

co

mm

e a

r

c

hit

ect

ur

e

et

p

la

n d

e co

ntr

ô

l

e

unifi

é

pour

les

r

és

e

a

ux d

e

p

aq

u

ets

et

d

e c

ir

c

u

i

ts

.

D

a

n

s ce

pr

o

j

et

,

nou

proposons

exp·

rim n

t al

m

ent

d

e

ux

o

lu

t

i

ons

b

as

·

es ,

ur Op

e

nF

l

ow

pour

co

ntr

ôle

r

à

l

a

fois

les

r

ésea

ux d

e

paqu

ets et

l

es

r

ésea

ux

optiques

:

(1

)

Op

e

nF

l

ow

M

essage-

Mappin

g

t

(2)

Op

e

nF

l

ow

Ext

n

si

on.

Op nFlow

M

ssage

-

N

l

a

ppin

g e

t

b

asée

ur

l

'association d m

esage

Op

e

nF

l

ow

à

d

es acti

ons

ap

propri

ées

.

L

a

d

e

uxi

ème sol

uti

o

n

que n

o

u pr

oposo

n

,

Op

e

nFl

ow

Ext

n ion

,

est

ba

ée s

ur

l

'

exte

n

s

i

on

du

protocole OpenFlow

sta

nd

a

rd

e

n

a

jou

ta

nt

d

n

o

uv

eaux

m

e sage

pour

up

o

rt

e

r d

e

inf

o

r-mations

d

'int

e

r

co

nn

xion

o

ptiqu

e

au

li

e

u

cl

'uti

l

i

er

d

es

m

essages

Op

e

nFlow

sta

nd

a

rd.

No

u

s

avo

n

s

impl

é

m

eté ces

deux olutions dans

un b

a

n cl

'

a

i

et

nou

a

von

s e

ff

ect

u

é

deux

ex

p

é

ri-e

n

ces

:

(

i

)

la c

r

éat

i

o

n

de

ca

n

a

ux

o

ptiqu

e

d

e

bout

e

n

bout.

(

ii

)

la r

esta

ur

at

i

o

n d

e c

hemin

opt

iqu

e

. L

s

mesures prise

à

partir

de

ce ex

p

é

ri

e

n

ce

ont

uti

l

i

ées

p

o

ur

l

'impl

é

mnt

at

i

o

n

d

'un

s

imu

late

ur J

ava.

Ce

s

imul

ate

ur

si

mul

e les

p

e

rf

o

rm

a

n

ces

d

e ces

d

e

ux

techn

i

que

ur d

eux

topo

l

og

ies

de

r

ésea

u

x

optiques

r

éel

s

et

l

es

co

mp

a

r

e

av

c

le

proto

cole

G

fPLS

sta

ndard

.

L

e

r

's

u

l

tat

e

t

r

e

pr

ése

nt

é so

us

fo

r

mat d

e

g

r

ap

hi

q

u

s co

m

r

aratifs

po

ur

d

é

t

e

rm

i

n

er l

a

t

ec

hn

i

que

qu

i a le

m

ille

ur

temps d

'éta

b

l

i

sse

m

e

nt de

lie

n

s,

la

p

l

u

s

p

eti

t

e

charge d

e co

ntr

ôle e

t

1

r

a

pport

de blocage

l

e

plus ba

. L

a

fai

a

bilit

é

d

e ce

o

luti

o

n

a été

vérifée

d

a

n

s

n

ot

r

e

banc d

'essa

i

et

l

eu

r

s

performances

sont

quantitativement

'

valuée

et compa

r

ées

dans deux réseaux

opt

i

ques

r'

e

l

.

(6)

So

f

twa

r

e

D

e

fi

ned Netwo

rk

(S

D

N) a

ff

o

rd

s

t

h

e

poss

ibili

ty to

co

n

t

r

o

l n

etwo

rk

s

u

s

in

g so

f

twa

r

e

r

unni

ng o

n

a

n

etwo

r

k ope

r

at

i

ng

syste

m

in

a

n

exte

rn

a

t

co

n

t

r

o

ll

e

r

, w

hi

c

h p

rov

i

des max

i

mum

fi

ex

ibili

ty,

s

impli

c

i

ty an

d m

a

n

agea

bili

ty.

Op

e

nFl

ow (

OF

), o

n

e

o

f

t

h

e

w

id

e

l

y

u

sed

S

D

im-p

l

eme

n

tat

i

o

n

s,

i

s

pr

es

nted

as a

unifi

ed co

n

t

r

o

l pl

a

n

e

a

n

d a

r

c

hi

tect

ur

e

f

o

r

pac

k

et

and

c

i

rc

ui

t

sw

i

tc

h

ed

n

etwo

rk

s

. B

ase

d

o

n thi

s,

in

t

hi

s

t

h

es

i

s,

w

e ex

p

e

rim

e

n

ta

ll

y

pr

o

p

ose two

so

lu

t

i

o

n

s

based o

n O

pen

Fl

o

w

to

co

n

t

r

o

l

bot

h

pac

k

et

a

nd

o

p

t

i

ca

l

n

etwo

rk

s

:

(

1

)

Op

e

n

F

l

ow Message

-Map

pin

g an

d

(2)

Op

e

nFl

o

w

exte

n

s

i

o

n

.

O

pen

Fl

ow Message

-M

ap

pin

g

i

s bas

d

o

n

mapp

in

g

t

h

e

O

pe

n

F

l

ow

m

essages

in

to a

ppr

op

ri

ate c

r

o

-

co

nn

ect

act

i

o

n

s

. Th

e seco

nd

so

lu

t

i

o

n

we p

r

opose

,

O

pe

n

F

l

ow Extens

i

o

n

,

i

s

b

ase

d

o

n

extend

in

g t

h

e

ta

nd

a

rd O

pe

n

F

l

ow p

r

otoco

l

by

a

ddin

g

n

ew

m

essages to ca

rr

y

t

h

e c

r

oss-co

nn

ect

inf

o

rm

at

i

o

n in

stea

d

o

f

u

s

in

g sta

nd

a

rd O

pe

n

F

l

ow mes

-ages. We

im

p

l

eme

n

ted t

h

ese

two so

lu

t

i

o

n

s o

n

a tes

tb

ed an

d

co

n

d

u

ct

two

expe

ri

me

n

ts:

(

i

)

E

nd-

to-E

nd

li

g

h

t

p

at

h

e

t

ab

li

s

hm

e

n

t.

(

ii

)

B

ac

kup

li

g

htp

a

th r

esto

r

at

i

o

n. Th

e

m

eas

ur

e

m

ents

ta

k

en

f

ro

m

t

h

ese

exper

i

me

n

ts a

r

e

u

se

d in

w

ri

t

in

g a

c

u

stom

-buil

t

J

ava eve

n

t

-dri

ven

s

i

m

ul

ato

r

.

Thi

s s

i

m

ul

ato

r

s

imul

ates t

h

e

pe

rf

or

m

ance of t

h

ese two tec

hni

q

u

es

o

n

two

r

ea

l

opt

i

ca

l n

etwor

k

to

p

o

l

og

i

es a

nd

co

mp

a

r

e

t

h

e

m

w

i

t

h

t

h

e

sta

nd

a

rd

GM

PLS

p

r

o

t

oco

l.

Th

e

r

es

ul

t

i

s

dep

i

cted

w

i

t

h

comparat

i

ve g

r

a

ph

s to

m

ake

i

t

easy

to

d

eter

min

e

w

hi

c

h

tec

hni

q

u

e

h

as t

h

e

fastest

estab

-li hm

e

n

t t

im

e,

l

owest co

n

t

r

o

l l

oa

d and

l

owest

bl

oc

kin

g

r

at

i

o

. Th

e ove

r

a

ll

f

eas

i

b

ili

ty o

f

t

h

ese

so

lu

t

i

ons

i

s

assessed

u

s

in

g o

ur

test

b

ed and t

h

e

ir p

e

rf

orma

n

ces

a

r

e

q

u

a

n

t

i

tat

i

ve

l

y

eva

lu

ated

(7)

RÉSUMÉ

ABSTRACT

LIST OF FIGURES

.

LIST OF TABLES .

ABREVIATIO S

I TRODUCTIO

STRUCTURE OF THIS DISSERTATION

CHAPTER I

PRINCIPAL CONCEPTS.

1.1 Circui

t-switched

and

P

acket-Swi

tched Net

work

1.1.1

I

nt

ernet Archit

ectur

e

1.1.2

Tran

sport Network Archit

ectur

e

1.2

Existing

Approach

es

1.2.1

IP

ov

r WD 1

1.2.2

G

1PLS

as a Unified Control iechanism

1.3 SOFT

\NARE

DEFlNED

ET\i ORKI

G

(

D

T

)

1.3.1

OPE

!

FLOW .

CHAPTER II

PROPOSED SOLUTIO S

2.1 OpenFlow

ch

ann

e

l

. . .

2.2 OpenFlow Optical Agent

2.2.1 Ports-Emulation

Modu

l

e

.

2

.

2.2 OpenFlow

/

TL1 Tran

lator

2.3 Path

Computat

i

on E

l

ment

(PCE)

2.3.1 Executor

.

.

.

2.3

.

2

0

S Ad

apter .

2.4

OpenFlow

Message-Mapping

So

lu

t

i

on

iv

v

viii

x Xl 1 7

8

10

12

12

14

15

17

.

.

.

.

. .

22

24

27

27

27

3

1

31

32

32

(8)

2.5 Op

e

nFl

ow

Ext

e

n

s

i

o

n

So

lu

tion

.

.

2

.

6 GMPLS

\

1\f

iTH P

CE

LIGHTPATH

SETUP

CHAPTERIII

CONDUCTED EXPERIMENTS

3

.

1 T

est

b

e

cl

S

e

tup

3.2 Scenar

i

o

1:

End-to-End

Li

g

h

tpath

S

et

up

a

nd R

e

l

ease .

3.3 S

ce

n

a

ri

o

2: B

ac

kup

li

g

h

t

p

at

h R

es

torat

i

on

3.4

GMPLS Approa

c

h Exp

e

r

i

ment

3.5 Exp

e

rim

e

n

tat

i

o

n R

es

ul

ts

.

CHAPTERIV

SIMULATION STUDY

4.

1 Th

e

C

u

tom

-built J

a

va Ev

e

nt-Dri

ve

n

Simulator

.

4

.2

Na

tion

a

l S

c

i

e

n

ce

F

o

und

atio

n

(N

SF)

topo

l

ogy ..

4

.3

E

ur

opea

n

Opti

cal

et

work Topo

l

ogy (COST239)

4.4

Summar

y

of

Simu

lat

i

on

R

es

ul

ts .

CHAPTER V

34

35

38

3

8

41

45

47

50

52

53

59

63

6

8

CONCLUSION

.

.

. .

.

.

.

.

.

.

.

.

.

.

. .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

70

CHAPTER VI

PUBLICATIONS

. . . .

.

..

.

.

.

. . . .

.

. .

6.1 A

cce

pt

ed

p

a

p

e

r

a

t

I

EEE GLOBECOM 2014

co

nf

e

r

e

n

ce

6

.

2

Submitt

e

d

pape

r

at

Opti

cs

lnf

o

B

ase j

o

urnal

(20

14

)

.

.

.

71

71

79

(9)

Fi

g

ur

e

IP

a

nd

Tran

s

port

Network

s

Op

e

r

at

in

g

L

aye

r

1

2

3

Op

e

nFl

ow

M

e. sages

/

TL1

Command

Tran

s

later

A

ge

nt .

T

es

tb

e

d

r

c

hit

ect

ur

e

.

1.1

IP

a

nd Tr

a

n

spo

rt

Networks

1.2 IP

Network

Ov

e

rl

ay

Tran

s

por

t

N

etwo

rk

s

1.3 Th

e

Ov

e

rl

ay

letworks

s

upp

o

rt

e

d b

y

Tran p

o

rt

Network

1.4

Tran

s

port

letwork

Control

&

Mana

ge

m

e

n

t

. . . .

.

1.5 IP

over WD

f

sce

nar

io

for th

e

Futur

e

of

th

e

letworks

1.6

Th

e

Tr

a

cli

tional Network

1

ode

hi

e

rarch

y

..

1.7

The Software

-D

e

fin

e

cl

Network

in

g et

work

ode

hi

e

ra

r

c

h

y

.

1.

Softw

a

r

e-

D

e

fin

e

d

e

tworkin

g etwo

rk

Nod

hi

era

r

c

h

y

1.9

Op

e

nFl

ow

n

et

w

o

rk

. .

. .

. .

1.10

Th

e se

paration b

et

w

e

n d

ata

p

l

a

n

e

and

co

ntrol

p

l

a

n

e

u

s

in

g

Op

e

nFl

ow

1

.

11 Packet flow t

hrough

an

Op

e

n

F

low witch

1.12 Pa

c

k

et

fl

ow

through

an

Op

e

n F

low

wit

c

h

2

.

1

Unifi

ed

architecture of

a co

nv

e

r

ge

cl

P

ac

k

et-C

ir

c

uit n

etwo

rk

.

2.2

Op

e

nFlow A

ge

nt

.

2.3

Op

e

nFl

o

w

Ch

a

nn

e

l

2.4

Op

e

nFl

w M

e sage

Fa

c

tory

2.5 Port

-Emu

l

at

i

o

n

Modul

e

.

2.6

P

at

h

Computati

o

n

E

l

e

m

e

nt

workflow

.

P

age

1

5

6

9

10

11

12

13

15

16

16

1

18

19

20

22

23

24

25

29

31

(10)

3.

1

Test

bed A

r

ch

i

t

ect

ur

e

.

.

..

3.2

Opt

i

ca

l

domain

I

nterconn

ect

i

on

3.3

phys

i

ca

l

equipments

i

n

the

Opt

i

ca

l

Transport

1

etwork Laboratory

3.4

W

i

reshark Screen hot

(Lightpath

Set

up

Message

Exch

ange)

3.5 Netwo

rk

configurat

i

on and

message exchange

.

..

.

3.6

C

i

sco

Transport

Contro

ll

er Screenshot

(

I

n

i

ti

a

l

State)

3.7 Cisco

Tran

sp

ort

Controller

Screenshot

(After L

i

ght

path

Estab

li

shment)

3.8

ML diagram

for

li

g

h

tp

ath

e

tab

li

shment

.

.

.

.

. .

.

.

.

.

3.9 Exch

anged messages

Backup

li

ghtpath Resto

rat

i

on

Scen

ario

3.

1

0 W

i

res

h

a

r

k Screens

h

ot (L

i

ghtpath

Set

u

p Me. sage

Exc

h

ange)

3.11

C

i

sco

Transport

Cont

r

o

ll

er Screen

shot

(After Lightpath Restorat

i

on

)

3.12 U ifL

d

i

ag

r

am of

l

ightp

ath recovery

.

313

GMPLS

Exper

i

ment

U

i

ng

DRAGO

3

.

14

GMPLS Sce

n

ar

i

o : W

i

res

h

ark

scree

n

s

h

ot

4

.

1

SF

topo

l

ogy (

1

4

nodes and 21 lin

k

s)

4.2

Lightpath

establ

i

shment

time

[m

]

vs.

network

l

oad

(

SF topology)

4.3 N

u

mber Of Hop Per Request

vs

.

network

l

oad

(

J

SF

topo

l

ogy)

4.4

umber

of

contro

l

messages

vs.

networ

k l

oad (

SF

topo

l

ogy)

.

4.5

Lightp

ath b

l

ock

i

ng

probab

i

lity

vs.

net

work

l

oad

(NSF

topology)

.

4.6 COST239

Topo

l

ogy (11

nodes a

nd

26

li

nks)

. . . .

.

..

..

..

.

4.7

Lightpath

establ

i

shment

time

[ms

]

vs.

n

etwork

l

oad

(COST239 Topology)

4.

8 N

u

mber Of Hop Per R

equest

vs

.

netwo

r

k

l

oad (COST239

Topo

l

ogy)

4.9

Number

of

contro

l

messages

vs

.

net

wo

r

k

l

oad (COST239

Topo

l

ogy)

.

4.

10 Lightpath

blocking

probabi

li

ty

vs.

netwo

rk

l

oad (C

OST239

Topology)

39

40

40

41

42

44

44

44

45

46

47

47

4

49

60

60

61

62

63

64

65

66

66

67

(11)

Tabl

e

2.1

l

lessage

Factory Example

.

.

. . .

.

.

..

.

..

.

2.2

Us

in

g

Message

Factory To

C

reat

a

Feat

ur Reply Message

2.3 Exec

u

t

in

g

TLl

Create

Li

g

h

tpat

h

Comma

nd

o

n

th

e Op

t

ical Switch

3.

1

Th

experim

e

n

ts

timing

4.

1

Summary

of

Simul

ated

Solutions

4.2

S

umm

ary of NSF topo

l

ogy s

i

m

ul

at

i

o

n

resul

ts

4

.3 Summary

of C

OST239

topo

l

ogy

simul

at

i

o

n r

es

ult

s

P

age

26

2

30

5

1

52

6

69

(12)

AS

CLI

Cost239

CSA

CTC

DRAGON

DWDM

EMS

/

NMS

FEDERICA

Autonomo

u

s

System

Command Lin

e

In

terface

European

union Ultr

a

-Hi

g

h

Capac

i

ty

Opti

ca

l

Transm

i

ss

i

on

etwo

rk

C

li

e

nt Sy

ste

m Agent

s

C

i

sco

Transport Contro

ll

e

r

D

y

nami

c

R

eso

ur

ce

A

ll

ocat

i

on via GMPLSnOpt

i

ca

l

etwor

k

s

D

e

n

se

W

ave

l

e

n

gth

Di

v

i

s

i

on M

ul

t

ipl

ex

ing

Element and Network Management Systems

F

e

d

e

rat

e

d E-

infr

astruct

ur

e

D

ed

i

cated

to European R

esea

r

c

h

ers

Inn

ovat-in

g

in

Computing network A

r

c

hit

ect

ur

es

flow

GENI

GMPLS

ITU

LSA

LSP

MPLS

Node

OADM

NSF

OF

n

etwor

k

traffic

The Glob

a

l

Env

ir

o

nm

e

nt for

etwo

rk Inn

ovations

Gene

r

a

li

zed

Mu

l

t

i-P

rotoco

l

L

abe

l

Sw

i

tch

i

ng

In

ternationa

l

T

e

l

eco

mmuni

cat

i

on Un

i

on

Link

State Adve

r

t

i

se

m

ent

L

abe

l Sw

i

tched

Path

M

ul

ti

-Pr

otoco

l L

a

b

e

l

Sw

i

tc

hin

g

networ

k

nod

e

(Sw

i

tch

, r

o

u

te

r

... )

Op

tical

Add-Drop Multip

l

exer

Nat

i

ona

l

Sc

i

e

nc

e

Foundation

(13)

OFP

ONF

ONS

OSPF

PCE

PSTN

QoS

ROADM

RSVP-TE

SDN

SLA

Switch

TCO

TDM

TED

UCP

VLSR

VPN

WDM

wso

Op

e

nFl

ow

Pr

otoco

l

Op

en

Networ

king

Foundati

o

n

Opti

ca

l

Network

Swit

ches

Op n

Short

est

P

at

h

Fir

s

t

P

ath Comp

utati

on E

le

m

e

nt

Publi

c

Swit

ch

d

Te

l

eph

one Netwo

rk

Qu

a

li

t

y-

of-Se

rvi

ce

Reco

nfi

g

urabl

e

Opti

cal

Add-Drop

Mu

l

t

iplexer

l

esource

R

ese

r

vat

i

o

n Pr

otoco

l-

Traffi

c

Engin

eerin

g

Software

De

fin

e

d

Networ

k

S

ervice

Le

v

l

Agr

ee

m

ent

ne

twork

node

(Switch

, r

o

uter

,

e

t

c .

.

)

T

otal

Co

t

o

f

Own

ership

Tim

e

Dev

i

sion iul

t

iplex

in

g

Tr

a

ffi

c

En

ginee

rin

g

D

ataba

e

U

nifie

d

Co

n

tro

l Pla

n

e

Virtu

a

l

L

a

be

l

Switch

R

o

u

t

ers

Virtu

al Pri

vate Network

Wa

ve

l

en

gth

-d

i

v

isio

n Mu

ltipl

exin

g

(14)

O

ver

v

iew

Th

e expo

n

e

nti

a

l

grow

th

of

I

nternet traffic

r

eq

uir

es

network

provid

e

r

s

to

co

n

str

u

c

t

e

ffi

c

i

e

nt

n

etwo

rkin

g systems.

Th

ese

l

a

r

ge

n

etwo

rk

s

n

ee

d

a com

p

l

ex a

nd

sop

hi

st

i

ca

t

e

d

cont

rol

system

es

p

ec

iall

y when

it includ

es

two

diff

e

r

e

nt infra

st

ru

ct

ur

es.

On

e so

luti

on

to

manag

e

th

i

s

probl

e

m

i

s

to

r

ed

u

ce

the differences

in n

etwo

rk

st

ru

ct

ur

e,

f

o

r

exa

mpl

e,

mo

st

network

pr

ov

i

de

r

s

h

ave

r

emoved

t

e

l

e

ph

o

n

y co

r

e sw

it

c

h

es

and

r

e

p

l

aced

them by

u

s

in

g

voice over

IP

se

r

v

i

ces.

Today

'

s

n

etwo

rk

s a

r

e com

p

osed o

f

an

o

p

t

i

cal

d

oma

in

(c

ir

c

ui

t

-

sw

i

tched

networks)

and

an

e

l

ect

ri

ca

l

dom

a

in

(packet-switched

n

e

t

wo

rk

s)

. Th

ese

two

n

e

twork

st

ru

ct

ur

es

operate on

different network

l

aye

r

s: c

ir

c

uit

-sw

i

tched networks

ope

r

ates o

n l

aye

r

o

n

e an

d

two,

w

hil

e

p

ac

k

et

-

sw

i

tched

n

etwo

rk

operates on

l

aye

r thr

ee a

nd f

o

ur (Figur

e

1). Ho

weve

r

, e

l

ect

rical

OSlModel Application

Layer 7 Presentation Layer 6

Session Layer 5

,-

- - - -

-1 Packet-switched Network )

1 1 1 • • 1

:

.

:

./

...___

-

.

·-

.

- - -

.

~

.

- - -

-

.

.

-.

.

.

(-

:

--

--

r-

---

;

-\

1 1 1 ~ ptical Clrcuit-switched Network ..____

___

_

_______

/ Transport Layer 3 Data-Link Physical Layer 1

Figure

1:

IP

a

nd Tr

a

n

s

por

t

etwo

rk

s

Op

e

r

at

in

g

La

ye

r

s

domain i

s

mor

e

fl

exi

bl

e

and

eas

i

e

r

to

m

a

n

age, se

r

v

i

ce

pro

v

id

e

r

s ca

n not r

e

pl

ace optical

d

ev

i

ees

w

ith

e

l

ect

ri

ca

l

deviees because optical network

ha

s

many

b

e

n

e

fit

s ove

r

e

l

ect

ri

ca

l n

etwor

k.

S

orne o

f

these

advantages a

r

e:

(15)

Opti

ca

l

n

etwo

rk

e

quipm

e

nt

s

upport up

to

10

tim

es

the

ca

pa

c

ity as

e

l

ect

ri

ca

l

eq

uipm

e

nt

s

upport.

Opti

cal

transmission

s

upport

s ve

r

y

long di

stances w

ithout

s

i

g

nal

attenuation.

Opti

cal

transmi

ss

ion i

s

int

e

rf

e

r

ence

fr

ee

which makes

it mor

e

robu

st

.

Mo

st

of infr

as

tructur

e

pro

v

ider

s

use two

diff

e

r

e

nt

control

plan

e,

one

for

e

l

ect

ri

ca

l

domain

a

nd

anot

h

e

r f

o

r

op

ti

ca

l

domain.

The

se

p

a

rat

ion

b

e

tw

ee

n

these two

n

etwo

rk

s

i

s

b

eca

us

e

of

the

diff

e

r

e

nt man

age

m

e

nt

methodology

o

f

estab

li

shing a

data path.

Op

t

i

ca

l

domain

w

hi

c

h i

s c

ir

c

ui

t-sw

i

tched

network

ope

rat

es

on l

ayer

1

a

nd 2

of

OSI

model, whil

e e

l

ect

ri

c

al domain

i

s

pack

et-sw

itched

n

etwo

rk

and

it

o

p

e

r

ates o

n

layer

3

o

f OSI

model.

Another

diff

e

r

ence

b

etwee

n

the

two

domain

s

i

s

pa

c

k

et

-

sw

itched

n

etwo

rk

meant

to

b

e

distributed

co

ntr

o

l

, eac

h

rou

te

r h

as

it

s ow

n lo

cal

l

y c

r

e

-ated

control

st

r

ategy,

whil

e

on the other hand

c

ir

c

uit-

sw

it

c

h

e

d

network

i

s

mostly

cent

r

a

lized

contro

l.

Thi

s se

p

a

r

at

i

o

n pr

ese

nt

s

a

l

ac

k

o

f

commo

n

co

ntrol m

ec

hani

s

m

whi

c

h

s

upp

o

rt

s

both

network domains.

Mo

st

Infrastruct

ur

e

provid

e

r

s

u

se

two

co

ntrol

mechani

s

m

s

to

operate

both n

e

twork

s

which

i

s

more

expens

i

ve and

in

effic

i

ent t

h

an operat

in

g one converged

network

w

i

t

h

a

unified

cont

r

o

l

m

ec

h

a

ni

sm

. Sorn

e effo

rt

s

h

ave

b

ee

n

done to

unif

y

the

co

ntr

o

l

and

m

a

nag

eme

nt

of

h

ete

r

oge-n

eo

u

s

n

etwo

rk

s

.

Th

e

most mature and

w

id

e

l

y com

mon

exam

pl

e o

f

these

effo

rt

s

i

s

G

e

n

era

li

zed

Mul

t

i-Prot

oco

l L

abe

l

Swi

tc

hin

g

(GMPLS)(Mannie,

200

4)

pr

otoco

l

w

hi

c

h

is

ve

r

y com

pli

cated

a

nd not

eve

n

co

mm

e

r

c

i

a

ll

y

adapted.

Ev

e

n

though

it

was

u

se

d

,

GMPLS did not

co

mpl

e

t

e

l

y

unify

t

h

e cont

rol

mechanism.

Indeed,

it

preserves the

separat

ion between the

two

n

etwor

k

s

.

Motivation

S

e

r

v

i

ce

pr

ov

ider

s a

r

e

obliged to own

a

nd

operate two di

st

inct

w

ide-area

n

etwo

rk

s

(packet-sw

it

c

h

ed and c

ir

c

uit-

sw

it

c

h

e

d

networks).

F

o

r

examp

l

e,

traditional

se

r

v

i

ce

providers

lik

e

AT&T,

V

e

ri

zo

n

,

Briti

s

h

T

e

l

eco

m

,

D

e

ut

sc

h

e

T

e

l

e

kom

,

NTT

and

oth

e

r

s

are

a

ll

ti

e

r 1

and

tier

2

ISP

s

(wikip

e

di

a,

20

14).

Th

ese

h

ete

ro

geneo

u

s

networks

r

eq

uir

e

two diff

e

r

e

n

t

d

es

i

g

n

and

management teams

even

with

in the

same o

r

ganizat

i

on

.

For

s

ur

e, own

in

g

and

ope

r

at

ing

two

sepa

rat

e

n

e

twork

s

i

s

in

e

ffici

e

nt and

it

ca

u

ses g

r

ea

t m

a

n

age

m

e

nt

overh

ea

d.

Th

e coo

rdin

a

tion

(16)

b

etwee

n

these two teams

i

s

a

l

so a

noth

e

r

c

halleng

e

to d

e

f

ea

t.

I

t a

l

so

incr

eases

th

e cost

of

n

e

t-works management

,

op

e

ratin

g,

d

es

i

g

n

i

n

g,

p

l

anning

,

a

nd m

a

int

e

nan

ce

which

e

ff

ec

t dir

ec

tl

y

th

e

Tota

l

Cost of

Own

ers

hip

(TCO).

Netwo

rk

s a

r

e

buil

t

basecl

o

n

closed-systems.

R

o

ut

e

r

s

and w

i

tches

fr

o

m

t

h

e

same ve

nd

o

r

h

ave t

h

e a

m

e

priv

ate

f

eat

ur

es a

nd

se

r

v

ic

es.

Th

ese

f

ea

tur

es a

r

e

closed

a

nd k

e

pt

sec

r

et

in id

e

e

a

c

h

ve

ncl

o

r

's

pr

o

clu

ct.

Thi

s sec

r

ecy

and

clo

se

cl

-

box

c

har

acte

r

i

s

ti

c

of

n

et

work n

o

d

es

f

eat

ur

es

a

nd

e

r

v

i

ces s

l

ow

clown

the

n

etwo

rk

s

inno

va

tion

a

nd

i

mprov

e

m

e

nt

s

.

Using

propr

i

eta

r

ies

man-age

m

e

nt

syste

m

s

b

y

eac

h

venclor

c

r

eates

b

a

rr

i

e

r

s

on

f

ace

of

n

etwo

rk d

eve

lopm

e

n

t

in

bot

h IP

a

nd

t

r

a

n

s

port n

etwo

rk

s

. Thu

s

,

it i

s

clear that

man

ag

in

g

two

se

p

a

r

ate

n

etwo

rk

s o

p

e

ratin

g

clifi'

e

r

e

nt

l

y

i

s

i

n

e

ffi

c

i

e

nt

.

Software

d

e

fin

e

d n

etwo

rkin

g

(SD

)

propo

ses

a

n

e

w

a

r

c

hit

ect

ur

e

capab

l

of

man

ag

in

g

d

i

ff

e

r

e

nt n

et

work

s

with

diff

e

r

e

nt

i

nf

r

ast

ru

c

tur

es eve

n

though

w

i

th

differ

e

nt

o

p

e

r

at

ion

a

l

l

aye

r.

Thi

s e

m

e

r

g

in

g

c

oncept

, SD

,

e

n

co

ur

ages

u

s

to

pr

ese

nt

a

com

mon

a

b

st

r

ac

t

that

fit

s

with

b

ot

h

ty

p

es

o

f n

etwo

rk

a

nd prov

i

d

es a co

mmon

a

r

c

h

i

tecture

for

co

ntrollin

g

both n

etwo

rk

s

.

Sorn

e

e

ffort

s

h

ave

b

ee

n clon

e

to

pr

ese

nt SD

-based UCP to

co

ntrol p

ac

k

et a

nd

c

ir

c

ui

t

s

witch

es

u

s

in

g

the

mo

st co

mmon

l

y

known pr

otoco

l

(

Op

e

nFlow).

Most

not

a

bl

y

,

PA

C.C

Da

s

et a

l.

(2010)

b

as ex

p

e

rim

e

nt

e

e[ w

i

th

a

l

ternative

a

pproa

c

h

es

. Oth

e

r pap

e

r

s

Liu

et a

l.

(2

011

,

2013

,

2012)

h

ave

pr

es

nt

e

cl

s

imilar work

as

PAC.C b

y

providin

g

an

e

xp

e

r

i

m

e

n

ta

l

st

ud

y

or

a

Pr

oof

-

o

f-Con

cep

t

to s

upport th

e

u

se

of

Op

e

nFlow

as a

un

i

fi

e

d

co

ntro

l

p

l

ane

.

H

o

w

eve

r

,

Gior-gett

i

e

t

a

l.

(

2012

)

pr

ese

nt

s

a

co

mpari

so

n

st

ud

y

b

et

w

ee

n Op

e

nFlow

a

nd

GMPLS

so

lu

t

i

ons

bas

e

d on

a

i

mul

at

ion.

I

n

th

i

s

work

,

we

propo

se

two

a

ppr

oac

h

es

b

ase

d

o

n Op

e

nF

l

ow

pr

otoco

l

to

c

ontro

l

b

ot

h

optica

l

a

nd

e

l

ec

tr

i

ca

l

n

et

work

s

. Th

e

n

,

w

e expe

r

i

m

e

n

ta

ll

y co

mp

a

r

e t

h

e e

two

so

lution

s

with

a

r

ea

l

imp

l

e

m

e

nt

at

ion

of G

iPLS

a

ppro

ac

h. To

the

b

es

t of our

knowl

edge,

thi

s

i

s

th

e

fir

s

t

work who

co

n

s

i

d

e

r

s

both Op

e

nF

l

ow a

nd

GMPLS UCP

so

l

ut

i

ons

,

a

nd

co

mp

a

r

e

t

h

rn

v

i

a

t

es

tbecl

e

xper

i

m

e

nt

at

i

o

n

.

W

e co

nclu

ct

a

r

eal

case st

ud

y

of

i

mpl

e

menting

e

nd

-to-e

ncl

li

g

htp

at

h

a

nd

a

li

g

htp

at

h r

esto

r

at

ion b

y esta

b

l

i

s

hin

g a

cl

y

n

am

i

ca

l

co

nfi

g

ur

e

d backup

li

g

ht-path

.

Fin

a

ll

y,

we

co

ndu

ct

th

e

co

mp

ar

i

so

n

between

t

h

e

Op

e

nF

l

ow

so

luti

o

n

a

nd

th

e

GMPLS

a

ppro

ac

h b

y s

imu

l

at

ion

on two

r

ea

l

n

etwo

r

k

to

po

l

ogies.

Références

Documents relatifs

Using silica nanoparticles for modifying sulfonated poly(phthalazinone ether ketone) membrane for direct methanol fuel cell : A significant improvement on cell performance..

The total number of wavelengths C net obtained with our method (CPLFT) and SPFF1+1 on Eurocore, NSFNet, EON and UKNet real mesh network topologies, for different connection

Deutlich werden aber auch damit die sehr engen Grenzen, die der EU beim Abschluss völkerrechtlicher Verträge mit Drittstaaten in Bezug auf die Einrichtung

We establish three types of keys for each sensor node according to each type of communication patterns: an individual key shared with the base station, a cluster key shared

Soulignons que, dès le départ, la condition mise à l’indemnisation et propre au non-lieu au sens de l’article 28, paragraphe premier, b, la preuve de l’innocence, avait suscité

Different proposed and standardised routing protocols for LLN exist, including RPL (Routing Protocol for Low-power and lossy networks) [3] and LOADng (Lightweight On- demand Ad

René Maran, considéré par les fondateurs du mouvement et par nombre de spécialistes comme étant son précurseur, émettra des réserves quant au concept voyant dans la

(a) Cumulative global land C storage from 1860 with (red; O-CN) and without (blue; O-C) accounting for terrestrial N dynamics and (b) effect of N dynamics on cumulative global land