Comparison of SDN and a nd GMPLS control techniques in optical transport networks
by Name: Muhammad Muhamma d tif !a !aqub "e# $: MS%&&'(SP)*(++,
-n Partial .ulfillment of "equirements .or the ward of De#ree of
Master of Science in &lectrical &n#ineerin# School of &lectrical &n#ineerin#
Faculty of Engineering Sciences
The University University of Faisalabad Faisalabad
1
S-GN/0"&S
/itle: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
......................... ................... ........... ........
Name of Student: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
"e#istration No : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Super1isory Committee %Name 2 Si#natures' Si#natures'
1.
.........................
(Supervisor) . . . . . . . . . . . . . . . . . . . . . . . . .
2.
.........................
(Member) . . . . . . . . . . . . . . . . . . . . . . . . .
3.
.........................
(Member) . . . . . . . . . . . . . . . . . . . . . . . . .
4.
_________________ ____________________ ___
_______________________ _____ Special member member if any) __________________
2
Certificate Tis Tis resear researc c !or" !or" !as neite neiterr submit submitte# te# earlie earlierr nor !ill !ill be submit submitte# te# else!ere as a re$uirement of M%il & MS #egree. Moreover' no part of te tesis is plagiarie# & copie#.
Name of Student: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Si#nature: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Date: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Super1isor name: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Si#nature: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Date: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Declaration .orm 3
t is certifie# tat MS(EE) tesis title# * +omparison of S,- an# M%/S control tecni$ues in optical transport net!or" * prepare# by Mr. Muamma# 0tif a$ub un#er eg MS(EE)4S%134556 as been approve# for submission.
Tesis Supervisor
7
bstract
M%/S (eneralie# multi4protocol label s!itcing) is te control plane tat is #erive# from te %&M%/S tecni$ue. Te protocols use# in M%/S for routing (8S%F' S4S) an# signaling (S9%) are same tat are use# in %&M%/S net!or". n %&M%/S te /abels are use# to #istinguis bet!een te /S%s but in M%/S te /S%s are i#entifie# base# on T,M signal' !avelengt an# fiber.
8n te oter an# S,- (Soft!are #efine# net!or"ing) serves te main purpose of current current re$uirements re$uirements in optical optical transport transport net!or". t meets te re$uirement re$uirement of #ynamic' #ynamic' smart an# price operative operative services services !it more optimie# optimie# solutions. solutions. S,- uses a common common control plane for all layers use# in optical transport net!or"s !ic ma"e te control plane less complicate# as compare# co mpare# to M%/S. Tere may be multiple s!itcing layers in S,- base# on #ifferent ven#ors !ic may provi#e #ifferent services but te control plane is centralie# for te !ole multi4layer net!or".
n tis paper !e !ill ave comparison on te control tecni$ues use# in M%/S an# S,- for te current re$uirements an# te future tren#s in optical transport net!or". 0fter tat te recommen#ations an# conclusions !ill be ma#e base# on te a#vantages of S,control tecni$ue over M%/S. 0lso te possible strengts of S,- !ill be #iscusse# for future optical net!or"s. +ontrol plane is not $uite easy to implement in optical net!or" as !e ave to observe observe multiple multiple parameters parameters li"e optical po!er calculation' calculation' optical optical signal ban#!i#t an# pat calculation.
First an# foremost !oul# li"e to e:press my sincere gratitu#e to my tesis a#visor' ,r. Sye# 0li Mosin' for is gui#ance' encouragements an# unreserve# elp #uring my t!o years of stu#y. ;e motivate# me to start my master tesis #uring my tir# semester !ic
<
C3N45L&DG&M&N/
allo!e# me to e:ten# my researc stu#y tis far. ;is office #oor !as al!ays opene# for me to #iscuss long ours even #uring after office ours. ;is pursuit of ig stan#ar# in researc an# is maturity in te fiel# !ere te greatest source of my motivation.
My special tan"s soul# go to te Mr. Muamma# =asif for is in #ept researc gui#ance especially in te area of 8ptical fiber +ommunication. n my #aily !or" ave been blesse# !it a frien#ly an# supportive group of fello! stu#ents. !oul# li"e to tan" all my frien#s !o ave provi#e# constant elp an# encouragement an# gratefully ac"no!le#ge teir support in many !ays #uring t!o years of my stu#y. Furter my immense gratitu#e soul# go to all my professional colleagues for teir researc gui#ance' constructive criticisms an# invaluable suggestions.
Finally ta"e tis opportunity to embrace my belove# parents for everyting tey ave #one for me an# tan" my family for teir prayers an# emotional support trougout my stu#ies. #e#icate tis tesis #issertation to my parents.
. bstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
>
/able of contents
cknowled#ements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1i Chapter() -ntroduction to GMPLS and issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .)
1.1.
ntro#uction of
1.2.
M%/S. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 M%/S protocols ............................................ .....3
1.3
M%/S callenges an# te !ay out . . . . . . . . . . . . . . . . . . . . . . . . . . < 1.3.1
,ivision of eneralie#
label. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . > 1.3.2 Formation of /abel s!itce# pat in M%/S. . . . . . . . . . . . . . . . . . . ...? 1.3.3
Mecanism for #ata Transmission variety . . . . . . . . . . . . . . . . . . . @
1.3.7
Structuring of M%/S. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.7.1 %ropose# /abel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 1.3.7.2 ,uple: /abel s!itc pat . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.3.< Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.3.<.1 For!ar#ing 0#Aacency /abel s!itc pat . . . . . . . . . . . . . . . 12 1.3.<.2 +ategorie# label s!itc pat . . . . . . . . . . . . . . . . . . . . . . . 13 1.3.<.3 /in" 0ggregations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.3.>
+onsistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1>
1.3.?
Efficient Bse of available resources . . . . . . . . . . . . . . . . . . . . . . . 1?
1.3.?.1 -umberless connections . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.7
M%/S unresolve# problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.7.1 Safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.7.2 +onnectivity !it oter net!or"s . . . . . . . . . . . . . . . . . . . . . . . . . 1@ 1.7.3 Stability of net!or" . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1@
?
1.7.7 -MS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Chapter(6 Software defined networkin# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6+
2.1 Cat is S,- . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.2 S,- 0rcitecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.2.1 ,ata plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.2.2 +ontrol plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.2.3 ,ata for!ar#ing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.2.7 ,ata #istribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.2.< -et!or" specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.2.> 0#vantages of ,ecoupling of control an# #ata planes . . . . . . . . . . . . 27 2.2.? nteraction of S,- layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2< 2.2.?.1 Soft!are #efine# net!or"ing core controller . . . . . . . . . . . . . . 2< 2.2.?.2 Bse of ven#or base# -et!or" Soft!are . . . . . . . . . . . . . . . . . 2< 2.2.?.3 Bse of -et!or" Management System . . . . . . . . . . . . . . . . . . 2< 2.2.6 S,- layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2< 2.2.6.1 nfrastructure layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2< 2.2.6.2 Soutboun# interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2? 2.2.6.3 -et!or" ypervisors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.2.6.7 -et!or" operating systems . . . . . . . . . . . . . . . . . . . . . . . . 26 2.2.6.< -ortboun# interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2@ 2.2.6.> /anguage base# virtualiation . . . . . . . . . . . . . . . . . . . . . . . 35 2.2.6.? %rogramming languages . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.2.6.6 -et!or" applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3. ,ifferences bet!een M%/S an# S,- . . . . . . . . . . . . . . . . . . . . 31 3.1 ,istribute# an# +entralie# control . . . . . . . . . . . . . . . . . . . . . . 31 3.2 +omplication of control plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6
3.3
Non-Flexibility for new control plane . . . . . . . . . . . . . . .
. . . . . . . . . . . . . 33 3.7 ,ifferent services provisioning . . . . . . . . . . . . . . . . . . . . . . . . . 37 3.< n#ustrial application of S,- . . . . . . . . . . . . . . . . . . . . . . . . . . 3<
3.5.1 !portant bene"ts in Teleco! ind#stry . . . . . . . . . . . . . . . . . . . . . . 35 3.<.1.1 Bse of special soft!are tools . . . . . . . . . . . . . . . . . . . . . . . . 3> 3.<.1.2 +ombine# management . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3> 3.<.1.3 Syncroniation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3>
3.<.1.7 Fle:ibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3? 3.<.2 DTE S,- special features . . . . . . . . . . . . . . . . . . . . . . . . . . . 3? 3.<.2.1 ,ynamic e:tension of services . . . . . . . . . . . . . . . . . . . . . . 3? 3.<.2.2 estoration of services using common controller . . . . . . . . . . 3@ 3.<.2.3 %ossibility of availability . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
@
List of Figures
15
11
12
13
Chapter()
GMPLS
)7) -ntroduction of GMPLS
Tis is te time of ig spee# internet tat is re$uirement of eac an# every in#ivi#ual. 0lso te ig #efinition an# ig $uality vi#eo streaming as become $uite important for use. 0ll tese services re$uire# ig ban#!i#t allocations. For tis purpose te transport net!or" is very important !ic nee#s to ave ig ban#!i#t in it to fulfill te future an# current re$uirements. Te ,C,M is te type of optical transport net!or"s !ic covers te re$uirements in a better !ay. 0lso it is very important tat te control on tese ,C,M systems soul# be !ell optimie# for better use.
nternet Engineering Tas" Force (ETF) starte# !or"ing on te control planes of optical transport net!or"s an# #evelope# eneralie#
Multi4protocol /abel s!itcing
( M%/S). M%/S !or"s as a control tecni$ue in te optical transport net!or"s to calculate te pat bet!een source an# #estination. Te M%/S control plane is for te ol# T,M type transport net!or"s li"e S,;' %,;' !avelengt an# fiber s!itcing tecni$ues similar to M%/S tat is use# for % traffic bet!een s!itces an# routers. M%/S control plane !or"s in suc a !ay tat it #ynamically allocates te resources as per #eman#. Cenever tere is #eman# for resource allocation' ten te me#ia resources are allocate# an# remain spare !en no re$uirements. Tis #ynamic approac of M%/S as selecte# it as a better coice of control tecni$ue in te future transport net!or"s.
17
M%/S as te benefit of being use# for fast #ata transmission in #ata net!or"s comprising of s!itces an# routers. Since nternet protocol tecni$ue is a not connection oriente# tecni$ue but !en te M%/S is use# it as information of bot source an# #estination. Tis pat in #etermine# before te pac"ets are transmitte# from one station to te oter. n or#er to furter enance te pace of #ata transmission in tis type of system' a label is use# to i#entify te for!ar# pat of te pac"et. t not use# te % a##resses to cec" for te #estination % for pat calculation. Ce can also provi#e te uality of service to te traffic using M%/S !it te tables. Te combination of bot te label an# tables ma"es te /S%.
M%/S uses te $ualities of M%/S as a control plane. Te protocols use# in M%/S li"e routing an# signaling protocols are use# in M%/S as !ell for #ifferent T,M' !avelengt an# fiber s!itce# systems. Te main purpose of using te M%/S tecni$ues in M%/S is to provi#e automatic service provisioning meto#s tat re$uire fe!er loa#s on te management an# #ynamically uses te available net!or" resources.
M%/S as goo# features for efficient routing an# transmission of #ata in #ata net!or"s an# !it oter T,M s!itces an# !avelengt multiple:e# systems also a# te re$uirements to ave a generalie# control plane tat can control te #ata transport #ynamically troug tese ol# systems. For tis purpose te M%/S i#ea !as use# tat is also using te /S%s tat !ere use# in M%/S net!or". Te /S% not aving labels in M%/S as !ere in M%/S' but for T,M it is aving time slots' for C,M it is aving !avelengt an# also for fiber s!itcing it is aving fiber port in place of labels.
9ery important feature tat is $uite ne! in optical transport net!or"s an# use# in M%/S is #ynamic an# automate# net!or" management. 0s !e see in S,; all "in#s of services are configure# manuallay from one station to te oter an# !e ave to manullay create te pat traversing all te systems in te pat. Suc process is too lengty an# creates a lot of loa# on te management system operations teams. f one ring is fully use# an# no more resource is available in te ring ten it is essential to upgra#e te ring. 0fter tat te
1<
!olebet!or" !ill be informe# manullay about tis upgra#ation so tat te resources are furter use#.
)76 GMPLS protocols
0ll te routing protocols li"e 8S%F' S4S an# signaling protocols li"e S9%' +4/,% tat !ere use# in M%/S are no! also use# in M%/S control plane. 0ll tese protocols cover te T,M li"e S,; an# C,M for !avelengt net!or"s.
0noter very important protocol for signaling puposes in lin" management protocol (/M%). t is use# to ave better con#itions of control an# #ata planes bet!een te t!o nearby stations.
elo! table41 summaries tese protocols an# te e:tensions for M%/S.
/able()7 GMPLS protocols Description
Protocols
1>
outing
Signaling
/in" Management
8S%F4TE' S4S4TE
S9%4TE' +4/,%
/M%
Tese protocols are use# for #ynamic&auto #iscovery of te net!or"' all te necessary information li"e te available ban#!i#t in te net!or" is also sare# !it all net!or" elements. 0lso !or"s to cec" for te protection pats availability. t sares some very important information !ic is as un#er. t sares !at "in# of protection to te traffic is available (e.g 1 G1' 1 H1 or no protection) n or#er to improve te $uality it uses tese n e! lin"s t broa#casts te information of lin"s tat are not aving % a##resses but te ,s of te lin" -Es an# te interface ,s of bot en#s. t is also important to ave calculate#&i#entifie# te protection pat in case of failure (!ic is sare# ris" lin" group). Tese protocols are very important' use# for ma"ing te /abel s!itce# pats. Te important aspects are as belo!. Since T,M' !avelengt net!or"s are use# in M%/S so generalie# labels are re$uire# instea# of labels in pac"er transport. Te generalie# labels are al!ays bi4#irectional. /i"e routing protocols te signaling protocols also use protection pat for bac"up recovery. 0fter tat te labels are assigne# using te E:pe#iting label assignment via propose# labels. Tese protocols ave follo!ing aspects. Link mana#ement with control(channel H t sares lin" parameters' for e:ample te fre$uency sen#s te information to oter en# to cec" te lin" status an# also lin" strengt using ello protocol. P-NG(test 1erification H Te %- test is use# to cec" te pysical connectivity of te t!o Aoining systems so tat lin" management is verifie#. Correlation of link possessions H t !ill i#entify te properties of te lin" of all no#es connecte# togeter. Te properties may be of protection sceme use# by te no#es. .ault Separation: t !ill separate a single or many faults of te optical #omain.
0ll tese protocols can be seen in belo! Fig41.
1?
GMPLS protocols7
)7* GMPLS challen#es and the way out Since M%/S as to cover #ifferent types of e$uipments !it #ifferent s!itcing types to provi#e a unifie# control plane' teerfore folli!ng important issues are observe#.
1. Forwardin$ of the data is as not %#st related to pac&et. The '()*+ control plane sho#ld be able to provide the facility to forward the data of T,( wavelen$th and "ber switchin$ also as wor&s with pac&ets trac. /. The data carried by pac&ets has the facility that it cab be observed by the networ& N0s who carry this trac. The pac&ets receive the header of pac&et chec& for label and then the o#tp#t port havin$ the forward path is deter!ined and the pac&et !oves to this direction. #t when we tal& abo#t ti!e and wavelen$th division syste!s then the pac&ets are not identi"ed in these syste!s as these syste!s are not !ade for this p#rpose. 3. 2s we &now in T,( ,,( and fre#ency switchin$ the bandwidth reservation of a label switched path is done as per "xed #nits. #t when we see in pac&et networ&s the bandwidth of label switched path can be fro! 1 (bs to 166 (bs. *i&e is
16
T,( we have optical data rates of +T(-1 +T(-4 +T(-17 teherfeore the bandwidth allocation will be only as per earlier !entioned data rates. f we have a pac&et networ& that has *+) bandwidth re#ire!ent of 56 (bps so a "xed bandwidth pipe of +T(-4 is !ore than the re#ire!ents. Therefore the re#ire!ents sho#ld be as per act#al. 4. t is essential that the networ& sho#ld be !ade s#ch that is can cover all the networ& !odi"cation easily and #ic&ly. +ince the T,( and optical networ&s are very lar$e networ&s as co!pared to those of pac&et switched syste!s therefore it is re#ired that the proper networ& chan$e adaptability be ens#red. The wavelen$th and "ber switched syste!s will have to #se h#ndrds of wavelen$ths in it. 5. t is not ti!e savin$ %ob to have electronic and optical switches as ta&es !ore ti!e when there are h#ndreds of o#tp#t ports available for switchin$ fro! inp#t to o#tp#t ports. 2ll these types of networ&s create too !#ch delay in settin$ #p label switched path as !ane inp#t ports will be chec&ed for switchin$ to !any o#tp#t ports. 7. +ince in T,( networ&s li&e +,8 we observe that the a#to!atic protection is observed in these systse!s in case of fail#re of any lin&. The restoration ti!e is 56 !s. The '()*+ ssyste!s sho#ld also ens#re that the fast recovery !echanis! is also adopted in this syste! with dyna!ica#to recovery process. The protection !echanis! !ay be "xed or a#to!aticdyna!ic. 2ll these challen$es are disc#ssed in detail is the below section.
)7*7) Di1ision of Generali8ed label eneralie# s!itcing is re$uire# in te M%/S base# systems tat ave time slots' !avelengts an# fiver ports for s!itcing. Te generalie# approac is use# in it in !ic te #ata is for!ar#e# to te ne:t no#e !itout consi#ering !at "in# of #ata it is an# !at "in# of s!itcing is involve#. Cen !e tal" about
1@
generalie# s!itc it can ave a time slot' !avelengt or fiber as a generalie# label. M%/S also carries te pac"et M%/S labels as !ell to carry pac"et traffic as !ell. Follo!ing important #ata regar#ing lables is available in M%/S. a) /abel s!itce# pat nee# to be i#entifie# tat !at "in# of label is use#' for e:ample pac"et' T,M or !avelengt. b) S!itcing capability nee#s to be verifie# tat !at "in# of s!itcing ability (pac"et' !avelengt of T,M) te -E possesses. c) t is important to "no! !at "in# of #ata&payloa# !ill be !it te /abel s!itce# pat (#ata may be of Eternet' S,;' 0TM etc). /abel s!itcing router of te main upstream no#e sen# te lables to te last #o!nstream no#e. Te upstream no#e suggest labels for te #o!nstream an# ultimately it covers te !ole pat.
)7*76 .ormation of Label switched path in GMPLS -o! !e ave an e:ample to create /S% pat from source to #estination tat as pac"et s!itcing at start. Te T,M net!or" use# aving S8-ET systems an# after tat ,C,M system use# net!or" t!o net!or"s. Eac layer as its #ifferent /abels. Te source is /S1 an# #estination is /S7. n te initial layer from source !e ave 8+412 lin" bet!een pac"et s!itcing layer to T,M net!or". 0fter tat t is net!or" of T,M CT; 8+476 capacity. Ten multiple 8+476 lines combine in te 8I+1.
25
.ormation of LSP in GMPLS network 9 .i#76
Tis cross connect is connecting multiple !avelengts of 8+41@2 #ata rate. -o! !e !ant to create /S% from /S1 to last /S7. -o! it is very important tat te /S%s soul# be establise# suc tat 1st of all te #o!nstream pat /S%s soul# be establise#. Tese /S%s sou l# be tunnele# !it te iger level /S%s to ave complete pat bet!een source to #estication. 0s per Fig.2 te /S% bet!een /S1 an# /S7 !ill ave to carry all te labels of in bet!een net!or" (T,M G C,M). First of all a e$uest for label is for!ar#e# #o!n!ar#s to te net!or" from upstream no#e as te #o!n!ar# no#es !ill sare te /S% first. n tis Fig.2' te T,M system ,S+i !ill for!ar# te re$uest to 8I+1 mentioning te receiver !ic is ,S+e. Tis message receive# by 8I+1 causes an action of one label s!itc pat creation bet!een 8I+1 an# 8I+2. Cit creation of tis /S%' te main /S% bet!een ,S+i an# ,S+e is ma#e !ic is calle# /S%t,i. Cat information is carrie# in tis re$uest for label creation is #eman# for creation of generalie# label for te label s!itc pat. t carries te type of label s!itc pat (protocol of te layer an# te !at "in# of payloa# it !ill carry for e:ample 9+47' 9+43. Some special properties li"e type of signal' automatic protection mecanism' #uple: /abel s!itc pat an# te propose# labels are also available in tis label re$uest. Te #o!n!ar# system sen#s ES9&label mapping message !ic as one generalie# label an# can also ave many generalie# labels. 0s te source /S gets tese generalie# labels from #o!n!ar# systems ten it ma"es a label s!itc pat to te #estination /S !it S9%&%0T; message. Te actions performe# in our case are as un#er. 1. Since a label s!itc pat is establise# bet!een 8I+1 an# 8I+2 so !it tis a complete !avelengt of 15 is available for use. 2. /abel s!itc pat is ma#e bet!een ,S+i an# ,S+e.
21
3. Te T,M no#es at corners of te 52 T,M net!or"s are also aving label s!itc pat. efore tat te label s!itc pats intra net!or"s must be ma#e. 7. Te en# systems of /S2 an# /S3 also ma"e a label s!itc pat. <. 0t last te label s!itc pat establise# bet!een source an# #estination.
)7*7* Mechanism for data /ransmission 1ariety
0s !e "no! tat Multi4protocol label s!itcing systems ave te intelligence of i#entifying te pac"et ea#ers an# ten performing re$uire# actions of #ata for!ar#ing as per te information carrie# by te ea#ers inclu#ing labels an# te #estination system information. Te #ata an# control planes are separate# !en te labels are e:canges bet!een systems.
eneralie# multi4protocol label s!itcing uses te same "in# of approac tat mentione# above. n it te control an# #ata planes are separate# pysically or logically. Ce can ave te option to ave #ifferent pat for te control messages oter tan t e #ata pat. Ce may ave separate S,; signal for control. eneralie# multi4protocol label s!itcing !ill not give or#ers tat te control messages soul# be transmitte# in tis !ay.
Te cost is involve# in coosing any one option for transmission of te control information. t is also not feasible to ave Aust one S,; lin" to transport te control information in a scenario !en te systems are far apart in a S,; ring. 0 goo# i#ea for use of te control messages in te e:isting net!or" !itout involving any cost is to transmit te control information troug ,++ cannel of te S,; signal. Tese overea# bytes !ill be use# in eac S,; lin" of Aoining no#es in te ring' so !e !ill ave no single point of failure. Tis arrangement #oes not cause any ban#!i#t !astage as it may appen if te control cannel uses te space of #ata plane.
)7*7 Structurin# of GMPLS 22
n M%/S te label s!itc pats are create# from te last access si#e. Tese pats may also re$uire te /S% formation of interme#iate systems tat may ave #ifferent structures upto te #estination no#e. Eac of te interme#iate #evices as uni$ue $ualities !ic M%/S as to a##ress so tat /S% creation is fast an# efficient. n or#er to ave insigt about tese #ifferences follo!ing meto#s nee#s to be clear.
)7*77) Proposed Label
Tere is one possibility available tat !e #iscusse# earlier tat tere is an option !it te upstream system tat it can propose a label to te ne:t #o!n!ar# system to spee# up te process of /S% pats formation. ut te #o!n!ar# system can reAect tis label an# can create its o!n ma#e /S%. Tis is a lengty process !ic may ta"e too muc time. f te s!itcing capacity of T,M system is e.g >22 Mbps ten it !ill become long process to allocate labels for multiple traffic. %ropose# labels are use# to spee# up te process of ma"ing /S% bet!een source to #estination. Te S,; system is our e:ample #oes not !ait for te labels from #o!n!ar# 8I+ layers an# as its o!n label. t re#uces te time consumption if ma"ing te /S% en# to en#. ut if te #o!n!ar# system not !illing to use te propose# label ten it can use its o!n cosen label instea# of propose# one.
)7*776 Duple; Label switch path
0s !e "no! tat S,; T,M systems use ring topology for provisioning of te protection pat in case of fiber cut con#itions. /i"e T,M protection !e ave to protect te label s!itc pat to ave te same protection of te pat in case of any pat failure. t can be #one by aving t!o simple: /S%s. 8ne /S% !or"s as protection to te oter.
eneralie# multi4protocol label s!itcing as te option for use of #uple: /S%s !it te signaling protocols li"e S9%&%0T;. t becomes useful to avoi# unnecessary saring of control information' route formation in optical net!or"s.
23
)7*7< Scalin# )7*7<7) .orwardin# d=acency Label switch path For!ar#ing 0#Aacency label s!itc pat is use# in M%/S to get #ata of oter /S%s. t os contructe# among 52 M%/S systems as as its o!n caracteristics an# may be a##e# in routing protocols li"e 8S%F an# S4S. t can be use# in te lin" state algoritm of routing protocols to transport information of oter label s!itc pats. t !ill ultimately lo!er te loa# on #atabase an# time consumption in label searc from te table is re#uce#. Tese For!ar#ing a#Aacency label s!itc pats can be a number of unnumbere# #ata !ic may be attace# !it te oter lin"s information.
.i#ure *7 .orwardin# d=acency
n above Fig.3' !e observe tat te /S%t#m is te %S% !ic is connecting 52pac"et s!itce# net!or"s. t ma"es te sense of Aust one /S% pat bet!een t!o pac"et net!or"s instea# of combination of #ifferent net!or"s.
)7*7<76 Cate#ori8ed label switch path
27
0s !e "no! te net!or" categoriation in access' MET8 an# C,M to carry te traffic from source to #estination' terefore te /S% ierarcy soul# be use# for better us eof te net!or". Te #ata from source to #estination !ill pass troug #ifferent /S%s of #ifferent net!or"s. Te T,M systems ave fi:e# ban#!i#t for eac /S%.
.i#7 >ierarchy of network
Terefore it is not feasible to use a complete 15 !avelengt for Aust a #ata of 15 Mbps. For tis purpose it is better to aggregate te lo!er rate signals into iger rate for better use of resources.
n eneralie# multi4protocol label s!itcing te /S%s are groupe# togeter to aggregate te traffic in main /S% aving large capacity. Te belo! Figure so!s te arcitecture. Figure 5. Hierarchical LSPs
)7*7<7* Link ##re#ations
2<
Since te optical net!or"s are suppose# to ave un#re#s of te optical fiber lines bet!een 52 a#Aacent no#es an# eac fiber may also ave many !avelengts in it. Tis !ill give uge ban#!i#t availability for te eneralie# Multi4protocol label s!itcing net!or". n or#er to efficiently use te net!or" resources it is better to use bun#ling of te net!or" resources.
Cit tis approac of bun#ling all tis information is sare# !it routing protocols. Cit tis bun#ling !e may ave lost some #ata. ut tis meto# as great a#vantage tat te loa# on routing an# signaling protocols is re#uce# too muc e:tent. Tis bun#le carrying eavy #ata as only one control cannel for e:cange bet!een systems.
Te bun#ling can be #one of point to point as !ell as label s!itc pats' !ic are sare# !it te routing protocols.
Ce ave some limitations for te use of bun#ling te lin"s. 0ll tese mentione# belo!.
1. t is co!p#lsory that b#ndlin$ can be done only when the both end so#rce and destination are the sa!e. /. t is also essential that the lin&s added in a b#ndle will be havin$ sa!e type of the lin& for exa!ple point to point or point to !#lti point. 3. t is also bo#nd on the lin& that the lin& which is b#ndled will have si!ilar trac characteristics li&e type of protection and the band pipe #sed.
2>
4. t is also essential that the switchin$ type sho#ld be sa!e for all the !e!bers of the b#ndle for exa!ple Ti!e switchin$ wavelen$th switchin$ or "ber switchin$.
)7*7? Consistency
Tis is very important feature of tis control plane is its capability for #ynamic restoration of faults in te net!or". t is important tat !enever a fault occurs in te net!or" it must be separate# from te net!or" an# restoration must be #one after separating from net!or". Te label s!itc pats tat are alrea#y tunn ele# !it oter label s!itc pats must ave suc "in# of consistency. Te process of te faults resolution is so!n in belo! Figure.>
.i#ure ?7 .ault(Mana#ement Process in GMPLS
eneralie# multi4protocol label s!itcing gives protection of te faulty cannels among bot Aoining systems !ic is calle# span protection !ile source to #estination protection is pat protection. outing protocols broa#cast all te information of tis "in# of protection !it te !ole net!or" an# te protection route is calculate#. 0fter its calculation a signaling is use# !it te S9% an# +4/,% protocols. Te secon#ary pat is calculate# an# !en te fault occurs te signaling protocols s!itc te signals to protection pats.
2?
Cenever a fault occurs te # ynamic fault recovery is acieve# !it sifting traffic to protection pat. Te 52 #ynamic proce#ures of restoration are line an# pat. n line protection alternate pat is selecte# at te system in bet!een source an# #estination' !ile in pat receiver te restoration is starte# from te source s ystem. Te belo! mentione# Figure.? so!s tat system 1 starts for ne! pat. Since muc large time ta"en for s!itcing to protection pat but it is ban#!i#t efficient as no me#ia is alrea#y allocate# for tis protecte# pat.
Figure 7. Protection techniques used in GMPLS
)7*7@ &fficient 0se of a1ailable resources
t nee#s some "in# of optimiation !en !e tal" about use of T,M net!or" !it %4 base# control plane. t nee#s to be #iscusse# te maAor ten#ency of T,M net!or" to use % a##resses of pac"et net!or".
26
)7*7@7) Numberless connections n te time #ivision an# !avelengt s!itcing systems !e #o not use te % a##resses for eac suc traffic. t is calle# numberless connectivity tecni$ue. Te reasons bein# tis tecni$ue are as belo!.
1. +ince the ti!e division and wavelen$th channels are #ite lar$e in n#!ber for which ) address will be assi$ned. t ta&es too !#ch ti!e to assi$n ) addresses for all these channels. /. e do not have eno#$h ) addresses and these are shot for #se.
Tese lin"s !ic are not aving any numbers are i#entifie# by te router number an# some information of te lin" number. Te oter necessary information is carrie# by te signaling protocols as #one !it te lin"s aving i#entifie# by % a##resses.
)7 GMPLS unresol1ed problems
Te M%/S protocols are not stan#ar#ie# an# tere are many issues !ic nee# solution' !ic are mentione# belo!.
)77) Safety
0s !e "no! tat te alrea#y use# pac"et base# systems use te % a##ress to i#entify te traffic an# te pac"et ea#er as te complete information about originator an# te final en# no#e % a##resses. Tese % a##resses are e:clusive in te !ole !orl# an# ave better security penomena. ut te labels !ic are use# in M%/S an# M%/S are Aust limite# to te internal net!or" an# ave noting to #o !it te e:ternal net!or". Terefore tese labels cannot be use# for security policy to give access control to te systems. n or#er to ave security in tese net!or"s it is re$uire# to use tecni$ues li"e tat use# in 0TM an# I.2<.
2@
)776 Connecti1ity with other networks
M%/S success is #ue to its property for connectivity an# !or"ing !it oter net!or" tecni$ues li"e 0TM an# Frame relay. Suc connectivity !it oter net!or"s results in saring of #ata an# control information of 52 0TM net!or"s !en M%/S use# in bet!een 52 0TM net!or"s.
Follo!ing issues are observe# in ma"ing suc connectivity.
1. t is very dic#lt for connectivity of 6/ di9erent control protocols. For exa!ple the ro#tin$ networ& #sed in 2T( is private networ& to networ& interface while for '()*+ it is open shortest path "rst. /. t is #ite i!portant to !ana$e the #ality of service when data !oves in 6/ di9erent networ&s. 3. 2s we &now we can have T,( pac&et switched wavelen$th switched and "ber switched networ&s. This !a&es very few co!binations of '()*+ with 2T( and fra!e relay. The 2T( and fra!e relay have cells and fra!es of data.
)77* Stability of network Cenever a ne! net!or" element i# place# in net!or" or someone remove#' te muc large control #ata is sare# by -Es !it eac oter tan tat of ol# pac"et base# systems. eneric multi4protocol label s!itcing performs TE !ic as information about traffic parameters an# using important protocols li"e /in" Management %rotocol. 0s per my stu#y tis "in# of net!or" consumes large time to ave stability tan tose ol# type pac"et4base# systems.
35
)77 NMS 8l# nternet protocol base# systems use only te % a##ress to ave reacability to tat systems' but !en !e tal" about eneralie# M%/S it as to #eal !it tousan#s of /abel s!itc pats to cec" for traffic monitoring an# $uality of services. t ma"es te -MS more #ifficult to use in suc net!or"s.
Chapter(6
Software defined networkin#
2.1 What is SDN? +oftware de"ned networ&in$ is a techni#e that is re#ired to be #sed in the latest co!p#ter networ&in$ centers in the world. This techni#e perfor!s separation of data and control planes which provide co!!on control scenario in the networ&. The control plane is fo#nd in the software de"ned networ&in$ controller. The path for the pac&et !ove!ent is deter!ined with this controller and then shares this infor!ation with all networ& ele!ents. t all special applications that r#n with the application layer. These applications can r#n dyna!ically or thro#$h the !ana$e!ent syste!. This dyna!ic nat#re of control processes enables the +,N networ& to provide fast service extensions. This control plane !ini!i:es the hardware cost of the optical e#ip!ent with the #se of control infor!ation in software that r#ns over co!!on hardware instead of wor&in$ on optical chips.
31
676
SDN rchitecture
6767) Data plane ,ata plane is te movement of #ata bet!een te routers an# s!itces troug te interface ports. Te s!itces an# routers can be conn ecte# !it ig #ata rate lines li"e in bps. T!o important information #atabases are available in every e$uipment in te net!or". Te 1st is te #atabase to #ifferentiate te pac"ets an# te oter for!ar#s te pac"et to its #estination troug interface ports.
67676 Control plane t contains special proce#ures an# meto#s tat are use# to fin# te pat to route te pac"ets from source to #estination. n te s!itces te control plane cec"s if te pac"et M0+ a##resses matces te #atabase ten it for!ar#s to te #estination' oter!ise if te #atabase not contains te information of pac"et ten it sen#s te pac"et to all s!itces. outers use teir o!n routing protocols for calculating te pat.
Fig46 S,- 0rcitecture
32
Follo!ing are four important parts of tis arcitecture. •
The last two planes of data and control planes are not connected with each other in a sin$le networ& ele!ent. ;ontrol plane will
•
%#st send pac&et to next syste!s. hat and where to forward the pac&ets are only dependent of
•
to send the pac&ets fro! one end to the other. There is a separate syste! available for control of the whole networ&. t has software in it to perfor! these control f#nctions. 2ll re#ired pro$ra!!in$ of other networ& ele!ents is done with
•
this control syste!. There is i!portant software r#nnin$ above the networ& operatin$ syste! which is the !ain control syste!. The software wor&s with the plane of data
Te common control nature as many benefits li"e it is very easy to cange net!or" structure !it ig level soft!are versions. t becomes $uite #ifficult !it system
33
#epen#ent soft!are. Te common control !or"s imme#iately !enever ne! net!or" canges are observe#. Te control plane as complete information of te net!or" canges an# ten acts accor#ingly.
6767* Data forwardin# ,ata for!ar#ing is very important an# uni$ue feature in S,-. Te for!ar#ing information #oes not carry any information of te source ar#!are an# i#es its information.
6767 Data distribution n S,- te #ata #istribution control is also common. Te net!or" operating system carrie# tis #istribution plane. t performs #o!nloa#ing net!or" #istribution information of te net!or" elements an# gets all necessary information about te net!or" elements an# connectivity bet!een tem for proper #istribution.
6767< Network specification -et!or" specification is #one in net!or" operating system !it a re$uire# beavior of net!or"' !ic is suppose# to be not #one by control plane. t is a virtual tecni$ue in !ic te specific configurations are mappe# o n pysical net!or" setup basis. elo! Fig4@ so! te #etail.
Fig.@ S,- arcitecture !it abstraction
37
6767? d1anta#es of Decouplin# of control and data planes Follo!ing are important benefits of using te separate control an# #ata planes. •
t is #ite easy for r#nnin$ new applications as the
•
applications are shared d#e to co!!on control plane. t is seen that the applications r#n on the sa!e networ& for which this this control plane is #sed so the other
•
parts of control plane are free for other services. The control plane ta&es the actions for any networ& ele!ent. n this way no need to have acc#rate proced#re for new networ& re#ire!ents for data
•
trans!ission. 2ll &inds of applications can be interconnected. e can add di9erent applications b#t the priorities are $iven to certain applications.
3<
6767@ -nteraction of SDN layers 6767@7) Software defined networkin# core controller t is te main part in te operation an# maintenance layers of te S,- an# "eeps all te net!or" relate# information !it it. t performs main operations to provi#e re$uire# actions in te net!or".
6767@76 0se of 1endor based Network Software Tis is soft!are !ic is ma#e by te e$uipment manufacturer. t is installe# in te system !ic manages te net!or". Te user gets it accesse# from anoter computer an# !or"s on it troug some comman#s. Te specific ven#or base# touc of tis soft!are as"s user to use specific comman#s !ic are only base# on te ven#or e$uipment. Tis soft!are is responsible to manage te !ole net!or" eiter single layere# or multi layere#.
6767@7* 0se of Network Mana#ement System -et!or" Management system is use# for some lo! level configurations of te transport systems. t is use# for ma"ing te arrangement so tat te net!or" elements are online an# accessible for furter operations. Tis soft!are ma"es configurations of transport no#es regar#ing teir ,' % an# te lin" establisment bet!een te n et!or" elements. Tis soft!are is also ven#or base# an# eac ven#or use# its o!n specific net!or" management soft!are. /i"e one ven#or uses B31 soft!are for tis purpose. Tis is B base# soft!are !ic enables users to ave performe# all operations efficiently.
6767, SDN layers 6767,7) -nfrastructure layer
3>
Tis is te layer !ic is relate# to ar#!are e$uipment installe# in te net!or". Tese e$uipment are not aving any control information in tem as S,- as te centralie# control plane. Te #ata plane is not aving any intelligence in it an# all is available in control plane.
Fig415 S,- in (a) planes (b) layers (c) arcitecture
Te intelligence is completely available in te net!or" control an# application plane. t is seen tat tese net!or"s are establise# along !it 8pen Flo! stan#ar#s so tat #ifferent "in#s of net!or"s elements can be use# togeter for #ata for!ar#ing. Te ol# transmission e$uipment li"e S,; are not aving tis fle:ibility as many built4in ar#!are re$uirements must be fulfille# in tese e$uipment. Te belo! Fig411 so!s te S,- control an# #ata #evices.
3?
Fig411 8pen flo! enable# S,- #evices
n te above Fig411 te S,- elements are so!n. 8ne is te control element an# oter is for!ar#ing one. Te control plane as a matc table !ic as parts of rules' actions an# statistics of tese #ata pac"ets. Te table so!s te process troug !ic te pac"ets ave to move. Ce ave te option to coose one of te rule out of tese mentione# for a pac"et to #istinguis. Te actions !it te pac"et may be sen#ing to te out ports' sen# to control plane' #iscar#' for!ar# to ol# for!ar#ing line !ic is not open flo!.
6767,76 Southbound interface Tese interfaces are in bet!een control an# for!ar#ing planes' !ic clearly separate tese t!o important fiel#s. 8pen flo! fiel# as ma#e it very easy for te investors to elp tem invest in te ne! tecnologies tat can integrate multiple #ifferent types of e$uipment. Soutboun# interface performs some important actions. t sares te information !it control plane !en te #ata port of for!ar#ing #evice is cange#' oter is te traffic information is also sare# !it te controller an# te tir# is tat it sare te information of #ata for !ic te #ata plane as no information about its for!ar#ing.
36
6767,7* Network hyper1isors Te ne! latest computer net!or"s are using a ne! tecni$ue of simulations. t as been ma#e effective in last fe! years. Te resources are use# base# on simulate# ar#!are not te pysical ar#!are resources. Tis layer as"s for te use of tese simulate# resources instea# of as"ing for te pysical resources. Te user #eman#s for resources only !en tere is te re$uirement. t is great benefit regar#ing revenue as te resources are al!ays sare# bet!een users. Tis tecnology as te important benefit tat tese sare# resources can be sifte# from any pysical macine to any oter pysical ar#!are !ic provi#es fle:ible net!or" nature. ut still rigt no! it is not freely use# an# te ol# pysical fi:e# resources are use# in net!or"s. Te capacity of any net!or" can be calculate# !it its structure an# te available place to ave net!or" information. Te net!or" layers of #ifferent type re$uire #ifferent types of net!or" resources. Te layers can be /ayer42' /ayer43 etc. 8ne net!or" structure cannot provi#e all "in#s of re$uirements. Te simulate# net!or"s are use# to ave best use of te resources. Te %v> !ill ave limitations for use as macines may be using %v7 sceme.
6767,7 Network operatin# systems t is observe# tat te ol# types of net!or"s ave te operating systems !it tem !ic can only !or" !it basic level of e$uipment an# operate tem !it ven#or specific soft!are. n a##ition to tat some special comman#s !it #ifferent programming tecni$ues are use# to furter better use tese #evices. ,ifferent ven#ors ave #evelope# teir o!n net!or" operating systems !ic are very !ell in te use an# are fulfilling te re$uirements in better !ay as per service #eman#. Te researc is tere to ave common control of te !ole #iverse net!or"s !it te routing protocols !ic !or" !it #ifferent "in#s of net!or" e$uipment. Soft!are #efine# net!or"ing as provi#e# a common net!or" operating system !ic gives protocols tat are use# by all "in#s of net!or". Tere is no re$uirement to see for te #etails of #ata tat is for!ar#e# bet!een #evices in tis operating system.
3@
Te control plane is run by tis net!or" operating system. Te maAor classification of te control planes is #one in 52 categories. Te one is te common an# oter is #ivi#e#. Te common control soft!are is a one source to manage all te net!or" entities. t as one soft!are source an# tere may be any resource limitations in tis "in# of net!or". Te common resources ave very ig limit for #ata for!ar#ing an# tousan#s of #ata for!ar#ing can be manage# !it tese controllers in one secon#. Te oter one !ic is #isperse# soft!are is ma#e as to an#le a sort or big net!or". t may be centrally manage# or pysically #istribute#. Te first one !ic is centralie# can be use# for con#ense# #ata net!or"s. Te oter is muc ar# to te #ifferent "in#s of failures. Te separate# control planes ave te a#vantage tat !en a system is faile#' ten te oter system ta"es loa# of tis faulty one. n a single control system' !e ave only one system to perform control actions an# all control function !ill be affecte# !en tere is fault to tis controller. Te controller performs important functions. t performs running of a program' activities base# on comman#s' service safety in case of any net!or" failure. n a##ition to tat te net!or" structure' control of #evices in te net!or"' #ata sen#ing on smaller pat an# #ata safety is also manage# by controllers.
6767,7< Northbound interfaces Te nort boun# interfaces are performing te operations of abstraction an# implement languages to program. Tere are #ifferent programming languages available for tis purpose. 8ne program use# /inu: language !ic use# 9irtual file system application. t is elpful for systems tat are !or"ing on layer41J2. Te programming is #ifferent for eac operation as te applications li"e #ata for!ar#ing' routing an# protection an# #ifferent for eac oter. Some applications are base# on involvement of te applications in te systems. Tere are certain rules !ic !or" for tese applications. Te net!or" resources are allocate# automatically !en tere is #eman# for a particular service. Cen tere is a re$uirement of voice an# #ata services ten it is possible tat uni$ue ban#!i#t is allocate# to tat service as per service re$uirement. Tere is also cec" to verify tat te ban#!i#t allocate# #oes not surpass te set one by te net!or" manager.
75
6767,7? Lan#ua#e based 1irtuali8ation 9irtualiation as te ability to provi#e te service tougt an# also ensures te $ualities li"e protection. t so!s te i#ea of one large s!itc !ic so!s a##ing many for!ar#ing e$uipment !ic for!ar# te #ata. Suc centralie# approac re#uces te comple:ity in #ata for!ar#ing as becomes easy to ma"e te net!or". 8ne suc "in# of language as te i#ea of obAects. Eac obAect as its o!n re$uirements of #ata for!ar#ing. Te obAects #o not sare te information !it eac oter for any particular service.
6767,7@ Pro#rammin# lan#ua#es Te researc base# on moveable program is $uite latest in te latest computer net!or"s. Te ;ig level co#es are elpful to resolve many issues face# on lo! levels. Te ig level as follo!ing important aspects. •
The codes re#ired to pro$ra! the devices >which forward the
•
data? are !ore si!pli"ed. The pro$ra!!in$ is !#ch proble! resol#tion oriented which
• •
beco!es #ite easy for the peoples involved in codin$. t enables to re#se the code as per re#ire!ent. (a&es the networ& virt#ali:ation possible
n soft!are #efine# net!or"ing it !ill become very #ifficult tat one application Aobs #o not affect eac oter. t is foun# tat every application as its o!n principles for #ata for!ar#ing !ic may be #ifferent for oter application. n tis !ay te systems tat are re$uire# to sen# te #ata to ne:t en# !it face some clases. Tis issue can only be resolve# !it te languages use# to program te applications. Te important benefit of tis programming is its ability to buil# for virtual structure. Te i#ea of obAect oriente# is also te same. t enables te programmers to consi#er only te re$uirements an# not tin" about te type of #ata use# an# o! to manage tis #ata. ;ig level languages serve te purpose of soft!are #efine# net!or"ing !it $ualities li"e its mo#ular approac' composition covering !ole net!or" an# #ata verifications. Te lo! level #esign as many callenges an# to cover tese callenges te ig level is use# !it approac as mentione# belo!.
71
1. This does not !ana$e the service con"$#ration on each syste! and foc#ses on the whole networ&. +#ch iss#es also observed in old syste!s. /. (ana$e!ent wor& is !ade easy with the networ& str#ct#re. 3. ,i9erent %obs for data forwardin$ are not connected with each other. 4. @esolves iss#es which avoid data forwardin$ with dyna!ic approach. 5. @ed#cin$ the ti!e for data forwardin$. 7. !!ediate addressin$ of any &ind of fa#lt.
6767,7, Network applications t is ta"en as te net!or" core. %articular control information is cange# into te form of instructions tat furter #river te #evices !ic are use# to sen# te traffic from one en# to te oter. Te i#ea in it is tat te pat is #efine# bet!een te source an# #estination. For tis purpose a special instruction if follo!e# by all systems from source to #estination. S,- is use# for any type of net!or" !ic may be small or large. Tese ol# systems can perform te oter applications of te net!or"s li"e te /aye42 an# /a yer43 protocols. ut special features can be performe# !it te S,- net!or" only. Soft!are #efine# net!or"ing as important features li"e 8ptimiation of #ata for!ar#ing' #ata protection an# interconnectivity about #evices. 9ery important application of Soft!are #efine# net!or"ing is to balance te loa# in net!or". t combines te multiple re$uests from #ifferent users an# #oes not nee# control function for every re$uest. t is $uite useful to balance te loa# as per net!or" re$uirements. t avoi#s e:erting loa# on a single net!or" element or lin". n soft!are #efine# net!or"ing it is possible to ave routers configure# #ynamically. t elps routers to ave less #ata base of routing tables an# routers are configure# !en tere is some #ata transmission re$uire#.
72
Chapter(*
*7 Differences between GMPLS and SDN *7)
Distributed and Centrali8ed control
n S,- !e enAoy te benefit of centralie# control plane !ile M%/S as te #istribute# control plane. n M%/S eac no#e as to perform te control functions regar#ing i#entification of lin" capacity' available ban#!i#t an# service security as per re$uirement of service. 0ll no#es ave te complete information about te mecanism to select a lin" route in case !en any service is re$uire# to be route# from Source to #estination in a net!or". n te #istribute# control scenario te 1 st no#e also as te complete information about te #estination an# te pat re$uire# to be use# for te pac"et. Eac net!or" element as to ma"e te lin"s in suc a !ay tat te service pat is complete# an# all tis is #one at te same time. n suc scenarios it becomes very #ifficult for te net!or" to ave optimum utiliation of te net!or" resources. 0ll no#es !or" togeter to ma"e te connections establise# an# te net!or" becomes inefficient for te traffic. Cile te centralie# one as te facility to use te net!or" in more optimie# !ay. Te #istribute# control cannot manage te efficiency of te routes from source to #estination an# te same is possible !it Soft!are #efine# net!or"ing. n tis control tecni$ue all te net!or" relate# information li"e pats' connectivity of no#es an# ban#!i#t available is store# in common soft!are. t computes te pat for te services centrally an# can manage in a better !ay as it as full picture of te net!or". Te ban #!i#t re$uirement if also fulfille# as te service pat canging is #one as per #eman#. Cit tis te net!or" is use# in efficient manner. Since it as complete information of te net!or" so centralie# control plane provi#es opportunity to control te issue of #elay in #ata transmission #ue to long pat usage. Cit tis tecni$ue !e can use te optimum pat to avoi# #elay in #ata transmission. Soft!are #efine# net!or"ing as 52 important interfaces. 8ne is for te #eman# of net!or" resources from te net!or" an# oter is for connectivity to -et!or" management systems !ic as full autority to use te net!or". 8ne #isa#vantage of
73
te centralie# control plane is tat it is not as fast as M%/S control plane !en te canges in net!or" are observe#. Cen any no#e is #o!n or any connectivity bet!een no#es is #isturbe# te same cannot be instantly conveye# to te centralie# control plane. M%/S as #istribute# control an# is efficient in tis scenario. n Soft!are #efine# net!or"ing' control #omain is not place# at te same location as it may be apart from eac oter for protection of te services. Eac separate control #omain nee#s to be in close connectivity !it all oter control #omains so tat all information bet!een tese controllers is sare# an# no ambiguity observe# in te net!or". t is alos possible tat eac separate control #omain may be of #ifferent bran# an# company. Te control systems soul# be able to arrange communication bet!een tem even if tey are not of same type. Te #ifference in tese t!o control tecni$ues is as u n#er.
Fig412 (a) S,- control plane !it centralie# control (b) M%/S control plane !it #istribute# control
77
*76
Complication of control plane
Te control plane becomes muc complicate# !it te use of M%/S. Te routing protocols use# in it li"e 8pen Sortest %at First are not beneficial for te net!or" tat is altering its con#ition $uite fre$uently. t ma"es te net!or" less reliable. Similarly te eservation protocol tat !as initially use# for te en# user to get te net!or" resources as no! been use# in Multi4%rotocol /abel s!itcing. t elps te net!or" to get te /abel s!itcing pat resources. Cit te same i#ea tis protocol use# in eneralie# Multi4%rotocol /abel S!itcing. 0ll tese a#vancements in tis protocol ma#e it more an# more complicate# an# re#uce# its efficiency. M%/S al!ays tries to ave information relate# to #ifferent types of e$uipment in te net!or". Te in#epen#ence of tese e$uipment an# retaining te i#entity ma"es te net!or" $uite more complicate#. -et!or" complication is re#uce# !it te use of Soft!are #efine# net!or"ing. n it tere are no 8pen sortest pat first an# eservation protocols an# te unifie# interface of 8pen Flo! is use# along !it many operating systems !ic are interconnecte# !it eac oter.
*7* Non(.le;ibility for new control plane n M%/S tere is no option available for resilience to te use of control plane tat is ne! in te net!or". Te people !o are to operate te net!or" #o not fin# it easy to us te ne! control plane !it te e:isting. t ta"es a long time !en te ne! control plane can syncronie !it te nee#s of te net!or". Terefore it is te continuous process tat te net!or" operators soul# be loo"ing for te ne! plane tat is very muc close in process to te e:isting. n S,- !e use to a## te control plane !it a ne! tecni$ue in !ic te ne! control plane is place# among te alrea#y !or"ing control planes. Tis control plane ten #ivi#es te circuit plane into many parts. Eac part is ten controlle# by a #ifferent control plane. Te net!or" operator as only control over te #ata plane tat as not been #ivi#e#' !ile all te oter #ivi#e# #ata planes are controlle# by te separate control planes. Cit tis #ivision tere is also benefir tat te control plane #oes not ta"e pac"et an# circuit
7<
s!itcing separately an# only one #ata plane is consi#ere# if controlle# by a single control plane.
3.4 Dierent services provisioning eneralie# multi4protocol label s!itcing provi#es #ifferent services for te users !ic ma"e #ifferentiate# from te Soft!are #efine# net!or"ing. M%/S provi#es services of optical virtual private net!or". n tis net!or" eac user is saring te optical fiber resources bet!een te no#es. Eac user is allocate# separate me#ia resources in te M%/S net!or" !ic ma"e tem in#epen#ent in service provisioning as !ell as operation an# maintenance.
Fig413 +onnectivity of 89%- users in a net!or"
7>
0s per above mentione# #iagram' te #ifferent optical virtual private net!or" users are connecte# to one management server. ot te users ave #ifferent lin"s for operation an# maintenance of te net!or". Tese users can manage teir M%/S services from te net!or". Eac net!or" management user first allocates resources in form of ban#!i#t to te customer. 0fter tat it monitors te traffic associate# !it tat customer an# performs necessary operations !en tere is furter re$uirement of me#ia resources for tat particular customer. 0noter important M%/S service is user net!or" interface (B-) service !ic is provi#e# troug te M%/S net!or". Cen a user connect to te M%/S systems troug tis interface ten it automatically sen# te M%/S re$uest to te net!or" to get te services. n tis service eac user !o is as"ing for connectivity to te oter user sen#s all information regar#ing te source no#e a# #ress an# sin" no#e a##ress. Te no#es on receiving tis information cange te no#e a##resses into teir i#entification numbers. 0fter tat te pat is calculate# an# te en# to en# trail is ma#e. Te service level is also available in te message information generate# by te user. 0fter tat te pat is i#entifie# as per service level.
*7<
-ndustrial application of SDN
S,- in transport net!or" is a ne! i#ea for te generaliation of te net!or"s. t is separate planes for transport an# control. t is use# in core net!or"s to ave ne! applications. Te use of S,- in Telecommunication in#ustry as ig importance regar#ing its ne! features. Soft!are #efine# net!or"ing is provi#ing follo!ing important features to Telecom in#ustry !ic !ere not provi#e# by M%/S an# ol# Transport tecnologies.
*7<7) -mportant benefits in /elecom industry Te important benefits gaine# by S,- over M%/S is telecom in#ustry as belo!.
7?
*7<7)7)0se of special software tools Te ol# transport tecnologies li"e S,;' ,C,M' 0S8- an# M%/S in transport net!or" !ere not aving any features to use special soft!are tools to properly plan te telecom net!or". Te planning !as #one only using te manual approac. Soft!are #efine# net!or"ing as given te goo# features to plan te net!or" using a soft!are tool. 0lso all te planning !as #one prior to commissioning of te net!or". So it #i# not provi#e te real time application. Cen tere !as some abnormality in te real time !or"ing of te systems ten it !as not an#le# by ol# systems. S,- gives real time solution to te problems using proper coor#ination bet!een te tools.
3.5.1.2 Co!ine" anageent Tis is te important feature available in S,- tat it provi#es a simple an# easy management of te tools to manage te e$uipments. t elps to lo!er te e:penses to maintain te e$uipment an# also can !or" to improve te e$uipment as per te re$uirements.
*7<7)7* Synchroni8ation S,- provi#es an important feature tat it provi#es te syncroniation bet!een transport e$uipments of #ifferent ven#ors' !ic !as not possible in legacy transport tecnologies. 0long !it tis S,- provi#es te syncroniation bet!een te core net!or" an# te transmission net!or" to ave a single control plane. t gives a simple un#erstan#ing tat managing a single control plane is easy oter tan to !or" on #ifferent control planes of #ifferent ven#ors.
*7<7)7 .le;ibility Te fle:ibility of te S,- planes ma"es it possible for interconnection of #evices tat are using multi4layer services. Te Multi4layer #evices ma"e it possible to !or" togeter !it a common control plane of S,-.
76
*7<7)7< Directness Tis is te important feature available in S,- tat it gives #irection tat any ven#or e$uipment can be connecte# to any oter ven#or e$uipment using te unifie# control arcitecture. 0ll tese important features are available in one of te top ven#or in Telecommunication ven#ors in te !orl#. Tis is DTE Tecnologies !ic as tis important feature. DTE is te no! te lea#ing company to provi#e te transport e$uipments in #ifferent parts of te !orl#. DTE as #evelope# te important features of S,- !ic are !ell beyon# te scope of legacy transport net!or"s. Te important features use# by DTE are so!n belo! in #etail.
*7<76 A/& SDN special features DTE provi#es follo!ing special features !it te use of S,-.
*7<767) Dynamic e;tension of ser1ices Since S,- is using te important feature of centralie# controller' terefore it is possible tat a common centralie# controller can !ell manage te net!or" to provi#e services from one en# to te oter. Since all type of te systems are controlle# by tis common controller' so tis controller as te #atabase of all te available resources of te net!or". t cannot be possible to provi#e suc #ynamic services using te #istribute# control plane of te ol# transport tecnologies. Te #istribute# control system of M%/S gives follo!ing #ra!bac"s !ic re#uce efficiency of te transport e$uipments.
•
The distrib#ted control syste! is #sed to cover only the fa#lts of speci"c areas and is #nable to cover the whole networ&. +ince the whole networ& co!prises of !#ltiple vendors or !#ltiple technolo$ies
therefore
proper
synchroni:ation
distrib#ted control planes is not possible.
7@
of
di9erent
•
hen a new networ& is $oin$ to be established or existin$ networ&in$ is expanded then it can be done #sin$ the infor!ation to plan the networ& for service con"$#rations. 2fter #sin$ a plannin$ tool the available data is exported to the networ& !ana$e!ent syste!. 2fter that con"$#ration is done.
•
2ll this lon$ process red#ces the syste! eciency. hen the new ro#tes are discovered for new service provisionin$ and the data is shared with the N(+ then the services can be created on the networ&. This also red#ces the eciency of the networ& as cons#!es too !#ch ti!e fro! networ& plannin$ to service provisionin$.
Te S,- provi#es a common controller. Cenever tere is te re$uirement of ne! services. Te controller automatically searces for te ne! service routes. n tis !ay te S,- feature ma"es te net!or" to give complete management. esultantly te efficiency is improve#' !ic is te main feature of S,-.
<5
.i#7 Dynamic e;tension of ser1ices *7<7676 "estoration of ser1ices usin# common controller Cen any fault is observe# on !or"ing or protection label s!itce# pat of any trail service ten te correspon#ing net!or" elements as" centralie# control plane for te restoration of te services. Te centralie# controller fin#s te most feasible route for te ne! pat. Te ne! pat is ten for!ar#e# to te specific net!or" elements. Te -Es get tis information an# ten configure te services. Tis restoration of te services is ma"es te net!or" more efficient. Te controller as all te information about te net!or". Terefore te rerouting of te services becomes $uite fast. Te service rerouting can be #one base# on te priority as !ell. t gives te planning information base# on te priorities.
Fig. estoration of services troug centralie# controller
<1
*7<767* Possibility of a1ailability Transport S,- as te option to prompt about te cance of any fault occurrence in te net!or". t gives analysis about any fault occurrence. t uses simulation to cec" for te net!or" status #uring any fault occurrence. Troug simulation te S,- control plane gives real time !arning about te net!or" survival. Te user performs te simulation to cec" te result of fault. Cen te fault is create# ten S,- control plane cec"s te available resources an# calculates tat !ic services !ill be interrupte#. t computes te pat of te services !ic !ill be reroute# an# igligts tose services !ic !ill not be restore#.
Fig. 0nalysis to cec" survival of resources
<2
"eferences K1L 0 %er4,omain %at +omputation Meto# for Establising nter4,omain Traffic Engineering (TE) /abel S!itce# %ats'N ETF F+ <1<2' February 2556. K2L 0 ac"!ar#4ecursive %+E4ase# +omputation (%+) %roce#ure to +ompute Sortest +onstraine# nter4,omain Traffic Engineering /abel S!itce# %ats'N ETF F+ <771' 0ugust 255@ K3L Te 0pplication of te %at +omputation Element 0rcitecture to te ,etermination of a Se$uence of ,omains in M%/S an# M%/S'N ETF F+ >65<' -ovember 2512 K7L %. ovanna' 0. ermoni' F. Testa' . +ossu' 9. /ope' . Sabella' * Multi4layer +ontrol for %ac"et48ptical -et!or"s'* in Oournal of 8ptical +ommunications an# -et!or"ing' 9ol.<' ssue 15 pp.06>40@@.
A5B Tho!as 2. *i!oncelli. CpenE?D4/4G 2#$#st /61/.
<3
A7B 2.T. ;a!pbell . Hat:ela H. (i&i and I. Jicente. Cpen si$nalin$ for at! internet and !obile networ&s >opensi$KLE?. 2;( +';C(( ;o!p#ter ;o!!#nication @eview /L>1?DLG16E 1LLL. AGB 2. ,oria F. 8ellstrand H. +#ndell and T. orster. 'eneral +witch (ana$e!ent )rotocol >'+()? J3. @F; 3/L/ >)roposed +tandard? I#ne /66/. AEB ,.*. Tennenho#se I.(. +!ith .,. +incos&ie ,.I. etherall and '.I. (inden. 2 s#rvey of active networ& research. ;o!!#nications (a$a:ine 000 35>1?DE6E7 1LLG. ALB ,.*. Tennenho#se and ,.I. etherall. Towards an active networ& architect#re. n ,2@)2 2ctive N0twor&s ;onference and 0xposition /66/. )roceedin$s pa$es /15. 000 /66/. A16B I.T. (oore and +.(. Nettles. Towards practical pro$ra!!able pac&ets. n )roceedin$s of the /6th ;onference on ;o!p#ter ;o!!#nications >NFC;C(?. ;iteseer /661. A11B ,evolved ;ontrol of 2T( Networ&s. httpDwww.cl.ca!.ac.#&researchsr$netosold-pro%ectsdcanMp#b . A1/B I. 0. Jan ,er (erwe and . (. *eslie. +witchlets and dyna!ic virt#al at! networ&s. n )roc nte$rated Networ& (ana$e!ent J pa$es 355 37E. ;hap!an and 8all 1LLG. A13B I.0. Jan der (erwe +. @ooney . *eslie and +. ;rosby. The te!pesta practical fra!ewor& for networ& pro$ra!!ability. Networ& 000 1/>3?D/6/E 1LLE. A14B I. @exford 2. 'reenber$ '. 8%al!tysson ,.2. (alt: 2. (yers '. ie I. =han and 8. =han$. Networ&-wide decision !a&in$D Toward a wafer-thin control plane. n )roc. 8otNets pa$es 5L74. ;iteseer /664. A15B 2. 'reenber$ '. 8%al!tysson ,.2. (alt: 2. (yers I. @exford '. ie 8. Oan I. =han and 8. =han$. 2 clean slate 4d approach to networ& control and !ana$e!ent. 2;( +';C(( ;o!p#ter ;o!!#nication @eview 35>5?D4154 /665. A17B O. Iarraya T. (adi and (. ,ebbabi P2 s#rvey and a layered taxono!y of software-de"ned networ&in$Q ;o!!#nications +#rveys T#torials
<7
000 vol. )) no. LL pp. 11 /614. A1GB N. Fea!ster I. @exford and 0. =e$#ra PThe road to +,NQ R#e#e vol. 11 no. 1/ pp. /6D/6/6D46 ,ec. /613. A1EB @. )res#hn PJersion / of the )rotocol Cperations for the +i!ple Networ& (ana$e!ent )rotocol >+N()?Q @F; 3417 >NT0@N0T +T2N,2@,? nternet 0n$ineerin$ Tas& Force ,ec. /66/. ACnlineB. 2vailableD httpDwww.ietf.or$rfcrfc3417.txt A1LB I. )an +. )a#l and @. Iain P2 s#rvey of the research on f#t#re internet architect#resQ ;o!!#nications (a$a:ine 000 vol. 4L no. G pp. /737 I#ly /611. A/6B N. Fea!ster and 8. ala&rishnan P,etectin$ ') con"$#ration fa#lts with static analysisQ in )roceedin$s of the /nd conference on +y!posi#! on Networ&ed +yste!s ,esi$n S !ple!entation - Jol#!e / ser. N+,K65. er&eley ;2 U+2D U+0N 2ssociation /665 pp. 43 57. A/1B @. arrett +. 8aar and @. hitestone P@o#tin$ snaf# ca#ses internet o#ta$eQ nteractive ee& 1LLG. A//B H. #tler T. Farley ). (c,aniel and I. @exford P2 s#rvey of ') sec#rity iss#es and sol#tionsQ )roceedin$s of the 000 vol. LE no. 1 pp. 1661// Ian /616. A/3B I. +herry and +. @atnasa!y P2 s#rvey of enterprise !iddlebox deploy!entsQ 00;+ ,epart!ent University of ;alifornia er&eley Tech. @ep. U;00;+-/61/-/4 Feb /61/. A/4B H. 'reene P(T Tech @eview 16 rea&thro#$h Technolo$iesD +oftware-de"ned Networ&in$Q httpDwww/.technolo$yreview.co! article41/1L4tr16-software-de"ned-networ&in$ /66L. A/5B ). New!an '. (inshall and T. *. *yon Pp switchin$S!dashat! #nder ipQ 0002;( Trans. Netw. vol. 7 no. / pp. 11G1/L 2pr. 1LLE. A/7B N. '#de T. Hoponen I. )ettit . )fa9 (. ;asado N. (cHeown and +. +hen&er PNCD towards an operatin$ syste! for networ&sQ ;o!p. ;o!!. @ev. /66E.
<<
A/GB 8. Ia!%oo! ,. illia!s and U. +har!a P,onKt call the! !iddleboxes call the! !iddlepipesQ in )roceedin$s of the Third or&shop on 8ot Topics in +oftware ,e"ned Networ&in$ ser. 8ot+,N K14. New Oor& NO U+2D 2;( /614 pp. 1L/4. A/EB 8. 2l&hatib ). Faraboschi 0. Frachtenber$ 8. Hasahara ,. *an$e ). *aplante 2. (erchant ,. (ilo%icic and H. +chwan P000 ;+ /6// report >draft?Q 000 ;o!p#ter +ociety Tech. @ep. Febr#ary /614. A/LB (. ;asado N. Foster and 2. '#ha P2bstractions for softwarede"ned networ&sQ ;o!!#n. 2;( vol. 5G no. 16 pp. E7L5 +ep. /614. A36B 2. ,oria I. 8. +ali! @. 8aas 8. Hhosravi . an$ *. ,on$ @. 'opal and I. 8alpern PForwardin$ and ;ontrol 0le!ent +eparation >For;0+? )rotocol +peci"cationQ nternet 0n$ineerin$ Tas& Force (ar. /616. ACnlineB. 2vailableD httpDwww.ietf.or$rfcrfc5E16.txt A31B 8. +on$ P)rotocol-oblivio#s Forwardin$D Unleash the power of +,N thro#$h a f#t#re-proof forwardin$ planeQ in )roceedin$s of the +econd 2;( +';C(( or&shop on 8ot Topics in +oftware ,e"ned Networ&in$ ser. 8ot+,N K13. New Oor& NO U+2D 2;( /613 pp. 1/G13/. A3/B T. ,. Nadea# and H. 'ray +,ND software de"ned networ&s 1st ed. CK@eilly /613 vol. 1. A33B N. (. H. ;howdh#ry and @. o#taba P2 s#rvey of networ& virt#ali:ationQ ;o!p#ter Networ&s vol. 54 no. 5 pp. E7/ EG7 /616. A34B . ,avie and I. 'ross P2 +tateless Transport T#nnelin$ )rotocol for Networ& Jirt#ali:ation >+TT?Q nternet 0n$ineerin$ Tas& Force 2pril /614. ACnlineB. 2vailableD httpDtools.ietf.or$ht!ldraft-davie-stt-67 A35B (. (ahalin$a! ,. ,#tt H. ,#da ). 2$arwal *. Hree$er T. +ridhar (. #rsell and ;. ri$ht PJ*2ND 2 Fra!ewor& for Cverlayin$ Jirt#ali:ed *ayer / Networ&s over *ayer 3 Networ&sQ nternet ,raft nternet 0n$ineerin$ Tas& Force Nove!ber /613. ACnlineB. 2vailableD httpDwww.ietf.or$iddraft-!ahalin$a!-d#tt-dcops-vxlan-67.txt A37B (. +ridharan 2. 'reenber$ O. an$ ). 'ar$ N. Jen&atara!iah H. ,#da . 'an$a '. *in (. )earson ). Thaler and
<>
