atm technology atm service categories by …pages.cs.aueb.gr/courses/networks/notes2016/... ·...

54
ATM Technology ATM Service Categories By C.Courcoubetis, V.Siris, G.D.Stamoulis

Upload: truongkhanh

Post on 03-Apr-2018

220 views

Category:

Documents


1 download

TRANSCRIPT

ATM TechnologyATM Service Categories

By C.Courcoubetis, V.Siris, G.D.Stamoulis

Network technology - 2

A Taxonomy of Communication Networks

Communication networks can be classified based on the way in which the nodes exchange information:

Communication Network

Switched Communication

Network

Broadcast Communication

Network

Circuit-Switched Communication

Network

Packet-Switched Communication

Network

Datagram Network

Virtual Circuit Network

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 3

Asynchronous Transfer Mode (ATM)The technology supported by the telecom operators as the solution for Broadband ISDN

Unified solution for LAN, MAN, WAN

Was fully standardized by the end of 90s, but …Internet was already very widely spread

Now mainly serves for the provision of backbone connections

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 4

Key Features of ATMAsynchronous: non-periodic transfer of information between ports in switches

Short fixed size cells

frame

STM (Synchronous Transfer Mode):

ATM:

...

...

header(VPI/VCI)

data

5 bytes 48 bytes

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 5

Key Features of ATM (cont.)Connection oriented: Virtual Channel (VC) connection is setup prior to information transfer

Labels vs. addresses: labels have local significance scalability

...

7 2

3 7

Network technology - 6

Switching in Packet NetworksDatagram (e.g. Internet): packet header contains full address information

Address has global significance

Virtual circuit (e.g. ATM, Frame Relay): packet header contains virtual circuit identifier

Address has local significance

Destination address: 147.52.16.2

VC: 16

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 7

Cell Switching

25 25 25

34

30

ATM switchCell

Header(VPI/VCI)

Payload

30 49

input port 1

output port1

output port 2

input port 2

Input Port VPI/VCI Output Port VPI/VCI1 25 2 341 30 1 49

Internal Routing Table

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 8

Virtual Channel Connection

30

49

17

7 Virtual Channel Link

Virtual Channel Connection

Two types of Virtual Channels:Switched Virtual Channels (SVC): connection setup with signallingPermanent Virtual Channels (PVC): connection setup with management

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 9

Routing and Call Admission Control (CAC)

Xsource

destination

Routing: find path from source to destination that fulfils user requirements (bandwidth, QoS)Call Admission Control (CAC): performed at every switch, determines whether there are enough resources to accept a call

CAC CAC

CAC

CAC

CAC

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 10

Virtual Channels and Virtual PathsCell label:

First 8 (12) bits: Virtual Path IdentifierLast 16 bits: Virtual Circuit Identifier

Virtual Path: group of VCs treated similarly

TRANSMISSIONPATH

VP

VPVP

VP VC

VCVC

VC

Network technology - 11

ATM Cell Header Format

GFC = Generic Flow ControlVPI = Virtual Path IdentifierVCI = Virtual Channel IdentifierPTI = Payload Type IdentifierCLP = Cell Loss PriorityHEC = Header Error Control

GFC/VPI VPI

VPI VCI

VCI

VCI PTI CLP

HEC

1

2

3

4

5

byte

GFC: only in UNI CLP=1 => cell has low priorityPTI: identifies user cell/data cell, congestion control

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 12

Input Port VPI Output Port VPI1 2 2 51 9 1 102 4 1 9

2/1 5/1

2/7 5/7

9/3 10/3

4/2 9/2

VC 1VC 7

VC 1VC 7

VP 2

VP 5

VP 9

VP 4

VP 10

VP 9

input port 1

input port 2

output port 1

output port 2VC 2

VC 2

VC 3

VC 3

VP Switching

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 13

VP and VP/VC Switching in a Network

source destination

26 / 43 33 / 57 63 / 127 47 / 52 39 / 41 45 / 3856 / 29

destinationsource

VPI/VCI

26 / 43 33 / 29 63 / 29 47 / 52 39 / 52 45 / 5256 / 29VPI/VCI

VP/VC switchVP switchor cross-connect

VPI: 56 VPI: 63 VPI: 47 VPI: 45

26/43

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 14

B-ISDN Interface Terminology

ATM switchATM station

User-Network Interface (UNI)

Network-Node Interface (NNI)

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 15

B-ISDN Interface Terminology (cont.)

B-ICI(B-ISDN Inter-exchangeCarrier Interface)

Public ATM

: Private ATM switch : Public ATM switch

Private ATM

Private UNI Private NNIPublic UNI

Public ATM

NNI

Network technology - 16

ATM Layer Service Categories

Offered at ATM UNI (User-Network Interface)

ATM UNI

ATMNetwork

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 17

Service Category TypicalApplication

Constant Bit Rate (CBR) Circuit emulation,videoconferencing,

Real-Time Variable Bit Rate(rt-VBR)

Compressedvideo/audio

Non-Real-Time Variable Bit Rate(nrt-VBR)

Critical data

Available Bit Rate (ABR) LANinterconnection,

Unspecified Bit Rate (UBR) File transfer,message transfer

ATM Forum Service Categories

Real-Time :

Non-Real-Time:

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 18

ATM Forum Real-Time Service CategoriesConstant Bit Rate (CBR):

real-time applications requiring a static amount of bandwidthGuaranteed Quality of Service (QoS) in terms of delay, delay variation, cell loss

Real-Time Variable Bit Rate (rt-VBR):real time applications with “bursty” trafficGuaranteed Quality of Service (QoS) in terms of delay, delay variation, cell loss

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 19

ATM Forum Non-Real-Time Service CategoriesNon-Real-Time Variable Bit Rate (nrt-VBR):

non-real-time applications with bursty trafficGuaranteed cell loss bound but no delay bounds

Available Bit Rate (ABR):“elastic” applications which can adapt their traffic rateclosed loop flow control supported

Unspecified Bit Rate (UBR):non-real-time applications, no service guarantees

Network technology - 20

Bandwidth Usage

Bandwidth unallocated

LinkBandwidth

time

Bandwidth allocatedand used

Bandwidth allocatedbut not used

CBR + VBR traffic

ABR + UBR traffic

LinkBandwidth

Network technology - 21

Traffic Contract Negotiation

ATM UNI

ATMNetwork

QoS parameters: cell delay, cell delay variation, cell loss ratio

Traffic parameters: peak rate, sustainable cell rate, burst size, minimum cell rate

Traffic Contract:

Conformance definition: Generic Cell Rate Algorithm (GCRA)

Network technology - 22

Traffic Parameters

Peak Cell Rate (PCR): Maximum instantaneous rateSustained Cell Rate (SCR): Average cell rate measured over some long intervalMaximum Burst Size (MBS): Maximum burst size (# of cells) that can be sent at the peak rateMinimum Cell Rate (MCR): Minimum cell rate user is guaranteed to have

Traffic parameters describe inherent characteristics of a traffic source

Network technology - 23

Traffic parameters (cont.)

time

Peak Cell Rate (PCR):

1/PCR

time

Peak Cell Rate (PCR), Sustainable Cell Rate (SCR), Maximum Burst Size (MBS)

...

...

Ti (minimum burst interarrival time)=MBS/SCR

cell

Tb (maximum burst durationat PCR)=(MBS-1)/PCR

SCR is also met with more “sparse” cells

...

Network technology - 24

ATM Quality of Service (QoS) Parameters

Max Cell Transfer Delay (CTD): Delay experienced by a cell between network entrance and exit pointspeak-to-peak Cell Delay Variation (CDV): max - min cell transfer delayCell Loss Ratio (CLR): Percentage of cells that are lostThe above are user negotiable

ATM QoS parameters characterise performance of an ATM layer connection

Network technology - 25

ATM QoS parameters (cont.)

fixed delay (min) peak-to-peak CDV

area ≤

CLR

max CTD

probabilitydensity

cell transferdelay

Network technology - 26

Conformance Definition

GCRA: Generic Cell Rate AlgorithmAbove applies to CBR, VBR, and UBRABR is a special case (includes feedback)

Conformance definition defines conformity at an interface with respect to traffic contract according to one or more instances of GCRA

Network technology - 27

Service Category Attributes and Guarantees

ServiceCategory

TrafficDescription

Guarantees FeedbackControl

Min Loss(CLR)

Delay/Variance

Bandwidth

CBR PCR

rt-VBR PCR,SCR,MBS

nrt-VBR PCR,SCR,MBS

ABR PCR,MCR+behavior

UBR (PCR) NO NO

NO

NO

NO

NO

NO

NO

NO

Network technology - 28

Traffic Control in ATMConnection Acceptance Control (CAC)RoutingUsage Parameter Control (UPC) or Source PolicingShapingPriorities, schedulingFeedback controlOther (e.g. pricing)

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 29

Connection interarrival time

Time scales of network control

Selective cell discard, frame discard,priority control and scheduling, Usage

Parameter Control (UPC), traffic shaping

Feedback controls

Routing, Call Admission Control (CAC)

Network management control

Cell time

Round-trip propogation time

Traffic & Congestion Control Functions Response time

order of minutes

Pricing months, years

Network technology - 30

Network control and pricingSet of feasible services depends on network control mechanismsEconomic incentives influence network control mechanismsNetwork control: guarantees contractsPricing: influences demand in order to increase efficiency

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 31

Network Control for Congestion

Guaranteed servicesCall Acceptance Control - CAC user-network contractUse charging as incentive for users to provide more accurate information about traffic profile

Best effort services (ABR):no CAC, except for MCR (Minimum Cell Rate)Flow controlUse charging to control congestion

acceptable

n1

n2

CAC:

(n1,n2: number of multiplexed sources of type 1,2)

non-acceptable

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 32

Prototype Problem of Call Acceptance (for Guaranteed QoS)

k traffic classesclass i contains

sources ni

n1

nkB: buffer

C: capacity... ......

What do not violate QoS constraints ?

StatisicalStatisical QoSQoS constraint: e.g. CLP less than p (e.g. p=10-8 )

( , , )n nk1 K

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 33

Prototype Problem of Call Acceptance (cont.)

Peak rate allocation: sufficient conditionacceptance region:

accomplishes zero cell losssimple acceptance criterion (linear in )dedicated capacity per source (reminds circuit switching)conservative: assumes worst-case demand arises simultaneouslyno multiplexing gain less revenue, more expensive to customers

n h n h Ck k1 1⋅ + ⋅ <K

n1

nkB: buffer

C: capacity... ......traffic

ni

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 34

Prototype Problem of Acceptance Control (cont.)

Mean rate allocation: only necessary conditionacceptance region:

ultimate multiplexing gainsimple (linear in )no performance guarantees (only CLP<1 !)

n m n m Ck k1 1⋅ + ⋅ <K

ni

n1

nkB: buffer

C: capacity... ......traffic

Network technology - 35

Effective Bandwidth and Many Sources Asymptotic

If , C=Nc and B=Nb, then

and thus

is the effective bandwidth of sources of type j,

Xj[0,t] = number of cells in [0,t]s = space scale parametert = time scale parameter

lim log Pr[ ] sup inf ( , ) ( )N t s j

k

jNI st a s t s ct b

→ ∞= − = − +

⎧⎨⎩

⎫⎬⎭

∑11

overflow ρ

a s tst

EejsX tj( , ) log [ , ]=

1 0

Pr[ ]overflow ≈ =− −e eNI γ

a s tj ( , )

n Ni i= ρ

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 36

Simplifying the Problem of Acceptance Control

Use Effective Bandwidths acceptance region:

depends only on source type, traffic mix, capacity, buffer and QoS requirements, summarises each source’s effect on QoSimmediate guarantee of QoSlinearizes a very complex problem of QoSstill allows easy estimation of available capacity

n eb n eb Ck k1 1⋅ + ⋅ <K $

ebi

iii hebm ≤≤

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 37

ON/OFF Source Model

h

ON/OFF source: ON & OFF periods are exponentially distributed

on

off

ab

timeon off

1/h

mean ON period = 1/a, mean OFF period = 1/b

mean rate m = b*h/(a+b)

effective bandwidth ( )ebs

mh

eish= + −

⎡⎣⎢

⎤⎦⎥

11 1log

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 38

CAC: Peak Rate vs. Effective Bandwidth

Effective Bandwidth CAC constraints correspond to tangents of the acceptance region boundary curve

The right selection of a tangent depends on the operating point

acceptable

n1

n2

non-acceptable

peak-rate constraint

(n1,n2: number of multiplexed sources of type 1,2)

Effective-bandwidth constraints

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 39

Usage Parameter Control (UPC) or Source Policing

UPCATM

NetworkATM UNI

Traffic Contract

UPC actions:

UPC

CLP bit

Cell passing:

Cell tagging:

Cell discarding:

UPC

UPC

0

0

0

0

1

Network technology - 40

UPC and Traffic Shaping

Traffic shaping at source prevents loss due to policing

UPCATM UNI

Shaping

Traffic Contract

ATMNetwork

nonconforming cellsMBS

No shaping

MBS

With shaping

MBS: maximum burst size

time time

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 41

UPC (or Source Policing) FunctionsAlgorithm implemented at UNI for ensuring that user traffic adheres to traffic contract.Generic Cell Rate Algorithm (GCRA):

“Leaky bucket” type algorithmOpen loop algorithmSpecified by ATM Forum

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 42

Generic Cell Rate AlgorithmGCRA(T, τ)t: cell arrival timetat: theoretical arrival time

tat0

cell nonconforming cell conforming

Ttat

if (t<tat- τ)cell nonconforming

elsecell conformingtat=max(t,tat)+T

“Virtual Scheduling Algorithm”

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 43

GCRA examplesGCRA(10,2)

Cell arrival times: 0 10 18 28 42 49

Cell arrival times: 0 10 15 25 35

t 0 10 15 25 35tat-τ 0-2 10-2 20-2 20-2 35-2result C C NC C Ctat after 10 20 20 35 45

if (t<tat- τ)cell nonconforming

elsecell conformingtat=max(t,tat)+T

t 0 10 18 28 42 49 tat-τ 0-2 10-2 20-2 30-2 40-2 52-2result C C C C C NC tat after 10 20 30 40 52 52

0 10 20 30 40 50

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 44

Leaky Bucket AlgorithmBucket size=1+τ/TLeak rate=1/T cells/sBucket contents increased by 1 for each conforming cell

Is provably equivalentto GCRA

B: bucket contents

If B+1> 1+τ/Tcell nonconforming

elsecell conformingB=B+1

virtual cell1+τ/T cells

1/T cells/s

actual cell

conforming ornonconforming

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 45

Traffic Contract for CBRFor each CBR connection:

PCR (Peak Cell Rate)CDVT (Cell Delay Variation Tolerance): takes care of slotted nature of ATM, physical overhead, ATM layer function overhead

– E.g. only with CDVT we can have PCR=2/3*R, where R is ATM layer rate

Connection must conform to GCRA(R/PCR,CDVT), PCR=15 Mbps, CDVT=2, ATM Layer Rate=150 Mbps =>

GCRA(10,2)

Customer Premises Network (CPN) UPC

Public ATM network

Private UNI Public UNI

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 46

Traffic Contract for VBR For each VBR connection:

PCR, CDVTSCR, MBS, CDVTBT=(MBS-1)(R/SCR-R/PCR), where R: ATM Layer Rate

Connection must conform to GCRA(R/PCR,CDVT) ANDGCRA(R/SCR, BT+CDVT)

Equivalent description: Dual leaky bucket

© C.Courcoubetis, V.Siris, G.D.StamoulisNetwork technology - 47

Leaky buckets for SCR and PCR policing

virtual cell1+(BT+CDVT) x SCR cells

SCR cells/s

arriving cell1+CDVT x PCR cells

PCR cells/s

Network technology - 48

Deterministic multiplexing of leaky bucket streams

Assume leaky bucket (r,b) and access rate h

Maximum time t user can transmit at rate h

Deterministic QoS constraints: no losses, delay < D for all cells

rhbtbrtht−

=⇒+≤

NB: buffer

C: capacity...

h: access rate

Network technology - 49

Multiplexing leaky bucket streams (cont.)Maximum backlog Q if N users (r,b) multiplexed on link with capacity C

Delay constraint D

Must also have:

Finally: CAC criterion , where is the effective bandwidth

rhbCNhCtNhtQ−

−=−= )(

CDrhb

bhNDCrh

bCNhDCQ

≤⎟⎟⎠

⎞⎜⎜⎝

⎛−+

⇒≤−

−⇒≤)(

)(

CNr ≤

CNa ≤

⎭⎬⎫

⎩⎨⎧

−+= r

Drhbbha ,

)(max

a

Network technology - 50

Available Bit Rate (ABR) Services within ATMIntended for elastic sources (i.e., sources which can increase/decrease their traffic rate)For each ABR connection:

PCR (Peak Cell Rate)MCR (Minimum Cell Rate)

No specific QoS parametersCLR (Cell Loss Ratio) is expected to be low for compliant sourcesfair share of available bandwidth

Rate-based closed loop congestion controlbinary feedback (Explicit Forward Congestion Indication - EFCI)rate-based (Explicit Rate - ER), Resource Management cell

Network technology - 51

ABR Congestion Control

a) EFCI marking mode

b) Relative Rate marking mode

c) Explicit Rate marking mode

source destinationCI bit setdata cell

RM cell

RM cell

CI bit set

source destination

RM cell

RM cell

ER set

source destination

Network technology - 52

Switch Operation with ABR Congestion ControlEach switch computes Explicit Rate (ER)

e.g., ER=C/N, where C is link capacity and N number of ABR connections

Sets minimum of value in ER field (in RM cell) and computed ER

RM cell

RM cell

ER

source destination

ER=min(ER_cell, ER_computed)

Network technology - 53

ABR Congestion Control: Source Behavior Same algorithm for any network (internal) modesIf no congestion ⇒ can increase sending rate (additive)If congestion ⇒ decrease sending rate (multiplicative)At all times: Sending Rate < Explicit Rate (ER)

Actually a bit more complicated (NI: No Increase, ACR: Allowed Cell Rate)If CI=1

Reduce ACR by amount proportional to current ACR but not less than MCR

Else if NI=0Increase ACR by amount proportional to PCR but not more than PCR

If ACR>ER set ACR=max{ER,MCR}

Network technology - 54

ABR versus TCPABR feedback controls rate of transmission

Rate control

TCP feedback controls window sizeWindow control

ABR feedback explicit from switches or destinationECN (Explicit Congestion Notification)

TCP feedback implicit (losses)ATM switch must perform:

Congestion control: Monitor queue lengthFair capacity allocation: Throttle back connections using more than fair share

Active Queue Management schemes similar to aboveRED (Random Early Detection)