part iv: carriers, traffic mgt, and trends
DESCRIPTION
Part IV: Carriers, Traffic Mgt, and Trends. Carrier Technologies SDH/SONET WDM xDSL Traffic Management Definitions and Traffic Models ATM Services Trends. Synchronous Digital Hierarchy – SDH. SONET: Synchronous Optical Network. ANSI-SONET (U.S.A.) and ETSI-SONET (Europe). - PowerPoint PPT PresentationTRANSCRIPT
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 1 ETH Zürich
Part IV: Carriers, Traffic Mgt, and Trends
• Carrier Technologies– SDH/SONET– WDM– xDSL
• Traffic Management– Definitions and Traffic Models– ATM Services
• Trends
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 2 ETH Zürich
SONET: Synchronous Optical Network. ANSI-SONET (U.S.A.) and ETSI-SONET (Europe). SDH: Synchronous Digital Hierarchy (international):
• Synchronous frames: 125 s
• Integration of ATM-and STM-baseddata.
• Compatibility withexisting equipmentand signaling.
• Support of varioustransmission rates.
STS-1STS-3STS-9STS-12STS-24STS-36STS-48STS-192
STM-1STM-3STM-4STM-8STM12STM-16STM-64
51,84155,52466,56622,08
1244,161866,242488,32
9953.28
50,12150,336451,008601,344
1202,6881804,0322405,376
9621.504
SONET SDHTransmissionRate in Mbit/s
User DataRate in Mbit/s
Synchronous Digital Hierarchy – SDH
STS-x: Synchronous Transfer Signal level xSTM-y: Synchronous Transport Modul level y
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 3 ETH Zürich
SONET – Architecture
Section: Fiber-optical cable between sender/receiver. Line: Sequence of sections.
• Unchanged internal signal and channel structure.
Path: Interconnection of two devices.
Repeater(STE)
SONETMultiplexer
(PTE + LTE)
Add-DropMultiplexer
(LTE)
SONETMultiplexer
(PTE + LTE)
Repeater(STE)
Terminals Terminals
Section Section Section Section
Line Line
Path
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 4 ETH Zürich
SONET – Frame (STS-1)
The frame length is 125 s. Rows and columns are used. Transmission from left to right by rows. Frames contain user data and additional control
data as well as timing information.
Sectionoverhead(3 rows)
Lineoverhead(6 rows)
Transportoverhead
(3 columns)
STS - Frame:
(3+6) * (3+87) Octett
810 Octett
Brutto data:
810 Octett / 125 s
51,84 Mbit/s
User data:
810 - [3*(3+6) + 1*(3+6)] Octett
49,536 Mbit/s
Synchronous PayloadEnvironment
(87-1 columns)
Path overhead (1 column) 125 s
0 s
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 5 ETH Zürich
SONET – Frame (STS-N)
Basic frame: STS-1 with 810 octets. Higher rate SONET channels formed by
octet-interleaving of multiple STS-1 inputs:• STS-N rate is formed from N STS-1 inputs.• Advantage: STS-1 line cards remain operable in an
STS-1-to-STS-N multiplexor.• STS-N frame:
90 * N columns per row, including 4 * N columns of interface overhead.
• Example: STS-3 = STM-1 (155.52 Mbit/s)
Sectionoverhead(3 rows)
Lineoverhead(6 rows)
Transportoverhead
(3*3 columns)
Synchronous PayloadEnvironment
((87-1) * 3 columns)
Path overhead (1*3 columns) 125 s
0 s
Payload150.336 Mbit/s
Overhead5.184 Mbit/s
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 6 ETH Zürich
SONET – Localization of Payload
Pointer H1 and H2 contain values for number of payload bytes inbetween H3 and J1. Direct access to single channels. No (de-)multiplexing necessary.
Payload may be located in two STS-1 frames.
FrameN
(9 rows)
FrameN+1
(9 rows)Path overhead
(1 column)
H1H2
H1H2
(9 rows)
H3 is used as padding byte.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 7 ETH Zürich
SDH Network Topologies
Point-to-PointTerminal
Point-to-PointTerminal
Point-to-PointTerminal
Point-to-PointTerminal
Point-to-point configurationwith 1:4 protection channel sharing
Point-to-PointTerminal
Point-to-PointTerminal
Point-to-PointTerminal
Point-to-PointTerminal
Linear Add/Drop Route
Add/DropMultiplexer
Add/DropMultiplexer
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 8 ETH Zürich
Fiber Optic Networks Revisited
Traditional use of fibers:
Optical Fiber
Laser Receiver
Current transmission capacities:• 2.5 Gbit/s (OC-48)
• 10 Gbit/s (OC-192)
Lasers available for 850 nm, 1310 nm and 1550 nm wavelength.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 9 ETH Zürich
Wavelength-Division Multiplexing
Dense* Wavelength-Division Multiplexing (DWDM):
Optical Fiber
Array of LasersArray of Photodetectors
Current available transmission capacities:• 96 lasers at 2.5 Gbit/s = 240 Gbit/s (OC-4608)
• 32 lasers at 10 Gbit/s = 320 Gbit/s (OC-6144)
• Soon 128 lasers at 10 Gbit/s > 1 Tbit/s (=1.000.000.000.000 bits/s)
* „Dense“ WDM: More than 10 lasers used simultaneously. Today: WDM usually means dense WDM.
1 2 3 4
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 10 ETH Zürich
Breaking the Internet Gridlock
Utilizing publically available infrastructure:• How to serve private users with sufficient bandwidth?• How to interconnect two enterprise sites with an at least
medium bandwidth solution?
Solution possibilities:• Hybrid fiber/coax (HFC) technology: any configuration of
fiber-optic and coaxial cable that is used to distribute local broadband communications:
– Shared downstream bandwidth, up to 30 Mbit/s.• Wireless cable.• xDSL (Digital Subscriber Lines).
Deployment: 650 M customers on twisted pair.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 11 ETH Zürich
ADSL Technology – Overview (1)
Twisted pair access to the information highway:• Delivering video und multimedia data.• Avoids the replacement of existing cabling.• Transformation of existing telephone network into
a multi-service network by applying modulation.• Use of full copper frequency spectrum (app. 1.1 MHz).
ADSLModem
ADSLModem
ExistingCopperCore Network
Internet
Server 16 … 640 kbit/s
1.5 … 9 Mbit/s *) depending on the implementationarchitecture
*)
144 kbit/s (POTS)
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 12 ETH Zürich
ADSL Technology – Overview (2)
ADSL Forum Reference Model:
ATU-C
ATU-C
ATU-C
ATU-RATM-SM
Splitter Splitter
Vc Va UC-2 U-C U-R U-R2 T-SM T-P T
POTS-C POTS-R
AccessNode
PSTN PhonesetsPremises
Distribution Network
T.E.
DigitalBroadcast
BroadbandNetwork
NarrowbandNetwork
NetworkManagement
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 13 ETH Zürich
DSL Comparison
1441,000
160 – 1,1682,048
1,500 – 8,0001,500 – 25,000
Downstream[kbit/s]
DSLScheme
IDSLUDSLSDSLHDSLADSLVDSL
Upstream[kbit/s]
VoiceSupport
144300
160 – 1,1682,048
64 – 8001,600
ActiveSplitterless
NoNo
PassivePassive
ADSL: Asymmetric DSlHDSL: High bit-rate DSLIDSL: ISDN DSLSDSL: Symmetric DSLUDSL: Universal DSLVDSL: Very high bit-rate DSL
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 14 ETH Zürich
ADSL Technology – Capabilities
Data rates depend on:• Length of copper line,• Wire gauge,• Presence of bridged taps, and• Cross-coupled interference.
95% of todays loop plantsmeet these measures.
Requires advanced digitalsignal processing and advanced coding schemes to deal with varying noise figures.
Data Rate[Mbit/s]
1.5 or 21.5 or 2
6.16.1
Wire Gauge[mm]
0.5 (26 AWG)0.4 (24 AWG)0.5 (26 AWG)0.4 (24 AWG)
Distance[km]
5.54.63.72.7
C Channels 1664
Optional 160 Channels 384
544576
DuplexBearer Channels
[kbit/s]
DownstreamBearer Channels
[Mbit/s]
n*1.536 1.5363.0724.6086.144
n*2.048 2.0484.096
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 15 ETH Zürich
ADSL Technology – DMT Modulation
To work simultaneously withPOTS on copper line.• Lower 4 kHz are used
by POTS.• Discrete Multi Tone (DMT):
256 separate sub-frequenciesfrom 64 kHz.
Amplifica-tion varies dependenton frequency.
Data rate = No of channels * no of bits/channel * modulation rateTheoretical max upstream: 25*25*4k = 1.5 Mbit/sTheoretical max downstream: 249*15*4k = 14.9 Mbit/s
Discrete Multitone (DMT) Modulation
POTS each 4 kHz (32 QAM) 1.4 MHz
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 16 ETH Zürich
ADSL Network Architectures (1)
ADSL-ATM network architecture, point-to-point:
DSLAM: Digital Subscriber Line Access Multiplexor
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 17 ETH Zürich
ADSL Network Architectures (2)
ADSL-ATM including L2TP:
LAC: Local Access CarrierLAC: Local Access Carrier
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 18 ETH Zürich
Part IV: Carriers, Traffic Mgt, and Trends
• Carrier Technologies– SDH/SONET– WDM– xDSL
• Traffic Management– Definitions and Traffic Models– ATM Services
• Trends
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 19 ETH Zürich
Traffic Engineering Definition
Traffic Engineering is the task of mapping traffic flows onto an existing physical topology.
The goals of traffic engineering are:• Minimization of packet loss and packet delay.• Optimization of network resources (avoiding overload
situations through load balancing).
Traffic engineering “applications” allow for a precise control of how traffic flows are placed within a routing domain.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 20 ETH Zürich
Policies and Mechanisms
Traffic engineering consists of:• Traffic management (short-term) and• Network planning (long-term).
Traffic management:• Set of policies and mechanisms for satisfying a range of
diverse application service requests.• Acting across: diversity and efficiency.• Subsumes traditional ideas of congestion control:
– An overloaded resource suffers from service degradation.
– Policies scale back demand or restrict access.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 21 ETH Zürich
Traffic Models
Goal of effectively managing traffic requires:• Requirements of individual applications and organizations.• Their typical „behavior“.
Traffic Models:• Summarize „expected behavior“.• Obtained by detailed traffic measurements or amenable to
mathematical analysis.• State of the art in traffic modelling:
– Telephone traffic model and– Internet traffic model.
• Change of applications make these models to change!
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 22 ETH Zürich
Telephone Traffic Model
Call arrival model:• How are calls placed?• Interarrival times drawn from an exponential distribution
(poisson process models all arrivals).• Memoryless (certain time elapse does not tell the future).
Call holding-time model:• Call holding-times drawn from an exponential distribution:
– Call longer than x decreases exponentially with x.• Heavy-tailed distribution in recent studies:
t
P(T>t)
0.010.00010.000001
10 20 30
Exponential
Heavy-tailed
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 23 ETH Zürich
Internet Traffic Model
Parameters to characterize applications:• Distributions of interarrival times between app. invocations.• Duration of a connection.• Number of bytes transferred during a connection.• Interarrival times of packets within a connection.
Note: There is little consensus on models!• E.g., interarrival times: Exponential or Weibull.• Effective means: Measurements to fit to statistical model.• LAN traffic differs heavily from WAN traffic.
– More local bandwidth, tendency for longer holding times, higher peak data rates.
– Expensive wide area bandwidth, less volume.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 24 ETH Zürich
Time Scales of Traffic Management
Scheduling, buffer managementRegulation, policingRouting (connection less)Error detection and correctionFeedback flow-controlRetransmissionRenegotiationSignallingAdmission-controlService pricingRouting (connection-oriented)Peak-load pricingCapacity Planning
Less than oneRTT(Cell level)
One or moreRTTs(Burst level)Session(Call level)
DayWeeks and more
Time Scale Mechanism Net Endsystem
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 25 ETH Zürich
Service Categories
ATM offers six service categories:• Real-time services using resource reservation.• Non-real-time services without resource reservation.• Non-real-time services with partial resource reservation.
Sources have to comply to a previously negotiated traffic characteristic (traffic contract).
Conforming traffic is transported with the negotiated quality of service guarantees.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 26 ETH Zürich
Real-Time Services (1)
CBR (Constant Bit Rate):• Traffic: constant, Peak Cell Rate (PCR).• QoS parameter: max. Cell Transfer Delay (maxCTD),
Cell Delay Variation (CDV), Cell Loss Ratio (CLR).• Example: uncompressed video/audio data.
Peak Cell Rate defines a temporal distance: T = 1/PCR.
Cells have to be evenly spaced in time.
T T T
marked or dropped
T
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 27 ETH Zürich
Real-Time Services (2)
rt-VBR (Real-Time Variable Bit Rate):• Traffic: Peak Cell Rate (PCR), Sustainable Cell Rate
(SCR), Maximum Burst Size (MBS).• QoS parameter: maxCTD, CDV, CLR.• Example: compressed video / audio data
marked or dropped
T = 1/PCRTS = 1/SCR
t T with mean value TS
t t t t
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 28 ETH Zürich
Non-Real-Time Services (2)
ABR (Available Bit Rate):• Traffic: Peak Cell Rate (PCR) and Minimum Cell Rate
(MCR), flow control mechanism mandatory.• "QoS parameter": minimum cell loss.• Flow control mechanism determines the
Allowed Cell Rate (ACR).
reserved
dynamic
Link Rate
PCR (Peak Cell Rate)
MCR (Minimum Cell Rate),may be 0.
ACR(Allowed Cell Rate)
Dynamically changedby the flow control.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 29 ETH Zürich
Usage Parameter Control
Test, whether a cell stream conforms to a given traffic characteristics.
Generic Cell Rate Algorithm: GCRA(T, ).• Virtual Scheduling Algorithm or• Continuous-State Leaky Bucket.
Input parameters: T = 1/PCR, = CDVT.
T
T
T
OKOK Not OKOK OK
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 30 ETH Zürich
Peak Cell Rate Conformance
For CBR traffic, it is sufficient to test peak cell rate. Usage Parameter Control takes places at the
network interfaces.
Shaping
Physical PrivateATM
PublicATM
UPCUPC
PrivateUNI
PublicUNI
GCRA(T,0)
GCRA(T, *)GCRA(T, )
Sha-ping
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 31 ETH Zürich
Part IV: Carriers, Traffic Mgt, and Trends
• Carrier Technologies– SDH/SONET– WDM– xDSL
• Traffic Management– Definitions and Traffic Models– ATM Services– IP Services
• Trends
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 32 ETH Zürich
Use of Network Protocols
0
10
20
30
40
50
60
70
80
1994 1996 1998 2000 2002
IP
SNA
IPX
RFC 1490
Others
IP is the only protocol that matters anymore!
Source: Gartner Group
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 33 ETH Zürich
Data Traffic is Overtaking Voice
Voice
Data
Time
Volume
Source: CIENA Corp.
POTS
Data
DATA
POTS
Voice-Centric Data-Centric
Today
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 34 ETH Zürich
Effect on (Carrier) Networks
Everything will be data, soon. The only protocol that matters is IP. Networks have to accomodate for the exponential
traffic growth. It makes sense to design networks for IP only!
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 35 ETH Zürich
Technology Trends
Chip performance doubles every 18 months (Moore‘s Law).
Modern chips can switch packets as fast as ATM cells.
New router architectures have appeared:• Routing at Gigabit/s speed• Routers support traffic management with thousands of
queues per interface• Routers interface directly to DWDM
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 36 ETH Zürich
DWDMDWDM
Layer upon Layer...
Sonet
ATM
IP
DWDM
ATM
IP
Sonet
IP
DWDM
IP
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 37 ETH Zürich
Traffic Multiplexing in the Backbone
SonetADM
OC-48OC-48
DWDM Ring
OC-48
DWDMADM
DWDM
Sonet
ATM
IP
N x OC-48
Multiplexing of IP traffic over ATM or Sonet no longer required.
Segmentation of IP packets into ATM cells not possible at OC-48.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 38 ETH Zürich
Optical Internet Backbones (1)
OC-48DWDM
IPDWDMADM
DWDM Ring
N x OC-48
Most important objective: high bandwidth. No „Quality of Service“, but „Classes of Service“ IP-centric Control (no SONET, no ATM). Traffic engineering using MPLS.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 39 ETH Zürich
Optical Internet Backbones (2)
Optical network: Provides point-to-point connectivity between routers („lightpaths“).
„Lightpaths“ have fixed bandwidth (e.g. OC-48). „Lightpaths“ define virtual topology, which may be static
by design.
Router network Optical network
Optical Crossconnects
IP routers
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 40 ETH Zürich
Conclusions
Transporting data using IP will be the key task of the „New Public Network“.
IP over ATM can not keep up with the very high-speed backbones (SAR!).
IP over DWDM or IP over Sonet needs to solve the traffic engineering problem.
IP over ATM will remain for small ISPs or large enterprise networks due to its proven reliability and traffic management capabilities.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 41 ETH Zürich
References (1)
• M.-C. Chow: Understanding SONET/SDH; 1995, Andan Publisher, Holmdel, New Jersey, U.S.A.,ISBN 0–9650448–2–3.
• The Sonet Home Page; URL: http://www.sonet.com, 1999.
• CIENA Inc.: Fundamentals of DWDM; URL: http://www.ciena.com, 1999.
• D. Ginsburg: Implementing ADSL; Addison-Wesley, Reading, Massachusetts, U.S.A., July, 1999,ISBN 0-201-65760-0.
• M. de Prycker: “Asynchronous Transfer Mode – Solution for Broadband ISDN”, 3rd Edition, 1995, Prentice Hall, Englewood Cliffs, New Jersey, U.S.A., ISBN 0–13–342171–6.
© 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM IV – 42 ETH Zürich
References (2)
• X. Xiao, L. M. Ni: Internet QoS: A Big Picture; IEEE Network Magazine, Vol. 13, March/April 1999, pp 8 – 18.
• C. Schmidt, M. Zitterbart: Reservierung von Netzwerkres-sourcen – Ein Überblick über Protokolle und Mechanismen; Praxis der Informationsverarbeitung und Kommunikation, Vol. 18, No. 3, 1995, pp 140 – 147.
• L. Zhang, S. Deering, D. Estrin, S. Shenker, D. Zappala: RSVP: A New Resource ReSerVation Protocol; IEEE Network, Vol. 7, No. 5, September 1993, pp 8 – 18.
• The SWITCHlan backbone network; available at the URL: http://www.switch.ch/lan, 1999.
• C. Metz: IP Routers: New Tool for Gigabit Networking; IEEE Internet Computing; November/December 1998, pp. 14-18.