Upload
demetrius-price
View
50
Download
11
Tags:
Embed Size (px)
DESCRIPTION
TCOM 513 Optical Communications Networks. Spring, 2007 Thomas B. Fowler, Sc.D. Senior Principal Engineer Mitretek Systems. Topics for TCOM 513. Week 1: Wave Division Multiplexing Week 2: Opto-electronic networks Week 3: Fiber optic system design Week 4: MPLS and Quality of Service - PowerPoint PPT Presentation
Citation preview
TCOM 513Optical Communications
Networks
Spring, 2007
Thomas B. Fowler, Sc.D.
Senior Principal Engineer
Mitretek Systems
2ControlNumber
Topics for TCOM 513
Week 1: Wave Division Multiplexing Week 2: Opto-electronic networks Week 3: Fiber optic system design Week 4: MPLS and Quality of Service Week 5: Optical control planes Week 6: The business of optical networking: economics
and finance Week 7: Future directions in optical networking
3ControlNumber
Virtual Session
End-to-End Messages
Physical
Presentation Presentation
Session Session
Network Network
Data Link Control
Data Link Control
PhysicalPhysical
Physical Link, e.g. electrical signals
Physical portion of code
Logical portion of
code
Virtual Network ServiceApplicationApplication
End-to-End PacketsTransport Transport
DLC DLC DLC DLC
NetworkNetwork
Bits
Packets
Frames
Physical Physical Physical
Originating site
Terminating site
Subnet node
Subnet node
4ControlNumber
Opto-electronic systems and networks
LAN protocols– Fiber distributed data interface (FDDI)– Fiber channel– Gigabit/10 Gigabit Ethernet
SONET/SDH Ethernet over optical networks
5ControlNumber
LAN protocols
Layers 1 and 2 Map into OSI reference model
Souce: Cisco
6ControlNumber
FDDI
Developed by American National Standards Institute (ANSI) Originally proposed as internal fiber optic I/O channel for
computers Later became generalized to high-speed LAN running at
100 Mbps– Can run on copper as well as fiber– Dual-ring is usual configuration– Can go up to 200 Mbps with single ring
Token ring architecture– Advantage of token-passing networks: deterministic– Possible to calculate maximum time before station can
transmit• Popular in real-time environments
7ControlNumber
Characteristics of FDDI
Token ring architecture– Two countercirculating rings– Only one used for data; other for backup
Ring size– Up to 200 km (on multimode fiber, single ring)– Dual ring size up to 100 km– Maximum of 500 stations
• Max distance between stations is 2 km Packet switched: utilizes variable length frames
– Max frame size is 4500 bytes– Frame header contains destination address
8ControlNumber
Characteristics of FDDI (continued) Guaranteed bandwidth availability
– Equality of access as in all token-ring systems– Guaranteed bandwidth for synchronous traffic
Token-ring protocol– Similar to IEEE 802.5 token-ring LAN– Differs in that it is dependent on timers
Ring stations– Each may connect to both rings or only primary ring
Ring monitor– Performed cooperatively by all stations rather than by
single active monitor• All look for errors; if found any station can request
reinitialization of ring– Each station does not have to have ring monitor
function
9ControlNumber
FDDI ring structure, with/without break
Source: Dutton
10ControlNumber
FDDI ring configuration
Source: Dutton
11ControlNumber
FDDI token ring protocol operation
Ring access controlled by special frame called a “token”– Only one token present at any time– When a station receives the token it has permission to
send– When station finishes sending it must place token back
on ring Each station on the ring receives and retransmits frames
– Ring is not a node
12ControlNumber
Timing on FDDI
3 timers required due to need to handle synchronous traffic– Token rotation timer (TRT)
• Elapsed time since last token received– Target token rotation timer (TTRT)
• Target maximum time between tokens time for token to traverse ring
• 4 msec < TTRT < 165 msec• Optimal value often around 8 msec
– Token holding timer (THT)• Governs max amount of data station may send• Max time allocated for station to send
13ControlNumber
Operation
When station receives token it compares time since last token (TRT) with target time (TTRT)– Normal operation: TRT < TTRT
• Station can send multiple frames until TTRT reached• TTRT-TRT = THT
– Overload: 2xTTRT> TRT > TTRT• Synchronous data only permitted
– Error: TRT > TTRT• Must be conveyed to LAN manager
Delays may occur– Stations must be capable of buffering data
Stations must remove data they send when it returns to them
May be many frames on ring, but only one token
14ControlNumber
Operation (continued)
When ring initialized, stations cooperate to determine TTRT value– Minimum of all requested TTRT values– Changed only if new station enters ring
15ControlNumber
Physical media for FDDI
Multimode fiber– Originally defined mode of operation
Single mode fiber– Included in standard but little used
Twisted pair copper wire– STP = shielded twisted pair
• Not as good as fiber, but cheaper– UTP-5 (=cat 5) unshielded twisted pair standard in 1994
16ControlNumber
Media specifications
Medium Fiber Light
sourceDetector Transmit power Receiver
sensitivity
Multimode 62.5/125
50/125
85/125
100/140
LED PIN diode (1) -20 to -14 dBm
(2) -4 to 0 dBm
(1) -31 to -14 dBm
(2) -37 to -15 dBm
Single mode
9 micron LED PIN diode (1)-20 to -14 dBm
(2)-4 to 0 dBm
(1)-31 to -14 dBm
(2)-37 to -15 dBm
17ControlNumber
Data encoding and clocking
Four data bits encoded as five bit group– 100 Mbps actually 125 Mbaud on ring– Allows adding of more transitions into bit stream to allow
for problem of too many 1s or 0s Uses Non Return to Zero Inverted (NRZI) encoding Each station has own clock
– Specification is accuracy of 0.005%– Max difference between stations 0.01%– 10 bit buffer inside each station to allow for differences in
clocks between stations• Gives average of 4.5 bit times to smooth out timing
differences– Determines max frame size
4.5 bits/0.01% = 45,000 bits = 9,000 symbols = 4,500 bytes
18ControlNumber
Physical layer operation
Source: Dutton
19ControlNumber
Comparison with standard token ring networks
Standard TRN uses Manchester encoding– Allows exact recovery of clock, but at cost of doubling
frequency FDDI uses optical signals at higher speed than TRN
– Does not have exact clock recovery, substitutes buffer
20ControlNumber
FDDI layers
Source: Dutton
21ControlNumber
FDDI layers (continued)
Physical Medium Dependent layer (PMD)– Optical link parameters– Cables and connectors– Optical bypass switch– Power levels
Physical Layer Protocol (PHY)– Access to ring– Clocking, synchronization, buffering– Code conversion– Ring continuity
22ControlNumber
FDDI layers (continued)
Media Access Control (MAC)– Tokens and timers– Frame check sequence
Station Management (SMT)– Ring Management (RMT)
• Ensures valid token circulating– Connection Management (CMT)
• Physical connections and topology– Operational Management
• Monitors timers and parameters• Interfaces to external network management software
23ControlNumber
SONET overview
SONET = Synchronous Optical Network– Should have been called Synchonous Opto-electronic
network (SOENET) Technology developed in 1980s for long-haul trunks
needed by Telcos– Formulated by Exchange Carriers Standards
Association (ECSA)• Industry group which sets standards for telecoms• 1984 work began
– Expected to serve as basis for Telcos for 20-30 years– Designed from ground up based on 64kbps channels
(DS0—voice channels)• Everything a multiple of this
24ControlNumber
SONET (continued)
Emphasis on qualities important to Telcos– Reliability– Availability– Millisecond recovery from outages
Optimal use of bandwidth of secondary concern Not originally intended as bulk data carrier or carrier for
asychronous packets Serves as transport only
– Does not do switching Utilizes optical components only because copper not fast
enough– Otherwise copper or fiber could transmit SONET
25ControlNumber
Advantages of SONET
Reduction in equipment Standardization of equipment to allow for plug and play Increased network reliability Provision of overhead and payload bytes Synchronous multiplexing format
– Allows carrying of different loads– Simplifies interfacing to switching equipment
26ControlNumber
Basic structure of SONET
Utilizes time division multiplexing to combine large number of individual signals
Structured in fixed-length frames Entire network operates synchronously Synchronous operation requires extremely precise
clocking throughout network– Utilizes Stratum atomic clock
• Known as “Primary Reference Clock” (PRC)• Accurate to 1 part in 1011
27ControlNumber
Basic structure of optical part of SONET
Modulator
Input signal Connector Optional optical amplifier
Amplifier Decoder
Output signal
Optical fiber Optical fiber
Light Wavelength = 800-1600 nm
Electricity Electricity
Light source
Detector
Input SONET
signal(time
multiplexed individual
signals)
28ControlNumber
Evolving SONET network architecture
Source Encoder(Time
Division Multiplexer)
Modulator/ transmitter
(Wavelength multiplexer)
ReceiverDecoder(Demux)
Receiver/ demodulator
(Demux)
Link
end user services
end userservices
SONET
SONET
DWDM
DWDM
SONET
SONET
end user services
end user services
1
n
29ControlNumber
SONET structure
First step in SONET multiplexing process: generation of lowest level or base signal– Referred to as Synchronous Transport Signal level-1, or
STS-1– 51.84 Mbits/second– Higher level signals are multiples of this, giving rise to
STS-N• N is not arbitrary, but restricted to certain values• STS-N signals composed of N byte-interleaved STS-1
signals– Optical counterpart known as “Optical Carrier level-1”
or OC-1
30ControlNumber
SONET hierarchy
Source: Tektronix
31ControlNumber
SONET frame format 810 bytes
– Logically a 90 column by 9 rows– Order of transmission: row by row, L to R within rows,
most significant byte first
9 rows
90 columns
Source: Tektronix
32ControlNumber
SONET frame format (continued)
One frame per 125 sec = 8,000 frames/sec 8,000 frames/sec x 810 bytes/frame x 8 bits/byte = 51,840,000
bits/sec Column = 9 bytes x 8000 per second x 8 bits/byte = 576K bits
SONET frame
TransportOverhead
Synchronous Payload Envelope (SPE)—783 bytes
STS PathOverhead
(POH)—9 bytes
Payload756 bytes(84 cols.)
Fixedstuff
18 bytes
33ControlNumber
SONET frame structure: SPE
Source: Tektronix
34ControlNumber
SONET frame structure (continued)
SPE does not have to be aligned with STS frame– Can begin anywhere in STS frame– Starting location designated by STS payload pointer in
transport overhead
Source: Tektronix
35ControlNumber
Overhead structure
Two types– Transport (27 bytes)
• Section (9 bytes)• Line (18 bytes)
– Path (9 bytes, embedded in SPE)
36ControlNumber
Overhead structure (continued)
Source: Tektronix
37ControlNumber
Detailed structure of overhead
Source: Tektronix
38ControlNumber
Function of overhead
Section (9 bytes)– Performance monitoring (STS-N signal)– Local orderwire– Datacomm channels to carry info for OAM&P– Framing
Line overhead (18 bytes)– Locating SPE in frame– Multiplexing or concatenating signals– Performance monitoring– Automatic protection switching– Line maintenance
39ControlNumber
Function of overhead (continued)
Path overhead (9 bytes)– Performance monitoring (STS SPE)– Signal label (contents of STS SPE)– Path status– Path trace
40ControlNumber
SONET alarms
Three levels to allow close monitoring of deteriorating conditions– Anomaly: discrepancy between observed and expected
• Does not constitute interruption in service– Defect: density of anomalies reached level where
service is interrupted• May be correctable
– Failure: Inability of function to perform required action (defect) persisted beyond allowable time span
41ControlNumber
SONET alarms
42ControlNumber
SONET Alarms (continued)
Source: Tektronix
43ControlNumber
Tributaries and Virtual Tributaries (VTs)
Need exists to transmit channels slower than full STS– Called tributaries or virtual tributaries– Only certain channel speeds allowed
Tributaries may occupy a number of consecutive columns within payload or be interleaved (time multiplexed) (usual)– US T-1 (1.544 Mbps) uses 3 columns
• Only requires 24 slots, given 27 = 3 slots wasted• Recall that each slot is 64 kbits, x 24 = 1.544 Mbps
– European E-1 (2.048 Mbps) uses 4 columns• Only requires 32, given 36 = 4 slots wasted
– Benefit is that single tributary can be demultiplexed without need to demultiplex entire stream
44ControlNumber
VT sizes
Used for T1
Used for E1
Source: Tektronix
45ControlNumber
Tributaries (continued)
An SPE carrying VTs is divided into 7 VT groups– Each group consists of 12 columns– 12 x 7 = 84 columns = payload capacity
Columns for each VT type are all factors of 12 Each VT group can carry only one VT type
– Cannot mix VT1.5 and VT3, even though they would fit– Separate VT groups within frame can carry different VT
types– Allowed combinations within a VT group
• 4 VT1.5• 3 VT2• 2 VT3• 1 VT6
Within group, VTs are interleaved (time multiplexed)
46ControlNumber
Multiplexing of VTs within group
Source: Tektronix
47ControlNumber
Multiplexing of VT groups
Source: Tektronix
48ControlNumber
Pointers
Used to compensate for frequency and phase variation Allow transport of synchronous payloads across
plesiosynchronous (almost synchronous) network boundaries
Avoid delays and losses of having to use 125 sec slip buffers
Dynamically and flexibly aligning payloads– Dropping– Inserting– Cross-connecting
Effects of jitter can also minimized
49ControlNumber
Pointers (continued)
Byte stuffing used to fix alignment dynamically– Positive: byte added– Negative: byte deleted
Does not affect data
50ControlNumber
Pointers (continued)
Source: Tektronix
51ControlNumber
Layers of multiplexing in SONET
Time division– (1) Data prior to sending to SONET
• E.g., several slow-speed channels multiplexed to make T1
– (2) Within VT group• E.g., several T1s
– (3) Among VT groups in STS frame– (4) Among STS frames for speeds greater than OC-1
• May be done multiple times, e.g., 4 OC-3 to OC-12, 4 OC-12 to OC-48, 4 OC-48 to OC-192
If WDM used, (5) wavelength multiplexing of SONET signals
52ControlNumber
SONET multiplexing (continued)
56K
128K
384K
x1001 Tbps
TDMLevel 1
TDMLevel 2
TDMLevel 3
TDMLevel 4
WDMLevel 5
53ControlNumber
SONET network elements
Terminal multiplexers– Level 3 or 4
Regenerator (repeater) Digital loop carrier (DLC)
– Concentrator at level 1 Add/drop multiplexer (ADM)
– Picks off multiplexed signals– Adds new signals
Source: Tektronix
54ControlNumber
SONET network elements (continued)
Digital cross-connects (DCS or DCX)– Accesses signals at STS-1 level and switches them– SONET equivalent of DS3 cross connect– Allows overhead to be maintained because network is
synchronous– Can make 2-way connections at DS3, STS-1, STS-Nc
levels• STS-Nc requires contiguous, not interleaved bytes
Source: Tektronix
55ControlNumber
SONET network configurations
Point-to-point Point-to-multipoint Hub Ring
56ControlNumber
Point-to-point
Two terminal multiplexers connected by optical link– May or may not use repeaters– Simplest SONET application
Source: Tektronix
57ControlNumber
Point-to-multipoint
Linear add/drop architecture– Circuits added, dropped along the path
SONET ADM designed for this task– Avoids need to completely demux signal, cross-connect
channels, remux– Typically placed along path to allow adding, dropping
channels where needed
Source: Tektronix
58ControlNumber
Hub Concentrates traffic at one or more sites Allows for easy reprovisioning Two implementations
– Cross-connecting tributary services• Requires 2 or more ADMs, cross-connect switch
– Cross-connecting at tributary and SONET level• Requires cross-connect switch
Source: Tektronix
59ControlNumber
Ring architecture Most popular architecture
– Used by all major carriers Basic building block is ADM Bi-directional or uni-directional traffic Main advantage: survivability
– If fiber cut, multiplexers canreroute in milliseconds
Source: Tektronix
After cut
60ControlNumber
Ring architecture (continued)
Source: Tektronix
61ControlNumber
Limitations of SONET ring architecture
SONET ring architecture very complex Main problem is scalability
– To increase capacity or add new locations requires building a new set of rings, which is very expensive
– Mitigated to some extent by DWDM But hardware is standardized and available from multiple
sources– SONET does its job well– Is established and low-risk technology
62ControlNumber
SONET and SDH
SDH=Synchronous Digital Hierarchy– Used widely outside of US, Japan– Same 125sec frames– Developed to accommodate different world standards
• T1-based• E1-based
– Original SONET standard changed from bit interleaving to byte interleaving
– SONET is subset of SDH
63ControlNumber
SONET/SDH hierarchies
Source: Tektronix
64ControlNumber
Non-synchronous hierarchies
Source: Tektronix
65ControlNumber
Address for following slides:
http://www.cisco.com/networkers/nw00/pres/pdf2000.htm
Presentation # 3003
66ControlNumber
67ControlNumber
68ControlNumber
69ControlNumber
70ControlNumber
71ControlNumber
72ControlNumber
73ControlNumber
74ControlNumber
75ControlNumber
76ControlNumber
Ethernet
Primarily of interest because of newer, high-speed versions– Gigabit Ethernet (GBE)– 10 Gigabit Ethernet
Fast Ethernet (100 Mbps) can run on fiber, but normally implemented with Cat-5 UTP
77ControlNumber
Brief review of Ethernet operation
All stations connected to bus, which is in effect a node Ethernet uses “Carrier Sense Multiple Access with
Collision Detection” (CSMA/CD) to control bus traffic Stations transmit independently and asychronously
– If a frame is received, all stations check to see if it is addressed to them
– If two stations transmit simultaneously or closely in time, a “collision” occurs
No guarantee that data will get through without error– Requires higher level protocol to monitor and indicated
need for retransmission
78ControlNumber
Clarification (continued)
Most modern Ethernet network interface cards (NICs) can operate either half duplex (with bus or hub) or full duplex (with switch)
Switches are sold by all major vendors– Improve throughput on slower speed LANs– Not much more expensive than hubs– Allow more devices to be connected to LAN
79ControlNumber
Source: Luxpath/IEC
80ControlNumber
Operation of Ethernet (continued) Operation of CSMA/CD
– If a station wishes to send, it must listen to see if another station is transmitting
• If so, must wait until bus is free
• If not, it can begin to transmit
– Because of signal propagation delays down the bus, a station may be unaware that another has begun to transmit
• If this occurs, called “collision”, garbage is result
• Transmitting station must listen to bus to monitor for collisions
• If collision detected, transmitting station sends “jamming” signal to improve chance that other station detects collision, then stops transmitting
– If collision occurs, all transmitting stations must cease transmission and wait for (different) random periods before retransmitting
81ControlNumber
Ethernet and OSI reference model
Application
Presentation
Session
Transport
Network
Data Link
Physical
TCP
IP
Applications:TelnetFTP
SMTPHTTP
Ethernet (802.3)
LLC SublayerMAC Sublayer
Physical signalingMedia attachment
TCP/IP
ApplicationProtocols
OSI Reference ModelTCP/IP Implementation
Using Ethernet
Source: IBM
82ControlNumber
Bus and hub architectures
Source: Dutton
83ControlNumber
Half-duplex and full-duplex
Meaning of half duplex (HDX) and full duplex (FDX)– Terms going back to teletype days– Half-duplex = same physical line (or bus) used for both
transmit and receive• Requires special protocol to prevent simultaneous
transmission and reception– Full-duplex = different physical line used for both
transmit and receive• Does not require special protocol, but does require
dedicated (at least temporarily) connection
84ControlNumber
Half-duplex and full-duplex (continued)
Original Ethernet: half-duplex because all transmitting and receiving on same bus
85ControlNumber
Implementation of Ethernet
Physical bus rarely used anymore– Too difficult to manage and repair– Unwieldy to add or change workstations– Requires coax cable in most cases
Implementations done with hub and Cat-5 UTP– Logically looks like bus
Manchester encoding always used– Signal always has transition with every bit
• Logic 0: 0 to 1 transition at bit center• Logic 1: 1 to 0 transition at bit center
– Effectively doubles frequency
86ControlNumber
Implementation of Ethernet (continued)
Example of Manchester encoding
Manchester encoding important for collision detection– Because a 0 level and a 1 level occur for each bit, code
is “balanced”• Average DC level is ½ of logic 1 level
– If collision occurs, signals are “ORed”, which raises average DC level
– Detected and interpreted as collision by transceivers
1 1 10 0 0
87ControlNumber
Implementation on fiber
Collision detection– Light pulses converted to electricity in transceivers– Average DC value will also change when light pulses
collide on fiber Uses LEDs at 850 nm
88ControlNumber
CSMA/CD performance and propagation delay
Propagation delay is main factor limiting performance of Ethernet– Delay means station may begin transmitting when bus
not free– Also means stations will learn that bus is free at
different times Collisions reduce utilization of Ethernet LAN because they
force two or more retransmissions Maximum utilization (maximum throughput) given by
1/(1+6.44)
where
= end-to-end delay/transmission time
89ControlNumber
Ethernet throughput vs. offered load
Source: Dutton
90ControlNumber
CSMA/CD performance and propagation delay (continued)
On copper wire, transmission speed about 5.2 sec/km For 10 mbps Ethernet, with 1000 bit frame size, utilization
estimated as• = 2 x 5.2 sec/100 sec = 0.104• Max utilization = 1/(1+6.44x0.104) = 0.60 = 60%
For 100 mbps Ethernet, same frame size,• = 2 x 5.2 sec/10 sec = 1.04• Max utilization = 1/(1+6.44x1.04) = 0.13 = 13%
For 1 Gbps Ethernet, same frame size,• r = 2 x 5.2 sec/1 sec = 10.4• Max utilization = 1/(1+6.44x10.4) = .0147 = 1.5%
91ControlNumber
Ways to fix deteriorating high speed performance Changing frame size not practical Only other variable is propagation delay
– Needs to be made shorter– Must shorten maximum length of cabling
Light in fiber takes about 3.3-5 sec to travel 1 km, not that different than electricity pulses in copper wire
Standard Ethernet: 1.6 km max LAN segment length (“collision domain”)
High speed Ethernet: 200 m max LAN segment length
92ControlNumber
Ways to fix speed problem (continued)
Gigabit Ethernet: would be ~ 20 m, but 200 was kept as spec– Other changes need to be made– Switches used instead of hubs– Minimum frame size 512 bytes, max same as before,
1524 bytes Switch is layer 2 device Reads addresses of frames and sends frame only to
destination– Reduces chances of collision significantly– Increases utilization seen by stations
Use of switches and routers also allows conventional Ethernet networks to span large areas
93ControlNumber
Ways to fix speed problem (continued) 10 Gigabit Ethernet uses only full duplex to avoid timing
problems associated with CSMA/CD protocol– Lower speed versions can use it as well– Requires switch which physically connects two devices
which are communicating– No collisions because both connected devices can
transmit and receive at same time
Terminal 1
Terminal 2 Terminal 3
Terminal 4SwitchT
R
94ControlNumber
Gigabit Ethernet standard
Shielded twisted pair up to 500 m UTP cat-5 available
– Requires 5-level encoding– 100 m max length
Cat-7 standard under development– Shielded twisted pairs
Single mode fiber at 1310 nm, up to 2 km Multimode fiber at 780 nm (CD-ROM lasers) or VCSELs at
850 nm– On 62.5/100 MM fiber up to 200m– May be extended to 1 or 2 km
95ControlNumber
Cabling standards
Source: 10 Gigabit Ethernet Alliance
96ControlNumber
10 Gbit Ethernet Fiber only Full duplex only, in combination with switches, will not
need CSMA/CD protocol required for half-duplex slower Ethernet
Standard called IEEE 802.3ae; see http://grouper.ieee.org/groups/802/3/ae/ for info on the spec
Source: 10 Gigabit Ethernet Alliance
97ControlNumber
10 Gbit Ethernet
For further info,
www.10gea.org
Source: 10 Gigabit Ethernet Alliance
98ControlNumber
Growth rate anticipated for Ethernet
Source: Luxpath/IEC
99ControlNumber
Ethernet over SONET
Ethernet over SONET inefficiencies
Source: Cisco
100ControlNumber
Future of Ethernet in WAN Objective is to replace SONET at 10 Gbit (OC-192) level Idea is to use Ethernet switch/routers to deal with failures
– Take advantage of extremely low error and failure rates with modern optical fiber
– Simplified architecture• Fewer network elements• No need for rings
– Not established technology yet, high risk compared to SONET
101ControlNumber
Future of Ethernet (continued)
10 Gig Ethernet may replace ATM Ethernet already dominates the LAN, where ATM never
made much headway ATM dominates MAN and WAN
– Ethernet could displace ATM because it would eliminate need to switch protocols
– Has not happened yet In 2004, an 18,500 km 10 Gbit Ethernet link from CERN to
Japan was put into service– Special hardware needed for servers, but switches
could handle the speed 100 Gig Ethernet in very early stages
– Deployment not expected until 2010
102ControlNumber
Other trends in Ethernet
All-optical Ethernet switches– Eliminate need for conversion back to electronic form– Useful in 10 Gbit WAN applications
Source: Luxpath/IEC
103ControlNumber
Fiber Channel
Developed by ANSI to address problems of existing computer channel interfaces
Main thrust: connecting disk drives or arrays of disk drives with computer systems– Allows systems managers to combine data warehouses
spread over a campus or—with repeaters—a metropolitan area
Primarily within computer, but can also be used as LAN Allows interconnection of computers and peripheral
devices– Point-to-point– Crosspoint switch– Arbitrated loop
104ControlNumber
Fiber Channel (continued)
Architecture is neither a channel nor a real network topology– An active intelligent interconnection scheme, called a
Fabric, to connect devices High performance serial link supporting its own, as well as
higher level protocols such as the FDDI, SCSI, HIPPI and IPI Speeds up to 4 Gbit/s (higher speeds planned for future)
– 8 Gbit standard ratified– 10 Gbit used now but only to interconnect switches
Can be converted for Local Area Network technology by adding a switch
Primary application is in storage area networks Can also run on copper twisted-pair
105ControlNumber
Fiber channel topologies
Point-to-Point
Crosspoint switch
Arbitrated loop
Source: Dutton
106ControlNumber
Characteristics of FDDI topologies
Source: Wikipedia
107ControlNumber
Fiber Channel Speeds
133 Mbit/sec 266 Mbit/sec 530 Mbit/sec 1 Gbit/sec 2 Gbit/sec 4 Gbit/sec Highest performance: 10 km at 1 Gbit/sec
108ControlNumber
Terminology
N_Port: connection for device to fiber channel F_Port: special connection to crosspoint switch fabric NL_Port: N_port in arbitrated loop FL_Port: F_Port connected to arbitrated loop
109ControlNumber
Media for Fiber Channel
Uses single mode or multimode fiber– Single mode
• Lasers at 1300 nm, 1550 nm• Data rates up to 1 Gbps• Distance up to 10 km at 1300, >50 km at 1550
– Multimode• Laser at 780 nm, 850 nm
– Distance up to 2 km• LED at 1300 nm
– Distance up to 1.5 km
110ControlNumber
Classes of service
Class 1: Dedicated (connection oriented)– 2 N_Ports– Maximum bandwidth guaranteed
Class 2: Multiplex– Connectionless– Acknowledgement of successful delivery
Class 3: Datagram– Connectionless– Best effort– No acknowledgement