SONET

- when fiber optics first used => every telephone company had own proprietary optical 
  TDM system

- after AT&T broken up in 1984, local companies had to connect to multiple long-distance 
  carriers all with different optical TDM systems

 - thus needed standardization

- 1985 - Bellcore (RBOC research arm) -> standard -> SONET; later CCITT joined & their 
   standard called SDH (Synchronous Digital  Hierarchy) -> only differs in minor ways 
   from SONET

* nearly all long-distance phone traffic in USA & much of elsewhere now use trunks 
   running SONET in physical layer

- as SONET chips get cheaper, SONET interface boards for computers may become more
  common

 - thus companies may be able to plug their computers directly into heart of phone network
   over specially leased lines

 4 Major Design Goals of SONET

1) make it possible for different carriers to interwork
  
- this required defining common signaling standard with respect to wavelength, timing, 
  framing structure, etc.

2) means of unifying USA, European, Japanese digital systems all of which based on 64-
  kbps PCM channels but all of which combined them in different & incompatible ways

3) way to multiplex multiple digital channels together

 - at time formed, highest speed was T3 = 44.736 Mbps but wanted hierarchy to go up to 
   gigabit/sec.

4) support for operations, administration & maintenance (OAM)

- traditional TDM system  ->  entire bandwidth of fiber devoted to one channel containing 
  time slots for various subchannels i.e. synchronous

 -  controlled by master clock with accuracy of 1 in 109

- bits on SONET sent out at precise intervals controlled by master clock

- consists of switches, multiplexers & repeaters

	see Fig 2.29 - SONET path

- basic SONET frame -> block of 810 bytes every 125 microsec. i.e. 8 x810 = 6480 bits
  8000 times/second
				see Fig. 2.30

- first 3 rows of overhead -> section overhead
   - generated & checked at start & end of each section

- next 6 rows of overhead -> line overhead
 - generated at start & end of each line

- remaining 87 columns  -> 87 x 9 x 8 x 8000 = 50.112 Mbps of user data -> SPE
  (Synchronous Payload Envelope)

- does not always begin in row 1 column 4
- can begin anywhere
- pointer to first byte of SPE in first row of line overhead
- first column of SPE is path overhead (header to end-to-end path sublayer protocol)

			see figure 2.30

* useful for ATM where 53-byte cells do not fit exactly into frames


Multiplexing of multiple data streams - tributaries

		see figure 2.31

 - done byte for byte
 - e.g. 1 byte from trib. 1, 1 from trib. 2., 1 from trib. 3,  => back to trib. 1, etc.

 - thus for STS-3 figure comparable to figure 2.30 would have (from left to right) columns 
    from trib. 1, trib. 2, trib. 3, trib. 1, ..., out to 270 columns wide

- so one 270 x 9 byte frame sent every 125 micro sec. giving 155.52 Mbps data rate


SONET physical layer divided into 4 sub-layers	see figure 2.33

-section -> point-to-point fiber run -> generates standard frame at one end &processes it at 
  other end

- sections can start & end at repeaters which can amplify & regenerate bits but do not 
  change or process them in any way

- line sub-layer -> multiplexes multiple tributaries onto single line & demultiplexes at other 
  end

- to line sub-layer, repeaters are transparent

  - protocol on line sublayer between 2 multiplexers & deals with issues such as how many 
    inputs are being multiplexed together & how

  -path sub-layer -> protocol deals with en-to-end issues

- in design of ATM want 155 Mbps to run on SONET- OC-3 trunks

			see figure 2.32

  STS ->  electrical;	OC -> optical         see figure 2.25

SONET continued

- in terms of future use - needed to allow for more sophisticated service offerings

 e.g. virtual private networking, time-of-day bandwidth allocation, & support of broadband 
  ISDN ATM transmission

- thus much emphasis on NW management capabilies in synchronous signals  i.e. overhead

HIERARCHY

Photonic
 - physical layer
- specification of type of optical fiber -> e.g. minimum powers & dispersion characteristics 
  of transmitting lasers, required sensitivity of receivers

Section
 - creates basic SONET frames
* - converts electronic signals to photonic ones
- some monitoring capabilities

Line
 - responsible for synchronization, multiplexing of data onto frames, protection & 
   maintenance functions, & switching

Path 
- responsible for end-to-end transport of data at appropriate signaling speed

transmission - one row at time from left to right, top to bottom

- line overhead has pointer to where path overhead starts

		Fig. 14.5,  Table 14.4


Pointer Adjustment

note:  in conventional circuit-switched NWs, most MUXs & telephone company channel 
 banks require demultiplexing & remultiplexing of entire signal just to access piece of 
  information addressed to node

e.g. if T-1 MUX B receives data on single T-1 circuit from T-1 MUX A, & passes data to  
  MUX C

 - in signal received, single DSO channel (64kbps) addressed to node B; rest will pass on to 
   node C & then further into NW

 - to remove single DSO channel, B must demultiplex every bit of 1.544  Mbps signal, 
    remove data, remultiplex every bit

 - few proprietary T-1 MUXs allow drop-and-insert capability but this does not allow 
   communication with other vendors

* SONET provides standard drop-and-insert capability  ->  applies not just  to 64 kbps 
  channels but to higher data rates also

- SONET  ->  set of pointers that locate channels within payload, & entire payload within 
   frame  ->  thus info can be accessed, inserted & removed with simple adjustment of 
   pointers

- pointer info contained in path overhead that refers to multiplex structure of channels 
   contained in payload

- pointer in line overhead plays similar role for entire payload

- recall:  SPE can float with respect to frame so actual payload (87 col. x 9 rows) can 
  straddle 2 frames  ->  H1,H2 octets in line overhead indicate start of payload

			Fig 14.6

* - atomic timing sources can differ by small amounts so SONET must cope with timing 
  differences

- thus each node must recalculate pointer to let next receiving node of exact location of start 
 of payload  ->  so payload allowed to slip through an STS-1 frame, increasing or 
 decreasing pointer value at intervals by one byte position

- if payload rate higher than local STS frame rate  =>  pointer decreased by one octet 
  position so that next payload will begin one octet sooner than earlier payload

- to prevent loss of an octet on payload that is thus squeezed, H3 octet used to hold extra 
  octet for that one frame
						Fig. 14.7 (a)

OR - if payload rate lags behind frame rate, insertion of next payload delayed by one octet
  =>  octet in SPE that follows H3 octet left empty to  allow for movement of payload

		Fig. 14.7 (b)

Fiber Optics

Recall:  much made of speed with which computer technology improving

e.g. in 1970s fast computer (CDC 6600)  -> execute one instruction in 100ns while in 1990 
 Cray could execute instruction in 1 ns  -> factor of 10 improvement /decade

- during same period, data communications went from 56 kbps (ARPANET) to 1 Gbps 
 (modern optical communication)  ->  gain of more than 100/decade while error rate went 
  from 10-5 per bit to nearly 0

- single CPUs beginning to reach physical limits  e.g. speed of light, heat dissipation but 
  current fiber technology  ->  achievable bandwidth is > 50,000 Gbps (50 Tbps) & 
  searches for even better material

* current practical signaling limit of about 1 Gbps due to inability to convert between 
   electrical & optical signals any faster

 - in lab, 100 Gbps feasible on short runs
 - 1 terabit/sec considered likely in few years
 
- considered fully optical systems - including into & out of computer, within reach

- so in race between computing & communication --> communication wins!

- implies essentially infinite bandwidth (but not at zero cost) -> not fully accepted by 
  engineers focussing on copper wire

- now -> computers are hopelessly slow so NWs should avoid computation at all costs, no 
  matter how much bandwidth is wasted

Optical Transmission System - 3 parts

 1) light source
 2) transmission medium -> ultra thin fiber of glass
 3) detector -> generates electrical pulse when light falls on it

convention:  pulse of light  ->  1 bit;  absence of light  ->  0 bit 

- so attach light source at one end of optical fiber & detector at other end  => have 
  unidirectional data transmission system that accepts electrical signal, converts & transmits 
  it by light pulses  =>  reconverts output to electrical signal at receiving end

- such a system would leak light & be useless except for a principle of physics:  when light 
  ray passes from one medium to another, ray is reflected at boundary so for angles of 
  incidence > certain critical value  => light refracted back into silicon i.e. none escapes into
  air

- thus light ray incident at or above critical angle trapped inside fiber

		Fig. 2.5

- many different rays can bounce at different angles - each ray is said to have a different 
  mode  ->  multimode fiber with many rays

- if fiber's diameter reduced to few wavelengths of light  => fiber acts as wave guide & 
  light can only propogate in straight line without bouncing -> single-mode fiber  

-> more expensive but can be used for greater distances e.g. several Gbps  for 30 km - 
  even higher rates in labs for shorter distances

 - some experiments have shown powerful lasers can drive fiber 100 km long without 
   repeaters

Transmission

- optical fibers made of glass -> made from sand --> cheap

- attenuation of light depends on its wavelength

- 3 wavelengths used for communication with 25,000 to 30,000 GHz wide:

  0.85,  1.30,  1.55 microns respectively with latter two having good attenuation 
  ( < 5%/km) while 0.85 has higher attenuation but lasers & electronics made of same
  material (Gallium Arsenide)

- light pulses sent down fiber spread out in length as propogate -> dispersion - amount is 
  wavelength dependent

- to keep spread-out pulses from overlapping can increase distance between them -> can 
  only do that by reducing signaling rate

- have recently discovered that by making pulses in special shape all dispersion effects 
  cancel out so may be possible to send pulses for thousands of kms without appreciable 
  shape distortion -  these pulses called solitons - still in lab

Fiber Cables

- in multimode fibers  ->  core is 50 microns in diameter (about thickness of human hair)

- in single-mode:  8-10 microns

fibers can be connected in 3 ways:

1) terminate in connectors & plugged into fiber sockets

  - connectors lose about 10-20% of light but make it easy to reconfigure systems

2) spliced mechanically
 - lay 2 cut ends in special sleeve & clamp in place
 - can fine tune by passing light through & making adjustments to maximize signal - lose 
   about 10% of light

3) fuse 2 pieces of fiber to form solid connection
 - nearly as good as single drawn fiber but small amount of attenuation

Two kinds of lights used for signaling:
 1) LEDs (Light Emitting Diodes)
 2) semi-conductor lasers	

		Fig. 2.8

-receiving end of optical fiber -> photodiode -> gives off electrical pulse when struck by 
  light

 - typical response time: 1 ns which limits data rates to about 1 Gbps

- thermal noise is issue -> so pulse of light must carry enough energy to be detected so by 
  making pulses powerful enough, error rate can be made arbitrarily small

Fiber Optic LANs

 - tapping into it more complex than connecting to Ethernet

 - note that ring NW really collection of point-to-point links

		Fig. 2.9

- interface at each computer passes light pulse stream through to next link & also serves as 
  T junction to allow computer to send & accept messages


Two Types of Interfaces

1) passive interface consists of 2 taps fused onto main fiber - one tap  -> LED or laser 
 diode at end for transmitting, & one tap -> photodiode for  receiving

 - tap completely passive so extremely reliable because broken LED or photodiode does not 
  break ring - just takes computer off-line

			Figure 12.10

2) active repeater - Fig. 2.9

 - incoming light converted to electrical signal, regenerated to full strength if weakened & 
   retransmitted as light 

 - interface with computer ordinary copper wire 
- purely optical repeaters now also used - do not require optical to electrical to optical 
  conversions => can operate at extremely high bandwidths

- if active repeater fails  =>  ring broken & NW goes down

- but, since signal regenerated at each interface, individual computer-to-computer links can 
   be kms long  ->  virtually no limit on total size of ring


- passive interfaces lose light at each junction so number of computers & total ring length 
  greatly restricted

- can also build fiber LAN as passive star

			Fig. 2.10

 - each interface has fiber running from transmitter to silica cylinder with incoming fibers 
   fused to one end of cylinder

- fibers fused to other end of cylinder run to each receiver

- when interface emits light pulse -> diffused inside passive star to illuminate all receivers
   -> i.e. broadcast

- incoming energy divided among all outgoing lines -> so, number of nodes in NW is 
  limited by sensitivity of photodiodes

Fiber Optics vs Copper Wire

fiber: much higher bandwidths than copper

  - low attenuation -> repeaters only about every 30 km (vs every 5 km for  copper) so 
   cheaper

 - not affected by power surges, electromagnetic interference, or power failures, or 
  corrosive chemicals in air as in harsh factory environments

- phone companies like fiber because thin & lightweight
 - many cable ducts completely full so no room to add new capacity
 - removing all copper & replacing it by fibers empties up ducts & copper
  has excellent resale value

- fiber is lighter than copper:
  - 1000 twisted pairs 1 km long -> 8000 kg
  -  2 fibers -> more capacity -> 100 kg
 also reduces need for expensive mechanical support system & its maintenance

- fiber does not leak light -> so difficult to tap - better security characteristics

- also when electrons move in wire -> affect one another & are affected by  electrons 
  outside wire but photons in fiber do not affect one another   (have no electric charge) & 
  are not affected by stray photons outside fiber

on negative side:
 - fiber new technology -> most engineers don't have experience to work with it (this will 
  change with time)

 - optical transmission inherently unidirectional => 2-way communication -> 2 fibers or
   2 frequency bands on one fiber

 - fiber interfaces cost more than electrical interfaces

LANs

 
 - packet broadcast networks

- each station attached to transmission medium shared by other stations

- simplest form -> transmission from any one station broadcast to, & received by all other
  stations
			Fig. 12.2;   Fig. 12.4

Bus & Tree Topologies

 - multipoint medium
 - bus -> all stations attach through hardware called tap

design issue - signal balancing
- when 2 stations exchange data over link => signal strength of transmitter must be adjusted
  within certain limits

- signal must be strong enough so that after attenuation across medium, it meets receiver's 
   minimum signal 
- strength requirements

- must also be strong enough to maintain signal-to-noise ratio but signal must not be so 
  strong that it overloads circuity of transmitter as signal would become distorted

- easily accomplished for point-to-point link

- signal balancing difficult for multipoint link

- if any station can transmit to any other station  => signal balancing must be done for all 
  permutations of stations taken 2 at a time i.e. for n stations => n(n-1) balancing so if
  n=200  =>  39,800 signal-strength constraints must be satisfied simultaneously

- so often divide medium into smaller segments within which pairwise balancing possible 
  using amplifiers or repeaters between segments

Baseband vs Broadband on Coaxial Cable

           Table 12.3

Baseband LAN - one that uses digital signaling i.e. binary data to be inserted onto the cable 
  as sequence of voltage pulses using usually, Manchester or Differential Manchester 
  encoding  =>  entire frequency spectrum of cable used

* so not possible to have multiple channels

- transmission bidirectional  i.e. signal inserted at any point in medium propogates in both 
  directions to ends  =>  absorbed

			Fig. 12.8

- requires bus topology

- unlike analog signals, digital signals connot easily be propogated through branching 
   points used in tree topology

- baseband bus can extend only few kms at most because of attenuation of signal especially 
  at higher frequencies  ->  blurring of pulses & weakening of signal

- originally used by Ethernet bus

- lower the data rate, the longer the cable can be  ->  follows since when  signal propogated 
  along transmission medium, integrity of signal suffers due to attenuation, noise, etc.

- longer the length of propogation  =>  greater the effect  ->  increasing probability of error

- but at lower data rate, individual pulses of digital signal last longer & can be recovered

e.g. 10Base5  =  10 Mbps baseband 500 m. 


Broadband Coaxial Cable
- in LANs, broadband coaxial  ->  analog signaling 
- can use FDM
-separate channels -> data traffic, video, radio
- allows for splitting, joining -> so both bus & tree possible
- much greater distances than baseband

Bus vs. Ring
- bus or tree broadband LAN better for user with large number of devices & high-capacity requirements

- for moderate requirements -> no clear choice

- baseband bus -> simpler  ->  passive taps rather than active repeaters  ->  no complex 
  bridges & ring concentrators

- ring -> point-to-point communication links

 - signal regenerated at each node so transmission errors minimized

- greater distances covered than with baseband bus
- broadband bus/tree -> similar distances but need cascaded amplifiers
 -> loss of data integrity at high data rates

- ring -> optical fiber links -> very high data rates -> excellent  electromagnetic interference 
  (EMI) characteristics

- electronics & maintenance of point-to-point lines simpler than for multipoint lines


Twisted Pair Star LANs

- increasing interest in use of twisted pair as transmission medium for LANs

- twisted pair cable cheaper than coaxial cable but cost of installing can be same for coaxial
  cable & twisted pair

- coaxial cable -> superior signal quality  =>  support more devices over longer distances at 
  higher data rates than twisted pair

- unsheilded twisted pair (USP) -> simply phone wire so most office buildings already 
  have it running from wiring closets to each office  =>  virtually no installation cost  ->  as
  wire there vs. coaxial cable must be pulled

- it can be very expensive to run coaxial cable to each office

- most popular approach to UTP for LAN is star-wiring

		Fig. 12.13,  Fig. 12.14

- hub ->  each station connected by 2 twisted pairs (transmit & receive) -> acts as repeater
  -> when single station transmits, hub repeats signal on outgoing line to each station

* although physically star  ->  logically a bus
 - a transmission from any one station received by all other stations, & if 2 stations transmit 
   at same time  ->  collision

- multiple levels of hubs
  - one header Hub (HHUB)
  - one or more intermediate hubs (IHUB)

Wireless LANs

- introduced in late 1980s as substitutes for traditional wired LANs

- saves cost of installation of LAN cabling & relocation

- but as new buildings wired for LANs & use of existing UTP  ->  not as much demand
  as expected

- are examples where wireless LANs useful  e.g.  buildings with large open areas 
  (manufacturing plants, stock exchanges, warehouses), historical buildings where no 
  drilling or existing wire present, small offices where installation & maintenance costs 
  prohibitive

- typically backbone wired LAN (e.g. Ethernet) that supports servers, workstations, one or 
  more bridges or routers

- control module (CM)  ->  interface to wireless LAN  -  includes either bridge or router 
  functionality to link wireless LAN to backbone

- has some sort of access control logic e.g. polling, token-passing scheme to regulate 
  access from end systems

- can also have end systems as standalone devices such as workstation or  server  ->  also, 
  hubs or other user modules (UM) to control number of stations off wired LAN

		Fig 12.17,  Fig. 12.18

- can have single cell or multiple cell wireless LANs

- can have point-to-point wireless link between building

- can have wireless link between LAN hub & mobile data terminal equipped with antenna 
  such as laptop computer or notepad computer

- can have temporary ad hoc peer-to-peer NW e.g. employees having temporary NW for
  meeting via laptops
							Fig. 12.19

- wireless LANs categorized according to transmission technique:

Infrared (IR) LANs
 - IR LAN limited to single room as IR does not penetrate opague walls

Spread Spectrum LANs
 - mostly operate in Industrial Scientific & Medical (ISM) bands so no FCC licensing 
   required in USA

Narrowband Microwave
 - operate microwave frequencies but not spread spectrum
 - some at frequencies requiring FCC licensing while others use one of ISM bands

              Table  12.6

More on LANs

Table 13.1 ->  IEEE 802.3
Fig.  13.4
Table 13.2

100 VG-ANYLAN

 -intended to be 100-Mbps extension to 10-Mbps Ethernet (like 100BASE-T is) & to 
  support IEEE 802.3 frame types

* but also compatible with IEEE 802.5 token ring frames

-uses new MAC scheme -> demand priority to determine order in which nodes share 
  network

- since it does not use CSMA/CD => new IEEE group:  802.12

topology: hierarchical star
 - simplest configuration  ->  single central hub & some attached devices

- more complex design  ->  single root hub plus one or more subordinate level-2 hubs
 ->  level-2 hub can have additional subordinate hubs at level-3, etc.


Medium Access Control
- round-robin scheme with 2 priority levels

Single-Hub NW MAC
 - if station wishes to transmit frame => issues request to central hub  =>  waits for 
   permission to transmit

 * must specify each request as normal priority or high-priority

- central hub continually scans all ports for request in round-robin fashion

- hub has 2 pointers:  high-priority pointer, normal-priority pointer

- hub grants each high priority request in order in which receives them

- if at any time, no high-priority requests  => grants normal-priority requests in order


Hierarchical NW

- set of hubs (at different levels) treated logically as single hub

			Fig. 13.12

- depth-first search of tree

-lower level hub can grant normal-priority requests as long as no high-priority requests 
 come in to R  ->  if so  =>  preempt when current transmission completes

- second refinement:  lower-level hub only keeps control for one round-robin cycle through
  its ports to prevent lock up by this hub

Signal Encoding
- key objective of 100VG-ANYLAN  ->  achieve 100-Mbps over short distances with 
  ordinary voice-grade cabling since many buildings already have this

- uses special kind of encoding - 5B6B -> uses 4 pairs & half-duplex mode

ATM LANs

Apple, Bellcore, Sun, Xerox identified 3 generations of LANs:

1st Generation: e.g. CSMA/CD, Token Ring
 -> terminal-to-host connectivity
 -> supported client/server architectures at moderate data rates

2nd Generation e.g. FDDI
 -> provides backbone LANs
 -> support of high-performance work stations


3rd Generation e.g. ATM LANs
 -> provide aggregate throughputs & real-time transport guarantees needed for multimedia

typical requirements of 3rd generation:
1) support multiple, guaranteed classes of services e.g. video -> 2 Mbps guaranteed; file 
  transfer -> background class of service

2) provide scalable throughput both in per-host capacity & aggregate capacity

3) facilitate interworking between LAN & WAN

ATM ideally suited for these

 -> virtual paths, virtual channels, multiple classes of service easily accomodated either on
  preconfigured, permanent connections or on demand -> switched connections

-> easily scalable by adding more ATM switching nodes & using higher or lower data rates 
  for attached devices

- term "ATM LAN" used in different ways but as a minimum refers to use of ATM data 
  transport protocol (more later)

e.g.
1) gateway to ATM WAN
 - ATM switch acts as router & traffic concentrator for linking premises' NW complex to 
   ATM WAN

2) Backbone ATM switch
 - either single ATM switch or local NW of ATM switches interconnect other LANs

3) Workgroup ATM
 -high performance multimedia workstations & other end systems connect directly to ATM 
   switch

OR mixture of above	

	e.g. Fig. 13.13

- as on-site demand rises  =>  simply increase capacity of backbone by adding more 
   switches, increasing throughput of each switch, & increasing data rate of trunks between 
   switches

- but doesn't solve all problems as end systems (e.g. workstations) still attached to shared-
   media LANs with their limitations on data rates

- so more advanced & more powerful approach -> hub

				Fig. 13.14

- each port operates at different data rate & different protocols

* each end system has dedicated point-to-point link to hub

  -> so each end system has communications hardware & software to interface to particular 
   type of LAN but LAN ony has 2 devices:  end system & hub

 ->  thus gets full throughput of medium

* have advantage of both approaches  ->  still use existing LAN installations & hardware of 
  "legacy" LANs while introducing ATM

- disadvantage: "mixed protocol" requires a protocol conversion capability

- "pure" ATM LAN -> end systems have ATM capability

- more on ATM in next chapter

Fibre Channel

- with increase in speed & memory capacity of personal computers, workstations & servers 
 =>  applications have become more complex with more graphics & video

-> affects 2 methods of data communications with processor:  I/O channel &  NW 
   communications

I/O channel:
- direct point-to-point or multipoint communications link, predominately hardware-based, 
  designed for high speed over short distances

- transfers data between buffer at source device & buffer at destination device without 
  regard to format or meaning of data

NW communications:
 - collection of interconnected access points with software protocol structure that enables 
  communication

-allows many different types of data transfer using software to implement networking 
  protocols & provide flow control, error detection & error recovery

* transfers between end systems

- Fibre Channel designed to combine best features of both approaches i.e. simplicity &
  speed of I/O channel with flexibility of protocol-based NW communications

channel-oriented facilities
- data-type qualifiers for routing frame payload into particular interface buffers

- link-level constructs associated with I/O operations

- protocol interface specifications to allows support of existing I/O channel
   architectures  e.g. SCSI (small computer system interfaces)


NW-oriented facilities
- full multiplexing of traffic between multiple destinations

- peer-to-peer connectivity between any pair of ports on Fiber Channel NW

- capabilities for internetworking to other communication technologies


Fibre Channel Elements
- key elements -> end systems or nodes

 - network itself  ->  consists of one or more switching elements -> forms fabric

- elements interconnected by point-to-point links between ports on individual nodes & 
  switches  =>  communication -> transmission of frames across point-to-point links

		Fig. 13.15

- each node has 3 or more ports, N_ports for interconnection

- each fabric-switching element has one or more ports  ->  F_ports

- interconnection by bidirectional links between ports

- any node can communciate to any other node connected to same fabric using services of 
  fabric

- frames can be buffered in fabric  =>  different nodes can connect to fabric at different 
  rates

- more general structure
				Fig. 13.16

- Fibre Channel more like traditional circuit switched or packet-switched NW vs. shared-
  medium LANs

- thus not concerned with MAC issues

- can easily scale by adding new switches & F_ports to existing fabric


Fibre Channel Protocol Architecture
		
	Fig.  13.17		Table 13.6

Note:  FC-3 provides set of services common across multiple N_ports of node (proposed):

striping
- uses multiple N_ports to transmit single information unit across multiple links 
  simultaneously -> higher aggregate throughput e.g. large data-sets in real time -> video-
  imaging

hunt groups
 - set of associated N_ports at single node
 - set assigned alias identifier
 - allows any frame sent to alias to be routed to any available N_port in set
 - may decrease latency by decreasing chance of waiting for busy N_port

multicast
 - delivers transmission to multiple destinations
 - includes sending to all N_ports on fabric (broadcast) or to subset of N_ports on fabric


Mapping
 
FC-4 defines mapping of various channel & NW protocols to FC-PH (FC Physical & 
 Signaling Interface)

I/O channel interfaces include:

1) SCSI (Small Comp. System Interface)
- a widely used high-speed interface typically implemented on personal computers, 
  workstations, servers

- used to support high-capacity & high-data rate devices e.g. disks, graphic & video 
  equipment

2) HIPPI (High-performance Parallel Interface)
- high-speed channel standard primarily used for mainframe/supercomputer environments

- at one point HIPPI & extensions to HIPPI considered as possible general-purpose high-
  speed LAN solution, but Fibre channel has superseded


NW interfaces:
 - IEEE 802 MAC frames map onto Fibre Channel frames; ATM; IP i.e. FC-4 mapping 
   protocols make use of FC-PH capabilties to transfer upper-layer protocol (ULP) 
   information

 - each FC-4 specification defines formats & procedures for ULP

FC Physical Media & Topologies

* a major strength of FC standard is that provides range of options for physical medium, 
  data rate on medium & topology

	Table 13.7		Fig. 13.18

- routing in fabric topology transparent to nodes -> each port has unique address

- point-to-point  -> only 2 N_ports directly connected

- arbitrated loop -> can connect up to 126 nodes in loop -> ports must act as both N_ports 
  & F_ports -> called NL_ports

- operates in similar manner to token ring

Wireless LAN standards
- developed by IEEE 802.11 committee -> not part of all commercial products

			Fig. 13.19

BSS - basic service set
 - set of   stations executing same MAC protocol, all competing for access to same shared 
   medium

- may be isolated or connect to backbone through access point which acts as bridge

- MAC protocol may be fully distributed or controlled by central coordination function 
  housed in access point

- BSS -> corresponds to term "cell"

ESS - extended service set
 - consists of two or more BSSs interconnected by distribution system -> typically wired 
   backbone LAN

 - ESS appears as single logical LAN to LLC level

3 types of stations defined based on mobility:

1) No-transition -> either stationary or moves within single BSS

2) BSS-transition -> movement from one BSS to another within same ESS

3) ESS-transition
 - movement from BSS in one ESS to BSS in another ESS

 - only supported in sense that station can move -> maintenance of upper-layer connections
    of 802.11 not guaranteed

 - disruption of service likely

Physical Medium Specification
 3 defined by 802.11

1) Infared at 1-Mbps & 2-Mbps operating at wavelength 850 & 950 nm

2) Direct-sequence spectrum operating in 2.4-GHz ISM band  ->  up to 7 channels, each 
   with data rate of 1-Mbps or 2-Mbps

3) frequency-hopping spread spectrum -> 2.4 GHz ISM band

Medium Access Control

IEEE 802.11 committee 2 approaches for MAC:

1) distributed-access protocols like CSMA/CD -> distributed decision to transmit over all 
   nodes using carrier sense  ->  made sense to ad hoc NW of peer work-stations or bursty  
   traffic

2) centralized access protocol
 -> natural for configuration with number of wireless stations interconnected with each 
     other & base station that attaches to backbone wired LAN

Result ->  MAC algorithm -> DFWMAC (distributed foundation wireless MAC)  ->  
   distributed access-control mechanism with optional centralized control built on top

		Fig. 13.20

DCF - distributed coordination function
 - simple CSMA algorithm

 - no collision detection since can not effectively distinguish between incoming weak signal 
   from noise & effects of its own transmission

- to ensure smooth & fair functioning of alg., DCF has set of delays that amounts to 
   priority scheme

- as with Ethernet, binary exponential backoff

- series of ACKs

Point Coordination Function - PCF
- althernate access method implemented on top of DCF

- polling with centralized polling master (point coordinator)

- allow time for asynchronous use by defining "superframe" -> part of  this devoted to 
   asynchronous use (CSMA) & rest for polling so that PCF does not lock out 
   asynchronous access