Chapter 6: Local Area Networks
LAN (Local Area Network) is a small size network which may cover a small geographical area such as a room, a floor, a building, or a campus depending upon its size. There is no predefined size for a LAN. The term LAN can be used to designate pretty wide range of network sizes. For example, few computers networked in a laboratory can be called as a LAN, and a fairly large LAN covering all the computers in an organization located can be called as a LAN. Let’s look at the examples of various LANs in an increasing order of their sizes.
· Home LAN or Personal LAN—Few computers at a home that are connected to a hub (usually Ethernet hubs) or a router(usually a router connected to a Cable network ISP or DSL network ISP) can be called as a LAN. Home LANs are usually used for sharing data between home computers and also to share an Internet connection to an ISP.
· Departmental LAN—A small network connecting several computers in a smaller part of a larger organization. Depending upon its size, it can be controlled by a separate personnel who belong to the personnel managing at the higher level.
· Organizational LAN—A large network covering all the computers connected throughout an organization. There could be several hierarchies in a network. For example, an organization has many departments and each departmental LAN is managed separately but the entire LAN is managed by the central IT department.
We can also classify LANs according to their functions.
· Personal LAN/Office LAN—A LAN supporting few personal computers. Common usages include file sharing, backup storage, sharing devices (printers, faxes, etc.), and sharing Internet connection.
· Backbone LAN—A LAN whose main function is to interconnect small LANs. A backbone LAN usually employs a higher speed LAN technology than its component LANs. For example, many 10Mbps Ethernet LANs can be connected to a backbone LAN running at 100Mbps speed.
· Backend LAN—A LAN that is used to interconnect large computers such as mainframes, supercomputers. These computers often need to communicate each other, therefore requires a highspeed connections for efficiency. These LANs are for large companies, large research organizations, and large databases.
· Storage Area Network(SAN)—SAN is a highspeed LAN made up of storage devices(Tape drives, harddisk, CD-ROMs, DVDs, etc.).
LANs have two major purposes: (a) To share data(database, files, programs, e-mails, etc.) (b) To share resource (printers, storage devices, processing power, etc.).
6.1 Characteristics of LANs
The main characteristics of LANs are:
a) Geographic scope
b) Connection topology
c) Transmission media
d) Speed
e) Medium Access Control method
(a) Geographic Scope
LANs are limited in terms of the area they cover which usually ranges from a small LAN in a room to a campus-wide LAN. There is a classification of networks according to geographical areas.
· LAN(Local Area Network)—Up to few Kilometers up to few 10s of Kilometers
· MAN(Metropolitan Area Network)—From few 10s of Kilometers up to 100 Kilometers covering a large metropolitan area
· WAN (Wide Area Network)—Any size network larger than MAN. WAN is usually a higher level network which connects many LANs together.
(b) Connection Topology
Topology means the way computers are attached to a network. Common topologies are Bus, Ring, Tree, Star, and Hybrid topologies.
· Bus topology—All stations are attached to a common medium called “Bus”. The attachment to a bus is usually by a hardware interfacing commonly known as a tap(or sometimes called as a “vampire tap” since the tap digs into the wires for connections). Original Ethernets(10Base5) use coaxial cables as the medium(bus).
· Ring topology—Stations are connected to form a ring. Each station acts as a repeater so that a signal will travel to next station and so on. IBM Token rings use the ring topology. Also, FDDI(Fiber Distributed Data Interface) uses fiber optic ring.
· Tree topology—It is a generalized bus topology in which many buses are joined at one point called “headend”. Cable TV uses tree topology.
· Star topology—All stations are connected to a central point. The device at the central point is usually called as a “hub”. An example is an Ethernet hub.
· Hybrid topology—Any network connected using more than one topology can be using a hybrid topology.
Bus Topology
Ring Topology
Star Topology
Tree Topology used in Cable TV
(from http://www.iec.org/online/tutorials/cable_mod/topic01.html?Next.x=29&Next.y=15
(c) Transmission Media
LANs can employ any available transmission media except some media that are inherently for WAN such as satellite, telephone lines, long-haul microwaves. Here are the most widely used media.
· UTP—Cheapest, most widely used for LANs, 100meter maximum, Category 3 and higher, speed can go up to 1 Giga bps
· FTP (Foil Twisted Pair)—Has a thin layer of foil wrap over twisted paris
· STP—Not widely used due to higher price
· Coaxial cable—Used in old Ethernets, mostly replaced by UTP
· Fiber optics—Getting more shares on LAN usage but still used to interconnect smaller LANs at the campus level
· Infrared—Used in a room level LANs
· Wireless LAN—Getting popular, used at a building level LAN
(d) Speed
LAN technologies have been evolving from early 1980s to date. They can be classified into several generations:
· 1st generation(1980 – 1990): 10 Mbps Ethernets, 4 or 16 Mbps Token Rings
· 2nd generation(1994): 100 Mbps Ethernets, FDDI, 100BaseVG
· 3rd generation(1998): 1000 Mbps Ethernets
· 4th generation(2002): 10 Gigabit Ethernets
(e) Medium Access Control Method
A medium that is used in a LAN can be either “shared” or “switched”. In a shared medium LAN, all stations must share the medium. Therefore, we need some kind of sharing mechanism. On a switched medium LAN, a central device switches packets to a destination.
LAN protocols and their implementations started with shared medium LANs first and then evolved to Switched medium LANs.
There are several methods to share the medium:
· Contention based—CSMA/CD (for Ethernet)
· Token based—Token Ring, FDDI
· Demand priority—100BaseVG
They will be described when 100 Mbps LANs are discussed.
6.2 LAN Standards
LAN standards are studied by all levels of standard organizations, national, regional, and international organizations. Among them, IEEE is the major player.
According to the web site of IEEE (www.ieee.org):
“The IEEE (Eye-triple-E) is a non-profit, technical professional association of more than 360,000 individual members in approximately 175 countries. The full name is the Institute of Electrical and Electronics Engineers, Inc., although the organization is most popularly known and referred to by the letters I-E-E-E.”
Before we look at the standard protocols for LANs, let’s first look at the list of IEEE Working groups.
802.1 High Level Interface (HILI) Working Group
802.2 Logical Link Control (LLC) Working Group
802.3 CSMA/CD Working Group
802.4 Token Bus Working Group
802.5 Token Ring Working Group
802.6 Metropolitan Area Network (MAN) Working Group
802.7 BroadBand Technical Adv. Group (BBTAG)
802.8 Fiber Optics Technical Adv. Group (FOTAG)
802.9 Integrated Services LAN (ISLAN) Working Group
802.10 Standard for Interoperable LAN Security (SILS) Working Group
802.11 Wireless LAN (WLAN) Working Group
802.12 Demand Priority
802.14 Cable-TV Based Broadband Communication Network Working Group
802.15 Wireless Personal Area Network (WPAN) Working Group
802.16 Broadband Wireless Access (BBWA) Working Group
802.17 Resilient Packet Ring (RPR)
802.18 Radio Regulatory Technical Advisory Group
802.19 Coexistence Technical Advisory Group
802.20 Mobile Wireless Access Working Group
As we can see from the above list, IEEE works on every facet of LAN technologies. IEEE has been leading in the LAN standards world. IEEE adopts a LAN standard as an “internal standard”, and IEEE recommends it to other standard organizations such as ANSI, ITU-T, and ISO and they adopt the standard. For example, IEEE 802.3 (CSMA/CD) is adopted as ISO 8802.3 by ISO.
6.2.1 Architecture of IEEE 802 LAN Protocols
IEEE 802 LAN protocols cover the bottom 2 layers of OSI Model. Their main concern is how computers or networking devices (hubs, switches, routers) are connected to a network. Since there are several choices in the physical medium (UTP, Coax, Fiber, Wireless) and also there are several choices in the medium sharing methods, IEEE 802 architecture divided the second layer (Data Link layer of OSI) into two sublayers: LLC (Logical Link Control) layer and MAC (Medium Access Control) layer.
The MAC layer is combined with the physical layer and each technology is numbered and standardized such as 802.3, 802.5, etc.
IEEE 802.X Protocol Architecture
IEEE 802.X Protocol Architecture
· LLC (Logical Link Control) layer: LLC is the upper part of the data link layer and defined in IEEE 802.2 . The LLC sublayer provides a uniform interface to the user of the data link layer’s service, usually the network layer. In other words, the functions that are common to all of the lower layer choices (802.3, 802.4, …) are grouped into one common sublayer. In short, LLC is responsible to provide the three functions that can a higher layer use: (a) Connection-oriented service (b) Acknowledged Connectionless service (c) Connectionless service.
These functions will be discussed in detail in a later chapter which deals with the Data Link layer.
· MAC (Medium Access Control Layer): The medium that devices are connected to can be either a shared medium or a switched medium. The mechanism for the access to the medium (to transmit and receive) is the MAC layer’s function. There are several ways to access a medium.
§ MAC algorithms for shared medium:
o Contention-based: CSMA/CD used for Ethernets and IEEE 802.3; CSMA/CA used for Wireless LAN (IEEE 802.11)
o Token-based: Token Ring (IEEE 802.5) and FDDI (ANSI X3T9.5)
o Demand Priority: Used for 100BaseVGAnyLan (IEEE 802.12)
Each will be described in the following sections.
· Physical layer: Physical layer provides services such as
§ Encoding and Decoding of signals (e.g. Ethernet’s Manchester code)
§ Synchronization
6.2.2 Ethernet and IEEE 802.3
Ethernet was invented at the Xerox Palo Alto Research Center by Dr. Robert M. Metcalf (founder of 3COM) in 1970’s. It was designed as a LAN technology using bus architecture on a coaxial cable.
Dr. Metcalf’s diagram for Ethernet
The formal specification for the Ethernet was published in 1980 by a multi-vendor consortium and called as DIX(DEC-Intel-Xerox) Ethernet specification. This specification resulted in an open, production-quality Ethernet system that operated at 10 Mbps on a coaxial cable. Ethernet version 1.0 and 2.0 were developed. This technology was studied by the LAN standards committee of IEEE (IEEE 802 project).
IEEE published the standard in 1985 with the formal title of “IEEE 802.3 Carrier Sense Multiple Access with Collision Detection (CSMA/CD) Access Method and Physical Layer Specifications” which is commonly known as Ethernet or IEEE 802.3 which is pronounced "eight oh two dot three."
The IEEE 802.x standards have been adopted by ANSI (ANSI X3.xxx) as U.S. national standard and also adopted by ISO (ISO 8802.x) as international standard. Ethernet (by DIX) has evolved to Version 2. Ethernet Version 2 and IEEE 802.3 can coexist.
IEEE 802.3 protocol covers MAC layer and Physical layer.
(a) CSMA/CD (IEEE 802.3 MAC layer): CSMA/CD (Carrier Sense Multiple Access with Collision Detection)
· In short, CSMA/CD is “Contention-based, listen-before-talk, binary-exponential backoff when collision detected.”
· Carrier Sense: “Listen Before Talk”—stations listen (sense) whether there is traffic on the shared medium.
· Multiple Access: More than one station can be connected on the shared medium and contend for the right to send.
· Collision Detection: Each station can detect when there has been a collision (transmissions (multiple access) from two or more stations that interfere with each other)
· Every station that collided uses BEB (Binary Exponential Backoff) algorithm to resolve the collision.
CSMA/CD works in the following way:
When a station needs to send, it first listens to the medium (senses whether the carrier is present or not). If there is a signal present (a station is already transmitting) then the station waits until it is done. If there is no signal (no station is transmitting), the station starts sending.
While a station is transmitting, it also listens to the medium. If a station detects a collision (by comparing the sending signal and received signal), the station stops the transmission immediately and start sending a short burst of “jamming signals” to ensure that every station can detect the collision.
After jamming, the collided stations will carry out the BEB (Binary Exponential Backoff) algorithm to resolve the situation.
BEB (Binary Exponential Backoff) algorithm:
Each station which experience a collision calculates a “Backoff Time” and waits for the Backoff Time and go back to the listen mode.
Backoff Time is calculated as follows:
Backoff Time = Slot time * Random number R
(Slot time is the round trip time for a signal traveling on the maximum length of
an Ethernet = 2500M. It is defined as 512 bit times for Ethernet networks
operating at 10 Mbps and 100 Mbps, and 4096 bit times for Gigabit Ethernet.)
0 <= R < 2k k= min(n, 10) n=
nth attempt
n=0 0 <= R < 1 ----> No wait---0th
attempt(=1st try)
n=1 0<= R < 2 -----> 0 or 1 wait
n=2 0<= R < 4 ----> 0, 1, 2, or 3 wait
...
n=10 0<= R < 1024
n=11 to n=15 0<= R < 1024
After 16th try, it is reported as an error.
(b) Physical Layer of Ethernet and IEEE 802.3
Ethernet has several options in physical layer depending upon the choice of wiring. They are: 10Base5, 10Base2, 10BaseT, 10BaseF, 1Base5, 10Broad36.
Let’s look at only the important ones among them.
The naming follows a convention:
The first number is for speed: e.g. 10 means 10Mbps
Base or Broad for “Baseband signaling” or “Broadband signaling”
The last number is for the maximum segment (a cut of a wire) length in hundred meters: e.g. 5 means 500 meters
· 10Base5: Original Ethernet using a thick coaxial cable
10Base5 Ethernet cable (RG-8)
This figure shows the wiring of workstations to the Ethernet coaxial cable.
Another diagram illustrates a more detailed connection:
A detailed tapping (physical connection to the coaxial cable) is shown below:
10Base5 uses 'thick' coaxial cable (RG-8 :Normally yellow in color). The minimum length between stations is 2.5m.
A cable is run in one long length making a 'Bus Topology'. Stations attach to it by way of inline N-type connections or a transceiver which is literally screwed into the cable (called a 'Vampire Tap'). A 15-pin AUI (Attachment Unit Interface) connection cables can run (maximum of 50m length) to the station.
Each segment (each cut of a cable) must be terminated with 50 ohm resistors and the shield should be grounded at one end only.
Ethernet’s 5-4-3 Rule:
5: An Ethernet can have Maximum of 5 segments of coaxial cable
4: Maximum number of repeaters is 4
3: Only 3 segments (out of maximum of 5) can have workstations attached.
Also, only 100 stations can be attached to each segment.
The following diagram illustrates the 5-4-3 rule.
5 sections of cables - 4 repeater - 3 sections with workstation connected
Therefore, a maximum span = 5 times 500meters = 2500 meters
· 10Base2: Also called as “ThinLan” or “Cheapernet”
The number 2 represents 185meters (maximum segment length). It uses RG-56 thin coaxial cables. Minimum spacing between stations is 0.5 meters and maximum of 30 stations can be attached to each cable segment. 5-4-3 rule applies to 10Base2 also
Thin Ethernet cables can be directly wired to workstations since they are more flexible and lighter than 10Base5 cables.
The following figure shows the comparison of wiring complexity between 10Base5 and 10Base2.
Figure (a) is for 10Base5 wiring, figure (b) is for 10Base2
· 10BaseT: IEEE standard for “Ethernet-over-twisted pair” came out in the fall of 1990. It is said, “LAN deployment exploded” meaning that with 10BaseT technology every building with category 3 telephone wiring can have Ethernets without rewiring the building. In previous Ethernets (10Base5 and 10Base2), new wires (coaxial cables) must be installed. Wiring in an existing building is very costly. 10BaseT solved this wiring problem.
An Ethernet hub
10BaseT connections using existing telephone wires
10BaseT wiring
· Most commercial building are wired according the EIA568B, "Commercial Building Telecommunications Cabling Standard”
· EIA568A specifies 4 pairs of wires for each connection and specifies categories of wires (category 1, 2 , up to 7)
· 10BaseT can use wires that are category 3 and above.
· Out of 4 pairs, only 2 pairs are used (one pair for transmitting and one pair for receiving).
· Each station’s wires are terminated with RJ45 connector
RJ45 connector
· Each station is connected to a centrally located device called as a “hub” as illustrated in the following figure.
10BaseT wiring
· 10BaseT also uses CSMA/CD protocol. The 10BaseT hub is a simple repeater, sometimes called as “multi-port repeater”. A hub receives signals from stations and it repeats to every port.
· Each station can send (using a pair of wires) and listen (using another pair) and compare the signals on the two pairs. If the signals on the two pairs are not same, then a collision is detected and the Binary Exponential Backoff algorithm starts.
RJ45 Connector (Pin 1 & 2 are for Tx(Transmit), Pin 3 & 6 are for Rx(Receive) from http://www.icc.com/NetworkAPP.htm
· 10BaseT requires category 3 or higher wires.
· Maximum segment length is 100 meters.
· 5-4-3 rule applies to 10BaseT: A hub counts as one repeater
· Maximum span of 10BaseT network is 500 meters.
· 10BaseT is Physical Star (stations are wired to a centrally located hub) and Logical Bus (Hub is functionally the same as a coaxial cable)
6.2.3 IEEE 802.5 (Token Ring)
Token rings are wired in a ring fashion, a station is connected to another station and that station is connected to another and so on and finally the connection comes all the way around.
When a ring is idling, a 3 byte long frame called “Token” is passed around. Each station in a ring is a one-bit-repeater (receives a bit and repeats it to downstream).
When a station wishes to send a frame, it waits until the token frame arrives. When the token arrives, it transmits a frame. The frame will reach the destination station but still travels on the ring. On Token ring, the sending station is responsible to purge the frame from the ring. In a slower speed ring (4 Mbps), a new token is release. In a higher speed ring (16Mbps), a new token is released as soon as the last bit is transmitted. This is known as ETR (Early Token Release). With ETR, there could be more than one frame being transmitted on a ring.
A typical Token ring operation is illustrated in the following diagram. The diagram shows the ring operation without ETR.
Token Ring Operation without ETR
Characteristics of Token Ring
· One bit delay for each station: Each station receives a bit at a time and repeats the bit downstream
· IBM defined 4 or 16 Mbps on STP (IBM Type 1 cable)
· 250 stations maximum on one ring
· Token Rings can have the following problem cases:
§ Lost token: A station holds the token and does not release it due to malfunction
§ Persistent busy token: A station begins transmitting and malfunctions. Therefore, the station will not remove the frame from the ring.
§ Garbled frame: The token frame is hit by noise and no station can recognize it
·
To solve the above problems, we need
a Monitoring station on a ring.
A station becomes the Monitor through an election process between stations and
monitors the ring all the time. When the monitor goes down (either turned off
or malfunctions), there will be another election
· Token rings have more sophisticated operations than Ethernets due to the above monitoring operations. Therefore, Token ring network cards have been more expensive due to the complexity.
6.2.4 Comparison of Ethernet and Token Ring
After IBM announced Token Ring in 1985, there have been debates about which LAN is better: Ethernet or Token Ring? Here are the comparisons of the two.
a) Ethernet
· Non-deterministic delay: The contention-based CSMA/CD does not guarantee a station a definite timing on the successful transmission of a frame. This aspect hinders Ethernet to transmit “isochronous (time-sensitive)” traffic such as voice and video on real time basis. These isochronous traffic need regular, periodic access to the medium.
· No Priority: On Ethernets, there is no provision on the priority of frames or stations. Therefore, we can not set a priority on frames.
· Collisions: When an Ethernet has heavy load (many stations transmit simultaneously), the effective transfer rate will deteriorate due to heavy collisions.
· Most widely available: Ethernet is the first technology in LAN field and there is accumulation of experience.
· CSMA/CD is simpler than Token Ring. Therefore, Ethernet cards are cheaper and easier to maintain than Token Ring cards.
· Ethernet provides adequate performance at light load.
· Adding or removing stations does not disturb the Ethernet LAN.
· Newer Ethernets (100 Mbps, Gigabit, 10Gigabit) are compatible with existing Ethernet
b) Token Ring
· Wiring can be either point-to-point or star wiring
· Predictable delay: Since token must be passed to next station, a station is guaranteed to get the token within certain time frame
· Utilization (efficiency = Time used for data transmission/ Total time) is high at heavy load: This is a reverse characteristic to Ethernet’s efficiency. Ethernet experiences more collision in heavy load, therefore the efficiency goes down quickly. Token ring’s efficiency goes up as the load increases.
· Delay at light load: Since a station must give up the token and wait for the token to come around again. At a light load (only few stations need to transmit), stations experience some delay (waiting time for the token).
· Adding/Deleting a station briefly disturb the ring operation.
Due to its complexity and its late introduction (Ethernet 1980, Token Ring 1985), Token Rings have never been very popular except in the IBM mainframe networks. The following figure shows the comparison of the number of installations for Ethernet and Token Ring.
From http://www.cisco.com/warp/public/cc/so/neso/ibso/ecampus/trh_bc.htm
(Business Case: Token Ring-to-Ethernet Migration)
From: http://www.iec.org/online/tutorials/opt_ethernet/topic02.html
