Network Topologies

Topology

Mathematical termRoughly interpreted as "geometry for curved surface"

Network Topologies

A network "topology" is the structure or organization of communications that links between hosts or devices on a network.

LAN topology A LAN is a shared medium that serves many DTEs( data

terminal equipment) located in close proximity such as in one building.

Three basic topologies associated with LANs: bus, ring, and star

WAN topology A WAN links networks that are geographically separated by

long distance through switches, routers, and/or bridges. Two topologies: mesh and tree

LAN Topologies

Bus topology: All hosts (DTEs) are connected to a common cable or medium.

LAN Topologies (cont’d)

Tree topology: Transmission medium is a branching cable with no closed loops.

Generalization of bus topology. Hub

Host

LAN Topologies (Cont’d)

Ring topology: Each device (DTE) is connected to another in sequence to form a "ring”

LAN Topologies (Cont’d)

Star Topology: All of the devices on the network are connected to a central "hub" or concentrator

Advantages and Disadvantages of Bus Topology

Advantages of bus topologies: Inexpensive to install (uses less cable) Easy to add new devices onto the bus or onto the

network

Disadvantages of bus topologies: Can be expensive to maintain and troubleshoot A naive user can easily "bring down" the entire bus Overall maximum length of the bus is limited (for

example, in a 10-Base-2 ethernet, 200m maximum from one terminator to the other along the cable)

Advantages and Disadvantages of Ring Topology

Advantages of ring topologies: Very predictable network performance May be slightly more secure than other topologies

Disadvantages of ring topologies: Expensive as compared to bus/star topologies

Hardware for ring topologies is less available and therefore more expensive

Many systems lack good support for networking in ring environments

Unique wiring requirements More complex networking and operational protocol

Advantages and Disadvantages of Star Topology

Advantages of star topologies: Each node has a dedicated connection to the network --

disconnecting a single node does not bring down the rest of the nodes on the network

Network and cable administration are centralized

Disadvantages of star topologies: More expensive to install -- require more cable and the

additional cost of a hub Maximum length of each spoke of the hub is limited to the

allowed maximum length of the medium (for example, on a 10-Base-T network using UTP cable, the maximum distance from the hub to a host is 100m)

Breakdown of the hub causes breakdown of the entire system

WAN Topologies

Mesh Topology: provides multiple paths between nodes or networks (N) usually implemented with switches and routers

N1

N2 N3 N4

N6

N5

WAN Topologies (Cont’d)

Tree Topology: A hierarchical architecture starts with header node and branches out to other nodes. Simpler to implement than mesh topology.

Data Link Layer

Specifies how two devices or hosts communicate with each other when they are connected to the same medium (e.g., connected via a common bus or a common hub).

Major functions of the layer: Flow control: prevents receiver’s buffer overflow Error detection: uses error-detecting code and

algorithm Error control: retransmits damaged frames upon

request or if no acknowledgement received from the receiver

Terminology

Subnet The devices which are linked together by a common

medium are collectively known as a subnet. Frame

Data are sent in blocks called frames. A frame, in addition to data, contains some header information such as source and destination addresses, control data bits, error-checking bits, etc. Frame size, the number of bits, depends on the underlying protocol. Actual frame format depends on the protocol.

Data Link Layer: Sharing Medium In A LAN

Shared medium used for all transmissions

Only one station transmits at any time Stations "take turns" using medium Media Access Control (MAC) policy

ensures fairness (MAC protocol)

Media Access Control Protocols

Media Access Control

determines the rules about when hosts on a subnet are allowed to transmit data onto the physical medium

Two broad control schemes: Centralized control

- greater control through priorities, overrides, and guaranteed capacity

- simple- but creates a bottleneck and a single point of failure

Distributed control

MAC Control Techniques

Round RobinEach station takes turn to transmitMay be centralized (polling) or distributed

(token passing)Efficient when many stations transmitHigh overhead when only few stations transmit

ReservationTransmitting station reserves slots (stream

traffic)May be centralized or distributed

MAC Control Techniques (cont’d)

Contention- Appropriate for bursty traffic- Distributed by nature- Simple to implement- Efficient for light to moderate load- Performance tends to collapse under heavy load - Examples:

- ALOHA- Slotted ALOHA- CSMA - CSMA/CD

ALOHA

Developed for packet radio networks Transmits whenever a station has a frame to send Wait and listens for an acknowledgement Wait time = maximum possible roundtrip

propagation delay plus a small fixed time increment Resends the frame if no acknowledgement is

received. Gives up after many repeated, failed transmissions Simple but poor utilization. Maximum utilization is

only about 18%

Slotted Aloha

Similar to ALOHA but stations are allowed to transmit during a time slot

Channel is organized into uniform time slots Size of the time slot = Frame transmission

time Some central clock synchronizes all stations If a frame collides with other one, it collides

completely Maximum utilization is about 37%

An Important Note

Both ALOHA and slotted ALOHA exhibit poor utilization and fail to take advantage of the fact that propagation delay is usually very small compared to frame transmission time for both packet radio and LANs.

CSMA (Carrier Sense Multiple Access)

Advantageous over slotted Aloha when propagation time is small compared to frame transmission time.

Listens if another transmission is in progress (carrier sense)

Transmits if medium is idle Wait for acknowledgement Wait time = maximum roundtrip propagation time +

medium access time for the receiver Retransmits if no acknowledgement is received Disadvantage: medium remains unusable for the

duration of transmission of damaged frames after collision

CSMA/CD Carrier Sense Multiple Access/Collision Detection)

A protocol for Ethernet

1. If medium is idle, transmit and go to step 3; otherwise, go to step 2.

2. If medium is busy, continue to listen; transmit immediately if idle.

3. If collision detected, transmit a brief jamming signal to inform other stations.

4. Wait a random amount of time after transmitting jamming signal (backoff), then go to step 1.

Note:With CSMA/CD, the amount of wasted capacity is reduced to the time it takes to detect a collision.

Collision Detection Mechanisms

Baseband Ethernet: Higher voltage swings than those produced by a single transmitter are detected. Cable length is limited to 500 meters.

Broadband Ethernet: RF Carrier is detected. Bit-by-bit comparison is done between transmitted and received data.

Twisted-pair star topology: If a hub detects presence of more than one input signal at its ports, it assumes a collision and sends out collision presence signal.

Backoff After Collision (Wait Time Calculation)

When collision occurs Wait random time t1, 0 < t1 < d

Use CSMA and try again If second collision occurs

Wait random time t2, 0 < t2 < 2d

Double range for each successive collision Called exponential backoff

CSMA/CA

Used on wireless networks Both sides send small message followed by data

transmission "X is about to send to Y” "Y is about to receive from X” Data frame sent from X to Y Purpose: inform all stations in range of X or Y before

transmission Known as Collision Avoidance (CA)

Example Bus Network: Ethernet

Most popular LAN Widely used IEEE standard 802.3 Several generations

Same frame formatDifferent data ratesDifferent wiring schemes

Identifying A Destination

All stations on shared-media LAN receive all transmissions

To allow sender to specify destination Each station assigned unique number Known as station’s address Each frame contains address of intended

recipient

Ethernet Addressing

Standardized by IEEE Each station is assigned with a unique 48-bit

address Address is assigned when network interface

card is (NIC) manufactured Each address is a physical address

Ethernet Address Recognition

Each frame contains destination address All stations receive a transmission Station discards any frame addressed to another

station Important: interface hardware, not software,

checks address. Does not utilize CPU to check address

Possible Ways to Direct Frames

Frames can be sent to: Single destination (unicast) All stations on network (broadcast) Subset of stations (multicast)

Some feature of destination address is used to distinguish type (unicast, broadcast, or multicast)

Broadcast on Ethernet

All 1s address specifies broadcast Sender

Places broadcast address in frame Transmits one copy on shared network All stations receive copy

Receiver always accepts frame that contains Station's unicast address The broadcast address

Multicast on Ethernet

Half of addresses reserved for multicast Network interface card

Always accepts unicast and broadcast Can accept zero or more multicast addresses

Software Determines multicast address to accept Informs network interface card

Promiscuous Mode

Designed to testing/debugging Allows interface to accept all packets Available on most interface hardware

Network Analyzer

Device used for testing and maintenance Listens in promiscuous mode Produces

Summaries (e.g., % of broadcast frames)Specific items (e.g., frames from a given

address)

Identifying Frame Contents

Integer type field tells recipient the type of data being carried

Two possibilities Self-identifying or explicit type (hardware

records type) Implicit type (application sending data must

handle type)

Conceptual Frame Format

Header Contains address and type information Layout fixed

Payload Contains data being sent

PayloadHeader

Example Ethernet Frame Format

8 6 6 2 46 - 1500 4

Preamble

Dest. Addr.

Src. Addr. Data In Frame

CRCFrameType

Preamble: Alternating 1s and 0s. Used by receiver synchronization

Example Frame Types

ValueMeaning0000-05DC0800080509000BAD1000-100F6004655980058008801480358038805C809B80C4-80C580D580FF-81038137-8138818DFFFFReserved for use with IEEE LLC/SNAPInternet IP Version 4 CCITT X.25Ungermann-Bass Corporation network debuggerBanyan Systems Corporation VINESBerkely UNIX Trailer encapsulationDigital Equipment Corporation LATFrame RelayHewlett Packard Corporation network probeAT&T Corporation Silicon Graphics Corporation network gamesInternet Reverse ARPDigital Equipment Corporation LANBridgeStanford University V KernelApple Computer Corporation AppleTalk Banyan Systems CorporationIBM Corporation SNAWellfleet CommunicationsNovell Corporation IPXMotorola CorporationReserved

When Network Hardware Does Not include Types

Sending and receiving computers must agree

To send one type of data To put type information in first few octets of

payload

Most systems need type information

Handling Frames of Many Types

Network interface hardwareReceives copy of each transmitted frameExamines address and either discards or

acceptsPasses accepted frame to system software

Network device softwareExamines frame typePasses frame to correct software module

Network Analyzer

Device used for testing and maintenance Listens in promiscuous mode Produces

Summaries (e.g., % of broadcast frames)Specific items (e.g., frames from a given

address)

Note: Check web for free network analyzer/packet sniffer.

10Base2 Ethernet Wiring (Thinnet)

Use coax cables (10Base2), NICs, BNC connectors, terminators

Twisted Pair (10Base-T) Ethernet Wiring

Use 10Base-T wire, hubs, NICs, and RJ-45 connectors

HubTwisted pair wiring

RJ-45 connectors

IEEE 802.3 10-Mbps Ethernet Specifications

Notation: <Mbps><signaling><length in 100m>

10BASE5- 50-ohm coax cable- Topology: bus- Maximum segment length: 500 meters- Nodes per segment: 100- Data rate: 10Mbps- 4 repeaters maximum (2.5 km)

IEEE 802.3 10-Mbps Ethernet Specifications (cont’d)

10BASE2- 50-ohm coax cable (thinner brand)- Maximum Segment length: 185 meters- Nodes per segment: 30- Topology: bus- Data rate: 10Mbps

IEEE 802.3 10-Mbps Ethernet Specifications (cont’d)

10BASE-T (Twisted pair)- Maximum segment length: 100 meters- Topology: star- Data rate: 10Mbps

10BROAD36- 75-ohm CATV coaxial cable- Maximum individual segment length is 1800

meters.- Maximum end-to-end span 3600 meters.

IEEE 802.3 100-Mbps Fast Ethernet Specifications

100BASE-TX Data rate: 100Mbps 2 Shielded twisted pair(STP) or high-quality Category 5

unshielded twisted pair(UTP) Maximum segment length: 100 meters Network span: 200 meters

100BASE-FX 2 Optical fibers Data rate: 100 Mbps Maximum segment length: 200 meters Network span: 400 meters

Token Ring

Token Ring

Most commonly used MAC protocol for rings IEEE 802.5 standard Token - a small frame

Token passing mechanism

A station seizes a token by changing one bit Changed token is a start-of-frame sequence Transmitted frame is absorbed by the

transmitting station after a round-trip The station will insert a new token

- at the end of transmission and- at the detection of leading edge of transmitted

frame after circulation

Advantages and Disadvantages of Token Passing Protocol

Inefficient at light load Efficient and fair at heavy load Principle advantage: Flexible and simple scheme Principle disadvantage: Token maintenance

Token Maintenance Problems

Token passing protocols are much more complex than contention-based protocols. For example, the protocol must deal with: What happens when a token gets lost What happens if two or more tokens show up

on the subnet

IEEE 802.5 Frame Format

SD AC FC DA SA Data unit FCS ED FS

1 1 1 6 6 >0 4 1 1

SD: starting delimiter SA: source addressAC: access control FCS: frame check sequenceFC: frame control ED: ending delimiterDA: destination address FS: frame status

Bytes

SD AC ED

1 1 1

A. General frame format

B. Token frame format

IEEE 802.5 Frame Format (cont’d)

Starting delimiter (SD). Indicates start of frame. Access control (AC). Used to identify data or token frame and

indicate priorities. Frame Control (FC). Indicates whether this is an LLC data frame End delimiter (ED). Error detection bit is set by a repeater. Frame status (FS). Contains the address recognized (A) and

frame-copied (C) bits. Set by the receiver and used by sender for checking.

A = 0, C = 0 meaning destination does not exist A = 1, C = 0 meaning frame not copied A = 1, C = 1 meaning frame received

Uses some priority scheme to transmit high priority frames.

Comparison of contention-based and token-passing protocols

Contention-basedToken-passingEqual access to all hosts onthe network (no "master"host)Equal access to all hosts on thenetwork (no "master" host)Access time can beunpredictableAccess time can be predicted easilyMaximum wait for sending amessage is unpredictableMaximum wait is predictableA host can begin transmittingimmediately if the network isquietA host must wait for the token before it can transmit

Comparison of contention-based and token-passing protocols (cont’d)

Contention-basedToken-passingAdding more "quiet"nodes doesn't affectnetwork efficiencyAdding more hosts adds delayaffect network efficiencybecause of increased number ofstops for the tokenSimpleComplexCommon protocol name:EthernetCommon protocol: IBM TokenRingCan have degradedperformance in heavilycongested subnetsHas better performance inheavily congested subnetsTypically less expensiveTypically more expensive

Ring LANs

Consists of repeaters Unidirectional transmission Single closed path Bit by bit transmission from one repeater to the next Each repeater regenerates and retransmits each bit Repeaters perform the data insertion and reception

functions Packet is usually removed by transmitting repeater

after one trip around Medium - Twisted pair, baseband coax, and fiber

optic cable

Ring LAN Repeaters: States

States of a repeater: listen state transmit state bypass states

1-bit delay

To station Tostation

Fromstation

A. Listen state B. Transmit state C. Bypass state

Ring Problems

Timing jitter - timing jitter accumulates as the signal travels around the ring. It limits the number of repeaters in a ring

Link failure/repeater failure disables the entire network

Installation of a new repeater is difficult and disruptive

Star-ring architecture eliminates some of the above problems partly.

Star Topology: Example Networks

ARCnet Developed by Datapoint Corporation in 1970 Has become a de facto microcomputer standard Speeds: 2.5Mbps, 20MBps Uses active and passive hubs Medium: twisted-pair wires or coaxial cables or fiber optic

cables for higher speed implementation

StarLAN Developed by AT&T Corporation Speeds: 1Mbps, 10Mbps Medium: twisted pair wires IEEE 802.3 standard

ARCNet Example

Active Hub

Passive Hub

Terminology: Hubs

Active hub Used in an ARCnet LAN that provides signal

regeneration allows nodes to be located up to 2000 feet from the hub

Passive Hub Used in an ARCnet LAN that does not provide signal

regeneration Nodes can be located no farther than 100 feet from the

hub

LAN Summary

EthernetIBM's Token RingARCnetStarLANSpeed10 or 100 Mbps4, 16, or 100Mbps2.5 or 20 Mbps1 Mbps or 10MbpsMediumTwisted-pairwires, coaxialcable, or fiberoptic cableTwisted-pair wiresTwisted-pairwires or coaxialcableTwisted-pairwireDistance500 meters forthick cable andmaximum span2500 meters with4 repeaters, 185meters for thinnetcable andmaximum span925 meters366 meters for themain ring; can beextended to 750meters withrepeaters and to4000 meters withfiber optic cable6110 meters;maximumdistance betweenactive hubs is 62meters andbetween passivehubs is 31 meters500 meters

LAN Summary (Contd)

EthernetIBM's TokenRingARCnetStarLANNumber ofStations100 perthick cablesegmet, 30per thinnetsegment260255Early StarLANs setlimit at 50StandardsIEEE 802.3IEEE 802.5De FactoIEEE 802.3 1Base5CostLowHighLowLow