special course in computer science: local networks

114
Lecture 5 4.4.2012 Special Course in Computer Science: Local Networks

Upload: others

Post on 02-Nov-2021

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Special Course in Computer Science: Local Networks

Lecture 5 4.4.2012

Special Course in Computer Science: Local Networks

Page 2: Special Course in Computer Science: Local Networks

Roadmap of the Course So far Basic telecom concepts General study of LANs LAN topologies Flow and error control Medium access methods Logical link control

Today Ethernet

Page 3: Special Course in Computer Science: Local Networks

Ethernet

Page 4: Special Course in Computer Science: Local Networks

Ethernet IEEE 802.3 Most widely used LAN protocol (wired networks) More than 35 years of use (1976) Started at 10 Mbps (1978) Nowadays 100Mbps – 10Gbps Classic Ethernet or Switched Ethernet

Page 5: Special Course in Computer Science: Local Networks

Architecture of the original Ethernet

Page 6: Special Course in Computer Science: Local Networks

CSMA / CD Stations abort transmission as soon as they detect a

collision => saves time and bandwidth Model At t0 some station finished transmission Any other station may start transmitting => collisions may

occur After a detected collision, stations wait a random time and

start sending again, sensing the channel first

Page 7: Special Course in Computer Science: Local Networks

CSMA/CD 2 Minimum time to detect a collision Time needed for a signal to propagate from a station to

another Worst case scenario τ – the time needed for a signal to propagate between

farthest stations It takes 2τ until a station is certain it seized the channel (if

transmits for 2τ without hearing a collision, then it is sure) A sending station must continually monitor the

channel To listen for noise bursts indicating a collision

Page 8: Special Course in Computer Science: Local Networks

Collision detection

Page 9: Special Course in Computer Science: Local Networks

Ethernet MAC

Page 10: Special Course in Computer Science: Local Networks

1-persistent CSMA/CD Station with frame to transmit Back-off factor k set to 0 Station monitors medium, waiting for idle channel If channel free: station waits IFG time and sends frame

IFG: time needed to send 96 bits (= 9.6 microsecs for 10Mbps) If channel busy, station continues to monitor medium Channel monitored during transmission If collision detected in the first 512 bits, transmission stops and

jam signal (32 bits) sent Station increments back-off factor k If k < 10, station waits another back-off time before sensing channel again Back-off-time = r × slot-time

r – random integer between 0 and 2k -1 k incremented at each collision, max 10

After that station should give up sending frame

Page 11: Special Course in Computer Science: Local Networks

Ethernet layers

Page 12: Special Course in Computer Science: Local Networks

MAC sub-layer in Ethernet Functions Responsible for the CSMA/CD operation Frames data received from upper layers (LLC) and

passes them on to lower layers (PLS)

Page 13: Special Course in Computer Science: Local Networks

Fields in Ethernet MAC frame Preamble: 7 bytes (56bits), each 10101010 Alert the receiver to the coming frame Enable synchronization in input timing Added at the PhL => not theoretically part of the

frame Start of Frame Delimiter: 1 byte: 10101011 Signals the beginning of a frame Last synchronization chance “11” bits say the destination address follows

Page 14: Special Course in Computer Science: Local Networks

Fields in Ethernet MAC frame 2

Destination address: 6 bytes Physical address of the destination station(s)

Source address: 6 bytes Physical address of the source station

Length/type: 2 bytes If value < 1518 then length, else PDU type

Data: minimum 46 bytes, maximum 1500 CRC: 4 bytes of error detection info (CRC-32)

Page 15: Special Course in Computer Science: Local Networks

Addressing Each Ethernet device has a NIC, which provides a 6-

byte physical address In hexadecimal notation

Page 16: Special Course in Computer Science: Local Networks

Address transmission

Page 17: Special Course in Computer Science: Local Networks

Unicast and multicast addresses Source address always unicast Destination addresses can be unicast or multicast or

broadcast Unicast: one recipient only First transmitted bit 0

Multicast: several recipients (group of addresses) First transmitted bit 1

Broadcast: recipients are all stations 48 1-bits

Page 18: Special Course in Computer Science: Local Networks

Unicast and multicast illustration

Page 19: Special Course in Computer Science: Local Networks

Globally and locally unique addresses Addresses should be unique Globally/locally

Globally unique Unique for all LANs Have to follow some IEEE rules About 7 × 1013 such addresses Second transmitted bit is 0

Locally unique Second transmitted bit is 1 Administrators should ensure the addresses in a LAN are

unique

Page 20: Special Course in Computer Science: Local Networks

Illustration of unique addresses

Page 21: Special Course in Computer Science: Local Networks

Minimum frame length

Page 22: Special Course in Computer Science: Local Networks

Main reason for minimum length To prevent station from completing the transmission

of a short frame before first bit has even reached the far end of the cable (where it may collide with another frame) If it completes transmission without hearing jam during

transmission, then it assumes all went well To be able to detect collision, a frame has to have the

length of at least 2τ 802.3 spec: 10Mbps, max length of 2500 m, 4 repeaters: round

trip time (incl. repeaters): 50 microsecs => shortest frame 500 bits, for safety 512 bits => 64 bytes => 46 bytes for data

Page 23: Special Course in Computer Science: Local Networks

Collision detection (again)

Page 24: Special Course in Computer Science: Local Networks

How long to listen for collisions? Slot time = round trip time + jam signal time 512 bits => in 10Mbps it is 51.2 microsecs If station has sent 512 bits and did not hear any

collision, it can assume no collision Signal reaches end of cable in less than half slot

time Collision can only occur during the first half of the

slot time and if it has, it can be sensed by the sender during the slot time

=> sender should listen for collision only during the first 512 bits

Page 25: Special Course in Computer Science: Local Networks

Maximum network length Maxlength = propagation speed × SlotTime/2 Maxlength = 2 × 108 × 51.2 × 10-6 / 2 = 5120 m Delays Repeaters, interfaces, time for jam signal

Max length for tradition Ethernet: 2500 m (48% of

the above theoretical calculation)

Page 26: Special Course in Computer Science: Local Networks

Maximum frame length

Page 27: Special Course in Computer Science: Local Networks

Physical layer

Page 28: Special Course in Computer Science: Local Networks

Sub-layers in PhL Physical Layer Signaling (PLS) Common for all implementations Encodes and decodes data at 10Mbps Uses Manchester encoding

Attachment Unit Interface (AUI) May be present or not in implementations Specification defining the interfaces between PLS and MAU Intended to allow for a medium-independent interface

between PLS and MAU Designed for thick Ethernet – the first

Page 29: Special Course in Computer Science: Local Networks

PLS

Page 30: Special Course in Computer Science: Local Networks

AUI

Page 31: Special Course in Computer Science: Local Networks

4 types of AUI Signals Each carried over a TP cable Transmit data Carries data from NIC to transceiver Manchester encoding used

Receive data Carries data from transceiver to NIC Manchester encoding used

Collision presence Sent from transceiver to NIC if there is collision on the line

Power Transceiver power provided by NIC (12V)

Page 32: Special Course in Computer Science: Local Networks

AUI Signals Illustration

Page 33: Special Course in Computer Science: Local Networks

AUI connector

Page 34: Special Course in Computer Science: Local Networks

AUI cable

Page 35: Special Course in Computer Science: Local Networks

Transceiver Both transmitter and receiver Also known as MAU (medium attachment unit) Medium dependent Creates the appropriate signal for each particular

medium There is a MAU for each medium type used in

10Mbps Ethernet Can be external or internal

Page 36: Special Course in Computer Science: Local Networks

Transceiver functions

Page 37: Special Course in Computer Science: Local Networks

MDI Medium dependent interface Needed to connect the transceiver to the medium Piece of hardware

Page 38: Special Course in Computer Science: Local Networks

Ethernet implementations

Page 39: Special Course in Computer Science: Local Networks

Ethernet cabling 10 10 Mbps Base baseband signaling 5 how many 100-meter units (can support

signals of up to 500 meters)

Page 40: Special Course in Computer Science: Local Networks

10Base5 (thick Ethernet) First Ethernet specification Bus topology External transceiver connected via a tap to a thick

coaxial cable Maximum 5 segments, each 500 m, 100

stations/segment

Page 41: Special Course in Computer Science: Local Networks

10Base2 (thin Ethernet) Second cable historically Much cheaper and easier to install, thinner Bus topology with internal transceiver, or P2P connection via external transceiver Maximum 5 segments, each 185 meters, 30

stations/segment

Page 42: Special Course in Computer Science: Local Networks

10Base-T (twisted pair Ethernet) Developed as the previous cables had problems

with detecting cable breaks Star topology using a hub Either external or internal transceiver Segments of 100 m, 2 stations/segment Adding/removing stations is simple, easy

maintenance Became quickly dominant

Page 43: Special Course in Computer Science: Local Networks

10Base-F(L) (fiber link Ethernet) Cable is fiber optic Star topology with hub, external fiber-optic

MAU (transceiver) MAU connected to hub by 2 fiber optics They are unidirectional

2000-meter segments More expensive alternative, excellent noise

immunity “The” solution for between-building cabling

Page 44: Special Course in Computer Science: Local Networks

Ethernet evolution ideas Bridged Ethernet Separated collision domains

Switched Ethernet Full-duplex Ethernet MAC control Pause packet

Page 45: Special Course in Computer Science: Local Networks

Bridges Devices controlling traffic management Divide a network into segments and relay frames

between separate segments of same type Filter traffic Effects on an Ethernet LAN Raising bandwidth Separating collision domains

Page 46: Special Course in Computer Science: Local Networks

Bridge illustration

Page 47: Special Course in Computer Science: Local Networks

Raising bandwidth 2 stations each with

heavy traffic share the medium in an 10Mbps Ethernet

In average each gets 5 Mbps bandwidth

Bridges separate net into segments Independent segments,

bandwidth-wise

Page 48: Special Course in Computer Science: Local Networks

Separate collision domains

Page 49: Special Course in Computer Science: Local Networks

Switched Ethernet Extends the bridged Ethernet idea (layer-two) Switch: N-port bridge with additional

mechanisms for faster handling N stations, N segments Bandwidth is shared between each station and the

switch Collision domain divided into N domains

Page 50: Special Course in Computer Science: Local Networks

Switched LAN illustration

Page 51: Special Course in Computer Science: Local Networks

Full-duplex Ethernet Half-duplex Ethernet 10Base5 and 10Base2 A station can either send/receive, but not at the

same time Due to CSMA/CD principle

Full-duplex Ethernet Capacity of each domain increases to 20Mbps

Page 52: Special Course in Computer Science: Local Networks

Full-duplex switched Ethernet

Page 53: Special Course in Computer Science: Local Networks

No more CSMA/CD Possible in full-duplex Ethernet Each station connected to the switch Via 2 separate links Each link is a point-to-point dedicated path Stations/switch can send/receive without worrying

about collisions No more need for CS or CD These functions are turned off

Page 54: Special Course in Computer Science: Local Networks

MAC control motivation Classical Ethernet designed connectionless At MAC sublayer No explicit flow/error control, no ACK/NAK Error sources Frame corrupted during transmission, detected with CRC Frame lost due to collision, detected with CSMA/CD

Full-duplex switched Ethernet, error sources Frame corrupted, detected with CRC Frame lost due to switch buffer being full => undetected by

sender !!

Page 55: Special Course in Computer Science: Local Networks

New sublayer In full-duplex switched Ethernet: there is need

for explicit ACK An extra sublayer is added, optionally

Page 56: Special Course in Computer Science: Local Networks

How does it work MAC control sublayer inserts special control

packets between data packets Encapsulated in MAC frames in the normal way Minimum length required for these frames (<= 46

bytes)

Page 57: Special Course in Computer Science: Local Networks

Encapsulation of a MAC control frame into a MAC frame

Page 58: Special Course in Computer Science: Local Networks

Fields of the MAC frame carrying a MAC control frame

Destination address Device at the other link end: 01-08-C2-00-00-01 This is a multicast address Blocked by all bridges/switches => does not pass link boundary Recognized by stations having MAC control, ignored by others

Source address Type/length: fixed, 0x8808 Payload: MAC control packet, 46 bytes FCS: frame control sequence (4-byte CRC)

Page 59: Special Course in Computer Science: Local Networks

The PAUSE packet The only MAC control packet defined Used for slowing down the flow between devices Not designed to solve long, continuous congestion Stop-start flow control

A device sends a PAUSE packet to ask the other device to stop sending frames for a period of time

Station receiving PAUSE starts a timer and stops sending data packets (it can still send its own PAUSE packets)

When timer expires, station resumes sending data Latest PAUSE packet the most significant Period of time is multiplied by time slot

Page 60: Special Course in Computer Science: Local Networks

Fast Ethernet, IEEE 802.3u

Page 61: Special Course in Computer Science: Local Networks

Fast Ethernet MAC Sublayer The idea of the 10-to-100-Mbps evolution: keep the

MAC sublayer untouched Same MAC method: CSMA/CD (or none needed, for full

duplex!) Same frame format Same maximum and minimum length Same addressing

Changes: slot time (in seconds) and maximum network length (collision domain)

Page 62: Special Course in Computer Science: Local Networks

Fast Ethernet slot time and network length

How many bits in a slot? still 512 To avoid changes in minimum frame length

Bit interval one tenth shorter (10 ns) => slot time = 512 × 10-8 = 5.12 μs

Collisions detected 10 times earlier Max network length: theoretically 512 m, in practice

250 m

Page 63: Special Course in Computer Science: Local Networks

Auto-negotiation New feature of Fast Ethernet Allows two devices to negotiate the mode and

data rate Designed for Allowing incompatible devices to connect to each

other Allowing devices to have multiple capabilities Allowing stations to check hubs capabilities

Page 64: Special Course in Computer Science: Local Networks

Auto-negotiation 2 Permissible between two hubs or a hub and a

station at a link’s ends (p2p link) Concerns only that link, not the network

Can occur only during link initialization Uses special frame format and signaling Each device at the end of the link advertises

its capabilities to the other Decision based on common capabilities Hierarchy of common capabilities defined

Page 65: Special Course in Computer Science: Local Networks

Auto-negotiation example

Page 66: Special Course in Computer Science: Local Networks

Auto-negotiation message format Selector field: 5-bit field defining the LAN type Ethernet type: 10000

Technology ability field: 8-bit field, each bit defining support for a different technology A device advertising its capabilities sets one/more of

these bits

Page 67: Special Course in Computer Science: Local Networks

Technology ability fields

Bit Supported Technology

0 10Base-T

1 10Base-T full duplex

2 100Base-TX

3 100Base-TX full duplex

4 100Base-T4

5 Pause operation

6 Reserved

7 Reserved

Page 68: Special Course in Computer Science: Local Networks

More on fields Remote fault field 1-bit field, when set, signals a fault has occurred

Acknowledge field 1-bit field, when set, signals a message was successfully

received Next page field 1-bit field, when set, signals another message is coming Next message can define the error reason

Page 69: Special Course in Computer Science: Local Networks

Fast Ethernet physical layer

Page 70: Special Course in Computer Science: Local Networks

Reconciliation Common to all implementations Replaces the PLS sublayer in 10Mbps Ethernet Encoding and decoding are done in the PHY sublayer

(transceiver) Encoding in Fast Ethernet is medium-dependent

Responsible for passing 4-bit format data to MII

Page 71: Special Course in Computer Science: Local Networks

MII (medium independent interface) Replaces AUI, improved interface Features Can work with both 10 and 100 Mbps data rate Parallel data path, 4 bits at the time, between

reconciliation and PHY sublayers Management functions added

Page 72: Special Course in Computer Science: Local Networks

MII illustration

Page 73: Special Course in Computer Science: Local Networks

MII signals Transmit data 4 wires: called TX data, 4 bits at a time Another wire: TX-clock, used to synchronize PHY

and reconciliation sublayers 25 Mhz

Another wire: TX-enable, used to send an enable signal to inform PHY that data is sent

Another wire: TX-error, used by a repeater when received data is in error; stations do not use this wire

Page 74: Special Course in Computer Science: Local Networks

MII signals 2 Receive data 4 wires: RX data, 4 bits at a time Another wire: RX-clock, used to synchronize PHY

and reconciliation sublayers 25 Mhz

Another wire: RX-valid, used to show that the data received from the medium is valid

Another wire: RX-error, used when data received from medium is in error

Page 75: Special Course in Computer Science: Local Networks

MII Signals 3 Control signals (2) Carrier sense: shows that the carrier was sensed Collision detect: shows that a collision was detected Used only with half-duplex operations

Management signals Power

Page 76: Special Course in Computer Science: Local Networks

MII Signals illustration

Page 77: Special Course in Computer Science: Local Networks

MII Connector

Page 78: Special Course in Computer Science: Local Networks

MII Cable

Page 79: Special Course in Computer Science: Local Networks

PHY and MDI PHY sublayer (transceiver) Encoding and decoding (function moved from PLS

layer to PHY layer) External or internal

MDI sublayer Connects the transceiver to the medium Hardware piece, implementation specific

Page 80: Special Course in Computer Science: Local Networks

Fast Ethernet implementation

Page 81: Special Course in Computer Science: Local Networks

100Base-TX 2 twisted-pair cables (5-UTP) One to hub, one from hub

Physical star topology Logical topology Star: for half-duplex mode (CSMA/CD needed) Bus: for full-duplex mode

Transceiver: either external or internal

Page 82: Special Course in Computer Science: Local Networks

100Base-TX illustration

Page 83: Special Course in Computer Science: Local Networks

Transceiver In Fast Ethernet responsible for Transmitting Receiving Detecting collisions Encoding/decoding data

Encoding/decoding Block encoding: 4B/5B Method: MLT-3 (three levels, multiline transmission)

Page 84: Special Course in Computer Science: Local Networks

Encoding/decoding in 100Base-TX

Page 85: Special Course in Computer Science: Local Networks

MLT-3 MLT-3 similar to NRZ-I, but with 3 signal levels Signal transition from one level to the next

signals a 1 bit No transition at beginning of 0 bit

Page 86: Special Course in Computer Science: Local Networks

100Base-FX 2 fiber optic cables in physical star topology Logical topology Star: half-duplex mode Bus: full-duplex mode

Transceiver: either external or internal

Page 87: Special Course in Computer Science: Local Networks

100Base-FX illustration

Page 88: Special Course in Computer Science: Local Networks

100Base-FX encoding/decoding

Page 89: Special Course in Computer Science: Local Networks

NRZ-I encoding

Page 90: Special Course in Computer Science: Local Networks

100Base-T4 Most buildings already wired with 3-UTP =>

100Base-TX not cost efficient 100Base-T4 uses 4 pairs of 3-UTP (or higher) Encoding and decoding more complex 3-level encoding used : 8B/6T (eight binary/six

ternary)

Page 91: Special Course in Computer Science: Local Networks

8B/6T Each block of 8-bit data encoded as units of ternary

signals (+1, -1, 0 V) 8-bit codes: 256=28

6-symbol ternary codes: 729=36

Encoding designed to keep synchronization and transparency

Page 92: Special Course in Computer Science: Local Networks

Transmission with four wires

Page 93: Special Course in Computer Science: Local Networks

Gigabit Ethernet 1Gbps=1000Mbps data rate, IEEE 802.3z

Page 94: Special Course in Computer Science: Local Networks

Gigabit Ethernet: MAC sublayer Goal: to keep MAC sublayer untouched At 1Gbps some changes were needed in the

access method

Page 95: Special Course in Computer Science: Local Networks

Half-duplex approaches Use CSMA/CD When a hub is used The hub does not buffer incoming frames Collisions are possible

Maximum length in network dependent on minimum frame size

3 methods defined Traditional Carrier extension Frame bursting

Page 96: Special Course in Computer Science: Local Networks

Traditional half-duplex How many bits in a slot? still 512 To avoid changes in minimum frame length

Bit interval one hundredth shorter (1 ns) => slot time = 512 × 10-9 = 0.512 μs

Collisions detected 100 times earlier Max network length: theoretically 51,2 m, in

practice 25m Length not suitable

Page 97: Special Course in Computer Science: Local Networks

Carrier extension half-duplex Extension in frame size => longer network Minimum length defined as 512 bytes (8 times

longer) A station is forced to add extra padding to any

frame smaller than 4096 bits Maximum network length: 200 m (8 times

longer) 100 m from hub to station

Padding added by sending hardware, removed by receiving hardware

Page 98: Special Course in Computer Science: Local Networks

Carrier extension illustration

Page 99: Special Course in Computer Science: Local Networks

Frame bursting Carrier extension very inefficient Different approach: frames are sent together, as

one long frame Padding added between the frames so that the

channel is not idle Non-transmitting stations are “fooled” into

considering a large frame is transmitted

Page 100: Special Course in Computer Science: Local Networks

Frame bursting 2 First frame as in carrier extension, max length

determined by this frame, if it goes through If no collision, station sends rest of the frames,

i.e. the burst frame Length maximum 8192 bytes minus last frame

Page 101: Special Course in Computer Science: Local Networks

Full-duplex MAC No CSMA/CD needed Almost all Gigabit Ethernet implementations use this

access method When a switch is present

Page 102: Special Course in Computer Science: Local Networks

Gigabit Ethernet physical layer

Page 103: Special Course in Computer Science: Local Networks

Sublayers Reconciliation Sends 8-bit parallel data to PHY via GMII

PHY Only internal; encodes/decodes

GMII Exists on NIC only => logical interface Operates at 1000Mbps Specifies a parallel data path

Page 104: Special Course in Computer Science: Local Networks

Gigabit Ethernet implementations

Page 105: Special Course in Computer Science: Local Networks

1000Base-X

Page 106: Special Course in Computer Science: Local Networks

Transceiver in 1000Base-X Internal Encoding: 8B/10B + NRZ

Page 107: Special Course in Computer Science: Local Networks

1000Base-T 4 5-UTP wires

Page 108: Special Course in Computer Science: Local Networks

Transceiver in 1000Base-T Encoding: 4D-

PAM5 (4 dimensional, 5-level pulse amplitude modulation)

Each code is modulated using 5 levels of PAM

Page 109: Special Course in Computer Science: Local Networks

Ethernet continues to evolve 2002, 2004, 2006: 802.3ae, 10 Gigabit Ethernet Standards for fiber, STP and UTP

10Gbps: prodigious speed Inside data centers Exchanges to connect high-end routers, switches,

servers Long-distance, high-bandwidth trunks between

offices: MAN based on Ethernet and fiber Full duplex (no CSMA/CD) Autonegotiation

Page 110: Special Course in Computer Science: Local Networks

X Gigabit Ethernet 10 Gigabit Ethernet is being deployed where needed

40/100 Gigabit Ethernet is under development (since 2007)

Page 111: Special Course in Computer Science: Local Networks

IEEE 802.3 Ethernet 1976: XEROX PARC Experimental Ethernet, 2.94

Mbps 1978: DEC, Intel, XEROX (DIX) DIX Ethernet, 10

Mbps 1983: IEEE 802.3 standard 1995: IEEE 802.3u Fast Ethernet, 100 Mbps 1997: IEEE 802.3x full duplex mode 1998: IEEE 802.3z Gigabit Ethernet, 1 – 2 Gbps 2002: IEEE 802.3ae 10 Gigabit Ethernet

Page 112: Special Course in Computer Science: Local Networks

What network do we use at ÅA? Gigabit Ethernet, 1Gbps Category 5 UTP cables + some optic fiber as part

of the backbone About 10 switches

Page 113: Special Course in Computer Science: Local Networks

Ethernet retrospective Used for over more than 35 years: rare phenomenon Simple and flexible => reliable, cheap, easy to maintain Interworks easily with TCP/IP, the dominant transport-

network layer protocol (they are connectionless) Able to evolve in certain essential ways

Speeds have gone up by several orders of magnitude Hubs and switches have been introduced The above changes did not require changing the software

Carrier-grade Ethernet Let network providers offer Ethernet-based services to customers

from MAN / WAN (2009)

Page 114: Special Course in Computer Science: Local Networks

Exercises regarding Ethernet I Solve the following exercises from the textbook

Chapter 10: 7-13, 23, 25, 26, 28, 29 Chapter 11: 1-8, 12-13, 19-22

II The word BLUE needs to be broadcast onto a 10Base5 Ethernet LAN. This LAN has 4 stations with addresses 44-AA-C1-23-45-32, 46-56-21-1A-DE-F4, 48-32-21-21-4D-34, 7C-AA-C1-23-45-32. Assume the first station (44…) transmits the info to the others. Show the MAC frame. Then show the communication between LLCs, if the frame is received fine by station 2 (46…), is damaged when arrives at station 3 (48…) and the acknowledgement for the well-received frame is lost from station 4 (7C…) to station 1.

III Assume the above LAN where stations 1, 3, 4 have simultaneously data to transmit. What is a possible scenario for a successful transmission? What happens if station 2 crashes? Any difference if the LAN was a 10Base-FL?

IV How many frames per second can gigabit Ethernet handle? Take into consideration all relevant cases.

V In 100Base-TX the clock runs at 125 Mhz, even though Fast Ethernet is supposed to deliver only 100 Mbps. What is the reason behind this?