• VLAN • High Speed Ethernet – Introduction – Fast Ethernet – Gigabit Ethernet – 10 Gigabit Ethernet

Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

The Beginning

• initial idea: shared media LAN – bus structure, CSMA/CD was access method – coax cable, transmission rate up to 10 Mbit/s – half-duplex transmission (two physical wires e.g. coax)

The Ethernet Evolution

collision domain From 10Mbit/s to 10Gigabit/s Ethernet Technology From Bridging to L2 Ethernet Switching and VLANs

Agenda Enlarging the Network

• Ethernet Evolution • VLAN • High Speed Ethernet – Introduction – Fast Ethernet – Gigabit Ethernet – 10 Gigabit Ethernet

repeater as signal amplifier used to enlarge the network diameter but no network segmentation !!! still one collision domain !!!

Page 07 - 1 Page 07 - 2 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Multiport Repeater Structured Cabling (2)

• demand for telephony-like point-to-point cabling represents four CU wires 2 for Tmt, 2 for Rcv using Twisted Pair wires Server Farm (e.g. 10BaseT) – based on structured cabling standard represents two FO wires – 10BaseT as new Ethernet type to support this demand e.g. 10BaseF – four physical wires (2 for tmt, 2 for rcv) • network stations are connected star-like to a 10 Base F multiport repeater – multiport repeater is called “hub” • hub simulates the bus: "CSMA/CD in a box" • only half-duplex 10 Base T – only one network station can use the network at a given Client-PCs time, all others have to wait

Structured Cabling (1) Bridging

• simple physical amplification with repeaters became insufficient Multiport Repeater, “Hub” “CSMA/CD in a box” – with repeaters all nodes share the given bandwidth – the whole network is still one collision domain – -> technology moved toward layer 2 • bridges segment a network into smaller collision domains – store and forward technology (packet switching) – the whole network is still a broadcast domain – Spanning Tree provides a unique path between each two 10 Base T devices and avoids broadcast storms

Page 07 - 3 Page 07 - 4 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Network Segmentation with Bridges Switching (2)

collision domain on point-to-point link reduced to a single link collision domain on Bridge Broadcast Domain shared media (only half duplex possible)

Switch

Collision Domains

Switching (1) Switching (3)

• "switching" means fast transparent bridging full duplex on point-to-point links collision domain on – implemented in hardware shared media (only half – also called Layer 2 (L2) switching or Ethernet switching duplex possible) • multiport switches allow full duplex operation on point-to-point links – no need for collision detection (media access control) on a link which is shared by two devices only • network station <-> switch port • switch <-> switch • multiport switches replaces multiport repeaters – a collision free Ethernet can be built, if network consists of point-to-point links only

Page 07 - 5 Page 07 - 6 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Switching (4) Switching (6)

full duplex everywhere = collision free Ethernet LAN

Server 100 Mbit/s 100 Mbit/s Server

1 Gbit/s 1 Gbit/s Flow Control possible

Clients Broadcast Domain 10 Mbit/s

Switching (5) Redundant Topology L2 Switching

PC3 PC6 • L2 switches can connect Ethernets with 10 represents four CU wires 2 for Tmt, 2 for Rcv Mbit/s, 100 Mbit/s or 1000 Mbit/s for example (e.g. 10BaseT) – clients using 10 Mbit/s either half duplex on shared media MAC D MAC F p1 p2 represents two FO wires or full duplex on point-to-point connection with switch e.g. 10BaseF – server uses 100 Mbit/s, full duplex, point-to-point S3 connection with switch t1 t2

– note: multiport repeater is not able to do this ! t1 Trunks t1

• L2 switch as packet switch operates with t2 t2 asynchronous TDM S1 S2 p1 p2 p1 p1 p2 – congestion can be avoided by using a new MAC based MAC A MAC B MAC E MAC C flow control (pause command)

PC1 PC4 PC5 PC2

Page 07 - 7 Page 07 - 8 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Spanning Tree Applied Agenda

PC3 PC6 • Ethernet Evolution • VLAN MAC D MAC F p1 p2 • High Speed Ethernet

S3 – Introduction – Fast Ethernet t1 F t2 F (Forward) – Gigabit Ethernet t1 F Trunk t1 B (Blocked) – 10 Gigabit Ethernet t2 F t2

S1 S2 p1 p2 p1 p1 p2

MAC A MAC B MAC E MAC C

PC1 PC4 PC5 PC2

Switching Table (L2) Virtual LANs (1)

Switching Table S3 pro VLAN PC3 PC6 • today's work-groups are expanding over the MAC-Addr. Port/Trunk D p1 whole campus in case of local environment MAC D MAC F F p2 p1 p2 • users of one workgroup should be kept A,B,C,E t1 separated from other workgroups S3 Switching Table S1 pro VLAN – because of security reasons they should see there MAC-Addr. Port/Trunk t1 t2 Switching Table S2 pro VLAN necessary working environment only A p1 MAC-Addr. Port/Trunk B p2 t1 t1 E p1 • end-systems of one workgroup should see D,F t1 t2 t2 C p2 C,E t2 broadcasts only from stations of same A,B,D,F t2 S1 S2 p1 p2 p1 p1 p2 workgroup MAC A MAC B MAC E MAC C • the network must be flexible – to adapt continuous location changes of the end- PC1 PC4 PC5 PC2 systems/users

Page 07 - 9 Page 07 - 10 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Virtual LANs (2) VLAN Assignment

• base idea of VLAN: • a station may be assigned to a VLAN – multiplexing of several LANs via same infrastructure – port-based (switches and connection between switches) • fixed assignment port 4 -> VLAN x • today's switches got the ability to combine • most common approach several network-stations to so-called "Virtual • a station is member of one specific VLAN only LANs“ – MAC-based • MAC A -> VLAN x – separate bridging/switching table maintained for every • allows integration of older shared-media components and single VLAN automatic location change support – separate broadcast handling for every single VLAN • a station is member of one specific VLAN only • each Virtual LAN is its own broadcast domain – protocol-based – separate Spanning Tree for every single VLAN • IP-traffic, port 1 -> VLAN x • note: IEEE 802.1w specifies a method to share one Rapid • NetBEUI-traffic, port 1 -> VLAN y Spanning Tree among all VLANs • a station could be member of different VLANs

VLAN Example Virtual Trunks - VLAN tagging

VLAN A • switches must be connected via VLAN-trunks on A4 A5 A1A2 A3 which each particular VLAN-frame is "tagged" (marked) with an identifier – examples for tagging standards: • IEEE 802.10 (pre 802.1Q temporary solution) • ISL (Cisco) Table VLAN A Table VLAN A • IEEE 802.1Q Table VLAN B Table VLAN B • so switches can distinguish between several VLANs and manage their respective traffic

B1 B2 B3 B4 B5 VLAN B

Page 07 - 11 Page 07 - 12 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

802.1Q VLAN Tagging 1 VLAN Operation (1)

802.3 802.2 LLC VLAN A A4 A5 802.1Q A1A2 A3 preamble DA SA length DSAP SSAP Ctrl data FCS Fields

A1 -> A3 A5 -> broadcast 2 Byte 2 Byte TPID … Tag Protocol Identifier TPID TIC TCI … Tag Control Information 0x8100 trunk untagged frames

3 Bit 1 Bit 12 Bit UP CFI VID B1 -> B5

UP … User Priority note: With tagging Ethernets maximal CFI … Canonical Format Identifier frame length = 1522, minimal frame VID … VLAN Identifier B1 B2VLAN B B3 B4 B5 length = 68

802.1Q VLAN Tagging 2 VLAN Operation (2)

Ethernet V2 VLAN A A4 A5 802.1Q A1A2 A3 preamble DA SA type data FCS Fields

2 Byte 2 Byte A1 -> A3 tag VLAN A TPID … Tag Protocol Identifier A5 -> broadcast TPID TIC TCI … Tag Control Information A5 -> broadcast 0x8100 B1 -> B5 3 Bit 1 Bit 12 Bit UP CFI VID tag VLAN B

UP … User Priority note: With tagging Ethernets maximal CFI … Canonical Format Identifier frame length = 1522, minimal frame VID … VLAN Identifier B1 B2VLAN B B3 B4 B5 length = 68

Page 07 - 13 Page 07 - 14 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

VLAN Operation (3) Multihomed VLAN 2

VLAN A VLAN A A1A2 A3 A4 A5 A1A2 A3 A4 A5

A5 -> broadcast tag VLAN A

A6->A3 A6 = B6->B2 B6

B1 -> B5 tag VLAN B

B1 B2VLAN B B3 B4 B5 B1 B2VLAN B B3 B4 B5

Multihomed VLAN 1 Multihomed VLAN 3

VLAN A VLAN A A1A2 A3 A4 A5 A1A2 A3 A4 A5

tag VLAN A A6 -> A3

A6->A3 A6 A6 = = B6->B2 B6 B6

tag VLAN B B6 -> B2

B1 B2VLAN B B3 B4 B5 B1 B2VLAN B B3 B4 B5

Page 07 - 15 Page 07 - 16 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Trunking between L2 Switches Trunking with LCAP / FEC / GEC

• on trunks between multiport switches full duplex VLAN A A4 A5 operation is possible A1A2 A3 – hence "200 Mbit/s" with Fast Ethernet – hence "2 Gbit/s" with Gigabit Ethernet Access Port Access Port • on trunks bundling (aggregation) of physical One logical trunk for STP

links to one logical link is possible Table VLAN A Table VLAN A Trunk 1 Trunk 2 – Fast Ethernet Channeling (Cisco) Table VLAN B Table VLAN B • 400 / 800 Mbit/s Load Balancing over two – Gigabit Ethernet Channeling (Cisco) physical trunk lines • 4 / 8 Gbit/s – IEEE 802.3 (2002) LACP (Link Aggregation Control Protocol B1 B2 B3 B4 B5 VLAN B

Trunking without LCAP / FEC / GEC Communication between VLANs

VLAN A • switches do not allow traffic between (different) A4 A5 A1A2 A3 VLANs • end-systems have to make use of routers Access Port • routers can be either part of several VLANs (via Access Port multiple physical ports), or Trunk 1 Table VLAN A Table VLAN A • routers provide VLAN-trunk capabilities -> router Table VLAN B Table VLAN B Trunk 2 (blocked by STP) must be able to recognize and change VLAN Bandwidth of trunk 2 not used tags

B1 B2 B3 B4 B5 VLAN B

Page 07 - 17 Page 07 - 18 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Agenda Full-Duplex Mode

• Ethernet Evolution • full-duplex mode is possible on point-to-point • VLAN links • High Speed Ethernet – except 100BaseT4 (Cat 3 cable), 100BaseVG which can work in half duplex mode only – Introduction – note: 10Base2 and 10Base5 are shared links and by – Fast Ethernet default half duplex medias – Gigabit Ethernet • if a network station is connected to an Ethernet – 10 Gigabit Ethernet switch via point-to-point link – CSMA/CD is not in necessary and can be switched off • now a network station can – send frames immediately (without CS) using the transmission-line of the cable and simultaneously receive data on the other line

IEEE 802.3 (2002) Flow Control

• the latest version of IEEE 802.3 specifies • speed-requirements for switches are very high – operation for 10 Mbit/s, 100 Mbit/s, Gigabit/s and 10Gigabit/sEthernet – especially in full duplex operation – full duplex Ethernet – auto-negotiation – also powerful switches can't avoid buffer overflow –flow control – earlier, high traffic caused collisions and CSMA/CD • it is still backward compatible to the old times of interrupted the transmission in these situations, now high Ethernet traffic is normal – CSMA/CD (half-duplex) operation in 100 and 1000 Mbit/s Ethernets with multiport repeater possible • L4 flow control (e.g. TCP) between end-systems – frame bursting or carrier extension for ensuring slot-time demands in is not efficient enough for a LAN 1000 Mbit/s Ethernet – switches should be involved to avoid buffer overflow • IEEE 802.3ae specifies (2004) – operation for 10 Gigabit/s Ethernet over fiber • therefore a MAC based (L2) flow control is • IEEE 802.3ak specifies (2006) specified – operation for 10 Gigabit/s Ethernet over copper – MAC-control-protocol and the Pause command

Page 07 - 19 Page 07 - 20 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

MAC-Control Frame The Pause Command 2

• identified among other frames by setting length • destination address is either field = 8808 hex – address of destination station or – broadcast address or always 64 octets – special multicast address 01-80-C2-00-00-01 8 octets 6 6 2 2 44 4 this special multicast address prevents bridges preamble DA SA 8808h MAC-ctrl opcode MAC-ctrl parameters FCS • to transfer associated pause-frames to not (Length) concerned network segments MAC-ctrl opcode ...... defines function of control frame • hence flow-control (with pause commands) MAC-ctrl parameters .... control parameter data; always filled up to 44 bytes, by affects only the own segment using zero bytes if necessary • currently only the "pause" function is available (opcode 0x0001)

The Pause Command 1 Demand for Higher Speed

• on receiving the pause command • higher data rates need more sophisticated – station stops sending normal frames for a given time coding which is specified in the MAC-control parameter field – 10 Mbit/s Ethernet: Manchester coding • this pause time is a multiple of the slot time – Fast Ethernet (100 Mbit/s): 4B/5B block code – 4096 bit-times when using Gigabit Ethernet or 512 bit- – Gigabit Ethernet 1000 Mbit/s): 8B/10B block code times with conventional 802.3 • new implementations should be backwards- • paused station waits compatible – until pause time expires or an additional MAC-control – old physical layer signaling interface (PLS), represented frame arrives with pause time = 0 by AUI, was not suitable for new coding technologies – note: paused stations are still allowed to send MAC- • AUI has been replaced control-frames (to avoid blocking of LAN) – MII (Media Independent Interface) for Fast Ethernet – GMII for Gigabit Ethernet

Page 07 - 21 Page 07 - 22 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

New Physical Sublayers PHY Sublayers

Logical Link Control LLC • PLS has been replaced with the Reconciliation MAC Control (optional) Data Link Layer sublayer Media Access Control MAC – Reconciliation layer transforms old MAC PLS-primitives PLS Reconciliation Reconciliation Reconciliation into MII control signals MII MII GMII • MII serves as an interface between MAC and PHY AUI PLS PCS PCS – hides coding issues from the MAC layer AUI PMA PMA PHY PMA (MAU) PMA PMD PMD – MII: often a mechanical connector for a wire; GMII is an MDI MDI MDI MDI interface specification between MAC-chip and PHY-chip Medium Medium Medium Medium upon a circuit board 1-10 Mbit/s 10 Mbit/s 100 Mbit/s 1000 Mbit/s – one independent specification for all physical media

AUI...Attachment Unit Interface, PLS...Physical Layer Signaling, MDI...Medium Dependent Interface, – supports several data rates (10/100/1000 Mbits/s) PCS...Physical Coding Sublayer, MII...Media Independent Interface, GMII...Gigabit Media Independent – 4 bit (GMII: 8 bit) parallel transmission channels to the Interface, PMA...Physical Medium Attachment, MAU...Medium Attachment Unit, PMD...Physical Medium Dependent physical layer © 2007, D.I. Manfred Lindner Ethernet Evolution, v4.6 45 © 2007, D.I. Manfred Lindner Ethernet Evolution, v4.6 47

PHY Sublayers PHY Sublayers

• Physical Layer Signaling (PLS) serves as • Physical Coding Sublayer (PCS) abstraction layer between MAC and PHY – encapsulates MAC-frame between special PCS delimiters • PLS provides – 4B/5B or 8B/10B encoding respectively – data encoding/decoding (Manchester) – appends idle symbols – translation between MAC and PHY • Physical Medium Attachment (PMA) – Attachment Unit Interface (AUI) to connect with PMA – interface between PCS and PMD • several new coding techniques demands for a – (de) serializes data for PMD (PCS) Media Independent Interface (MII) • Physical Medium Dependent (PMD) • today coding is done through an media- – serial transmission of the code groups dependent Physical Coding Sublayer (PCS) – specification of the various connectors (MDI) below the MII

Page 07 - 23 Page 07 - 24 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Bridging Aspects Gigabit Ethernet becomes WAN

• new PHY-sublayers preserves old Ethernet MAC • point-to-point full-duplex connections do not frame format limit the maximal network diameter as CSMA/CD – bridging from 10 Mbit/s Ethernet to 100 Mbit/s Ethernet does does not require a bridge to change the frame format – Gigabit over fiber optic cables reach 70 km length (and • Remark: bridging from 10 Mbit/s Ethernet to FDDI (100 Mbit/s even more) Token ring) requires frame format changing -> slower !! • trend moves towards layer 3 switching • therefore Ethernet L2 switches – high amount of today's traffic goes beyond the border of – can connect Ethernets with 10 Mbit/s, 100 Mbit/s or 1000 the LAN Mbit/s easily and fast – routing decisions enable load balancing and decrease network traffic • Gigabit Ethernet becomes WAN technology

Today: Gigabit Ethernet Agenda

• continues point-to-point and full-duplex idea • Ethernet Evolution • also backward compatible with initial 10 Mbit/s • VLAN shared media idea -> CSMA/CD capable • High Speed Ethernet • but nobody uses it as shared media! – Introduction – multiport repeater with Gigabit Ethernet seems absurd – Fast Ethernet because of small network diameter (20m) – Gigabit Ethernet • 200m with carrier extension and burst mode – 10 Gigabit Ethernet – bandwidth sharing decreases performance; every collision domain produces an additional delay for a crossing packet – full duplex means exclusive, unshared, high performance point-to-point connections between two stations (total 2Gbit/s!)

Page 07 - 25 Page 07 - 26 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

100 Mbit/s Ethernet Implementation Overview 2

• Access method disagreement split 100 Mbit/s LAN development into two branches: – Fast Ethernet - IEEE-802.3u (today 802.3-2002) preserves classical HP and AT&T own Ethernet specification for time – 100VG-AnyLAN - IEEE-802.12 (disappeared) sensitive applications • Fast Ethernet was designed as 100 Mbit/s and backwards-compatible 10Mbit/s Ethernet 100BaseT 100VG-AnyLAN – CSMA/CD but also Access method: Access method: demand priority – Full-duplex connections (collision free) CSMA/CD • Network diameter based on collision window IEEE 802.3 IEEE 802.12 requirement (512 bit times) – reduced by factor 10 – e.g. 250m compared with 2500m at 10 Mbit/s

Implementation Overview 1 Fast Ethernet

• AUI has been replaced with the Media different signaling (coding) schemes Independent Interface (MII) – New coding (4B/5B, 8B/6T, PAM 5x5) and bandwidth Fast Ethernet Fast Ethernet Fast Ethernet constrains demand for a redesigned abstraction layer 100BaseX 100BaseT4 100BaseT2 • MII defines a generic 100BaseT interface – Allows utilization of a 100BaseTX, 100BaseFX, 100BaseT4 or a 100BaseT2 transceiver 100BaseT4 100BaseT2 100BaseFX 100BaseTX • On-board or cable-connector with (half duplex) (full duplex) • 20 shielded, symmetrically twisted wire pairs -> 40 poles two optical two pairs UTP four pairs UTP two pairs • One additional main-shield fibers Cat 5 or STP Cat 3 or better balanced TP • 68 Ohm impedance; 2.5 ns maximal delay Cat 3 or better • 50 cm maximal length

"100BaseT"

Page 07 - 27 Page 07 - 28 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Typical Fashion FLP Burst Coding

Computer I/O Bus Fast Ethernet card 17 odd-numbered pulses (clock intern 100BaseTX pulses) transceiver 100 ns MAC

MII PHY

1 2 3 4 5 6 7 8 9 10 11 12

MII connector MDI RJ45 connector 62.5 µs Up to 16 even-numbered data bit-pulses MII-cable MII Media Independent Interface MDI Medium Dependent Interface PHY Physical Layer Device e.g. 100BaseFX = 1 1 0 1 0 1 .... PHY transceiver MAC Media Access Control Unit

MDI e.g.Fibre MIC connector

Autonegotiation Autonegotiation

• Autonegotiation support enables two 100BaseT • To avoid increase of traffic FLP-bursts are only devices (copper only) to exchange information sent on connection-establishments about their capabilities • 100BaseT stations recognizes 10 Mbit/s stations – signal rate, CSMA/CD or full-duplex by receiving a single NLP only • Achieved by Link-Integrity-Test-Pulse-Sequence • Two 100BaseT stations analyze their FLP-bursts – Normal-Link-Pulse (NLP) technique is already available in and investigate their largest common set of 10BaseT to check the link state features – 10 Mbit/s LAN devices send every 16 ms a 100ns lasting Last frames are sent 3 times -> other station NLP -> no signal on the wire means disconnected • responds with acknowledge-bit set • 100BaseTX uses bursts of Fast-Link-Pulses (FLP) consisting of 17-33 NLPs • Negotiated messages are sent 6-8 times – FLP- session stops here – Each representing a 16 bit word

Page 07 - 29 Page 07 - 30 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

FLP-Session Base Page

• The first FLP-burst contains the base-link • Remote Fault (RF) codeword – Signals that the remote station has recognized an error • By setting the NP bit a sender can transmit • Next Page (NP) several "next-pages" – Signals following next-page(s) after the base-page – Next-pages contain additional information about the • Acknowledge (Ack) vendor, device-type and other technical data – Signals the receiving of the data (not the feasibility) • Two kinds of next-pages – If the base-page has been received 3 times with the NP – Message-pages (predefined codewords) set to zero, the receiver station responds with the Ack bit – Unformatted-pages (vendor-defined codewords) set to 1 • After reaching the last acknowledgement of this – If next-pages are following, the receiver responds with FLP-session, the negotiated link-codeword is Ack=1 after receiving 3 FLP-bursts sent 6-8 times

Base Page Coding

• 4B/5B block encoding: each 4-bit group encoded S0 S1 S2 S3 S4 A0 A1 A2 A3 A4 A5 A6 A7 RF Ack NP by a 5 bit run-length limited "code-group” – Code groups lean upon FDDI-4B/5B codes Selector field Technology ability field – Some additional code groups are used for signaling purposes; remaining code groups are violation symbols provides selection of up to 32 Bit Technology -> easy error detection different message types; currently A0 10BaseT – Groups determinate maximal number of transmitted zeros only 2 selector codes available: A1 10BaseT-full duplex 10000....IEEE 802.3 A2 100BaseTx or ones in a row -> easy clock synchronization 01000....IEEE 802.9 A3 100BaseTx-full duplex • Keeps DC component below 10% (ISLAN-16T) A4 100BaseT4 • Code groups are transmitted using NRZI- (ISO-Ethernet) A5 Pause operation for full duplex links A6 reserved encoding A7 reserved – Code efficiency: 4/5 = 100/125 = 80% (Manchestercode only 50 %)

Page 07 - 31 Page 07 - 32 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

4B/5B Coding Signaling Types

MII 3 2 1 0 • Three signaling types : 1 1 0 1 16 code groups – 100BaseX: • refers to either the 100BaseTX or 100BaseFX specification 4 x 25 Mbit/s – 100BaseT4 PCS – 100BaseT2 4B/5B Encoder/Decoder 32 code groups

25 million code groups per second • 100BaseX 4 3 2 1 0 1 1 0 1 1 – combines the CSMA/CD MAC with the FDDI Physical Medium Dependent layer (PMD) – allows full duplex operation on link 125 Mbit/s PMA

Code Group Table Signaling Types

PCS code-group name MII group • 100BaseT4 11110 0 0000 Remaining code groups 01001 1 0001 – allows half duplex operation only 10100 2 0010 are not valid (triggers 10101 3 0011 –8B6T code 01010 4 0100 error detection) 01011 5 0101 – Uses 4 pairs of wires; one pair for collision detection, three 01110 6 0110 pair for data transmission DATA 01111 7 0111 10010 8 1000 – One unidirectional pair is used for sending only and two 10011 9 1001 10110 A 1010 bi-directional pairs for both sending and receiving 10111 B 1011 11010 C 1100 – Same pinout as 10BaseT specification 11011 D 1101 11100 E 1110 – Transmit on pin 1 and 2, receive on 3 and 6; bi-directional 11101 F 1111 on 4 and 5; bi-directional on 7 and 8 11111 I undefined Idle pattern between streams 11000 J 0101 Start of Stream Delimiter (1st part) Control 10001 K 0101 Start of Stream Delimiter (2nd part) 01101 T undefined End of Stream Delimiter (1st part) 00111 R undefined End of Stream Delimiter (2nd part) 00100 H undefined signals receiving errors © 2007, D.I. Manfred Lindner Ethernet Evolution, v4.6 66 © 2007, D.I. Manfred Lindner Ethernet Evolution, v4.6 68

Page 07 - 33 Page 07 - 34 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

100BaseTX and 100BaseFX Agenda

• 100BaseTX: • Ethernet Evolution – 125 MBaud symbol rate, full duplex, binary encoding • VLAN – 2 pair Cat 5 unshielded twisted pair (UTP) or 2 pair STP or • High Speed Ethernet type 1 STP – Introduction – RJ45 connector; same pinout as in 10BaseT (transmit on 1 and 2, receive on 3 and 6) – Fast Ethernet – Gigabit Ethernet • 100BaseFX: – 10 Gigabit Ethernet – 125 MBaud symbol rate, full duplex, binary encoding – Two-strand (transmit and receive) 50/125 or 62.5/125-µm multimode fiber-optic cable – SC connector, straight-tip (ST) connector, or media independent connector (MIC)

100BaseT4 and 100BaseT2 Gigabit-Ethernet: IEEE-802.3z / IEEE802.3ab

• 100BaseT4: • Easy integration in existing 802.3 LAN – 25 MBaud, half duplex, ternary encoding configurations because backwards compatible – Cat3 or better, needs all 4 pairs installed – Through integration of 3 different transceivers for 10, 100 – 200 m maximal network diameter and 1000 Mbit/s – maximal 2 hubs – No need to change existing equipment • 100BaseT2: – Supports also 10 Mbit/s and 100 Mbit/s (not with fibre) – Access methods: CSMA/CD or full duplex – 25 MBaud, full duplex, quinary encoding – 2 pairs Cat3 or better • Backbone technology; has also WAN capabilities – Reaches 70 km length using fibre optics – 1 Gbit/s data rate in both directions (full duplex mode, no collisions) – MAC based congestion avoidance (pause frame) © 2007, D.I. Manfred Lindner Ethernet Evolution, v4.6 70 © 2007, D.I. Manfred Lindner Ethernet Evolution, v4.6 72

Page 07 - 35 Page 07 - 36 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Implementation Overview CSMA/CD Restrictions (Half Duplex Mode)

• The conventional collision detection mechanism Media Access Control (MAC) full duplex and/or half duplex CSMA/CD – Requires that stations have to listen (CS) twice the signal Gigabit Media Independent Interface (GMII) propagation time to detect collisions 1000BaseX 1000BaseT – Collision window of 512 bit times at a rate of 1Gbit/s limits 8B/10B encoder/decoder encoder/decoder the maximal net expansion to 20m !

1000Base-CX 1000Base-LX 1000Base-SX 1000Base-T Shielded LWL SWL UTP Balanced Fiber Optic Fiber Optic Cat 5e Copper

802.3z physical layer 802.3ab physical layer

Physical Sublayers CSMA/CD Restrictions (Half Duplex Mode)

• Solutions to increase the maximal net Logical Link Control LLC expansion: MAC Control (optional) – Carrier Extension: Media Access Control MAC • extension bytes appended to (and removed from) the Ethernet Reconciliation frame by the physical layer GMII • frame exists a longer period of time on the medium PCS – Frame Bursting: PMA PHY • to minimize the extension bytes overhead, station may chain LX-PMD SX-PMD CX-PMD several frames together and transmit them at once ("burst"). LX-MDI SX-MDI CX-MDI Medium Medium Medium 1000 Base LX 1000 Base SX 1000 Base CX

Page 07 - 37 Page 07 - 38 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Frame Bursting 1 1000BaseX Coding

• With both methods the minimal frame length is • 8B/10B block encoding: each 8-bit group increased from 512 to 4096 bits encoded by a 10 bit “code-group” (symbol) – = 512 bytes – Half of the code-group space is used for data transfer – The corresponding time is called slottime – Some code groups are used for signaling purposes • If a station decides to chain several frames to a – Remaining code groups are violation symbols burst frame, the first frame inside the burst • -> easy error detection frame must have a length of at least 512 bytes – Groups determine the maximal number of transmitted zeros or ones in a 10 bit symbol – By using extension bytes if necessary • -> easy clock signal detection (bit synchronization) • The next frames (inside the burst frame) can – No baselinewander (DC balanced) have normal length (i.e. at least 64 bytes) • lacking DC balance would result in data-dependent heating of lasers which increases the error rate – Code efficiency: 8/10 = 1000/1250 = 80%

Frame Bursting 2 8B/10B Coding

Station may chain frames up to 8192 bytes • GMII 7 6 5 4 3 2 1 0 (=burst limit) 256 code groups 1 0 0 1 0 1 0 1 – Also may finish the transmission of the last frame even beyond the burst limit 8 x 125 Mbit/s • So the whole burst frame length must not PCS exceed 8192+1518 bytes 1024 code groups 8B/10B Encoder/Decoder – Incl. interframe gap of 0.096 µs = 12 bytes 125 million code groups per second 802.3 frame + byte ext. if-gap 802.3 frame if-gap ...... 802.3 frame 9 8 7 6 5 4 3 2 1 0 1 0 1 0 1 0 1 1 0 1 burst limit

whole burst frame length 1250 Mbit/s PMA

Page 07 - 39 Page 07 - 40 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Implementations Autonegotiation

• actually 2 different wavelengths on fibre media, • Both 1000Base-X and 1000Base-T provide both full duplex, SC connector autonegotiation functions to determinate the – 1000Base-SX: short wave, 850 nm multimode – Access mode (full duplex - half duplex) (up to 550 m length) – Flow control mode – 1000Base-LX: long wave, 1300 nm multimode or • Additionally 1000Base-T can resolve the data monomode (up to 5 km length) rate • 1000Base-CX: – Backward-compatibility with 10 Mbit/s and 100 Mbit/s – Twinax Cable (high quality 150 Ohm balanced shielded – Also using FLP-burst sessions copper cable) – About 25 m distance limit, DB-9 or the newer HSSDC connector

1000BaseT 1000BaseX Autonegotiation

• 1000Base-T defined by 802.3ab task force • 1000Base-X autonegotiation uses normal – UTP uses all 4 line pairs simultaneously for duplex (1000Base-X) signalling ! transmission! – Signaling part of the 8B/10B code groups • Using echo-cancelling: receiver subtracts own signal – No fast link pulses ! – 5 level PAM coding • Autonegotiation had never been specified for traditional fiber- • 4 levels encode 2 bits + extra level used for Forward Error based Ethernet Correction (FEC) • So there is no need for backwards-compatibility – Signal rate: 4 x 125 Mbaud = 4 x 250Mbit/s data rate • 1000Base-X does not negotiate the data rate ! • Cat. 5 links, max 100 m; all 4pairs, cable must conform to the • Only gigabit speeds possible requirements of ANSI/TIA/EIA-568-A – Only 1 CSMA/CD repeater allowed in a collision domain • 1000Base-X autonegotiation resolves • note: collision domains should be avoided – Half-duplex versus full-duplex operation – Flow control

Page 07 - 41 Page 07 - 42 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

Base-Page Agenda

full duplex half duplex acknowledge next page • Ethernet Evolution reserved reserved • VLAN

rsvd rsvd rsvd rsvd rsvd FD HD PS1 PS2 rsvd rsvd rsvd RF1 RF2 Ack NP • High Speed Ethernet – Introduction

technology ability field remote fault – Fast Ethernet – Gigabit Ethernet PS1PS2 description RF1RF2 description – 10 Gigabit Ethernet 0 0 no pause 0 0 no error 0 1 asymmetrical pause 0 1 offline 1 0 symmetrical pause 1 0 connection error 1 1 symmetrical and 1 1 autonegotiation error asymmetrical pause (no common capabilities)

1000BaseT Autonegotiation 10 Gigabit Ethernet (IEEE 802.3ae)

• Autonegotiation is only triggered when the • Preserves Ethernet framing station is powered on • Maintains the minimum and maximum frame size • At first the stations expects Gigabit-Ethernet of the 802.3 standard negotiation packets (replies) • Supports only full-duplex operation • If none of them can be received, the 100Base-T – CSMA/CD protocol was dropped fast link pulse technique is tried • Focus on defining the physical layer • At last the station tries to detect 10Base-T – Four new optical interfaces (PMD) stations using normal link pulses • To operate at various distances on both single-mode and multimode fibers – Two families of physical layer specifications (PHY) for LAN and WAN support – Properties of the PHY defined in corresponding PCS • Encoding and decoding functions © 2007, D.I. Manfred Lindner Ethernet Evolution, v4.6 86 © 2007, D.I. Manfred Lindner Ethernet Evolution, v4.6 88

Page 07 - 43 Page 07 - 44 Datenkommunikation 384.081 - SS 2007 Datenkommunikation 384.081 - SS 2007

L07 - Ethernet Evolution L07 - Ethernet Evolution

PMDs IEEE 802.3ae PMDs, PHYs, PCSs

• 10GBASE-L – SM-fiber, 1300nm band, maximum distance 10km • 10GBASE-E – SM-fiber, 1550nm band, maximum distance 40km • 10GBASE-S PCS – MM-fiber, 850nm band, maximum distance 26 – 82m 10GBASE-E 10GBASE-ER 10GBASE-EW – With laser-optimized MM up to 300m 10GBASE-L 10GBASE-LR 10GBASE-LW PMD • 10GBASE-LX4 10GBASE-S 10GBASE-SR 10GBASE-SW – For SM- and MM-fiber, 1300nm 10GBASE-L4 10GBASE-LX4 – Array of four lasers each transmitting 3,125 Gbit/s and four receivers arranged in WDM (Wavelength-Division Multiplexing) fashion LAN PHY WAN PHY – Maximum distance 300m for legacy FDDI-grade MM-fiber – Maximum distance 10km for SM-fiber

WAN PHY / LAN PHY and their PCS 10 Gigabit Ethernet over Copper

• LAN-PHY • IEEE 802.3ak defined in 2004 – 10GBASE-X – 10GBASE-CX4 – 10GBASE-R – Four pairs of twin-axial copper wiring with IBX4 connector • 64B/66B coding running at 10,3125 Gbit/s – Maximum distance of 15m • WAN-PHY • IEEE 802.3an working group – 10GBASE-W – 10GBASE-T • 64B/66B encoded payload into SONET concatenated STS192c – CAT6 UTP cabling with maximum distance of 55m to frame running at 9,953 Gbit/s 100m • Adaptation of 10Gbit/s to run over traditional SDH links – CAT7 cabling with maximum distance of 100m – Standard ratification expected in July 2006

Page 07 - 45 Page 07 - 46