Ethernet
Direct connection: point-to-point
• 2 nodes:
datagram
sending
node
frame
adapter (NIC)
• More than 2 nodes?
rcving
node
link layer protocol
frame
adapter (NIC)
•
Direct
connection:
broadcast
Shared media
Metcalfe’s Ethernet
Sketch (1973)
Ethernet “dominant” LAN technology:
•
•
•
•
cheap $30 for 100Mbs!
first widely used LAN technology
simpler, cheaper than token LANs and ATM
kept up with speed race: 10, 100, 1000 Mbps
Ethernet Format: Physical Layer
• Each bit has a transition
• Allows clocks in sending and receiving nodes to
synchronize to each other
– no need for a centralized, global clock among nodes!
Ethernet Format: Framing
• Preamble: (clearing your throat)
– 8 bytes, allows sender/receiver clocks to synchronize
•
Destination/Source Address: (hey Paul, Tom here)
– 6 bytes each
• Type:
– 2 bytes, indicates higher layer protocol
– 0x0800 is IP, 0x0806 is ARP
• Data: 46-1500 bytes
• FCS (CRC):
– catches most transmission errors - errored frames dropped
Ethernet Packet Structure
•14 byte header
•2 addresses
Graphic Source: Network Computing Magazine August 7, 2000
Ethernet Physical Layer Packet
Structure
•8 byte header (Preamble)
Graphic Source: Network Computing Magazine August 7, 2000
Ethernet Addressing
• 6 byte address (unique to each adapter)
– Example: 08-0b-db-e4-b1-02
– 2^48 = 281 trillion; can produce 100 million LAN devices every
day for 2000 years!
• Interpretation of address:
–
–
–
–
–
Upper 24 bits OUI (Organizationally Unique Identifier)
Lower 24 bits Organization-assigned portion
Unicast: lowest bit of first byte is 0
Multicast: lowest bit of first byte is 1
Broadcast: ff-ff-ff-ff-ff-ff
• Adaptor accept frame if and only if:
– Destination address matches adapter address, or
– Destination address is broadcast, or
– Destination address is multicast and adapter has been configured to
accept it
Ethernet Media sharing
• CSMA/CD (the polite
conversationalist)
– carrier sense: don’t transmit if
you sense someone else
transmitting
– collision detection: abort your
transmission if you sense
someone else transmitting
– random access: wait random
time before attempting a
retransmission
Ethernet Technologies
• 10Base2:
– 10Mbps, 200 meters max cable length
– thin coaxial cable in a bus topology
– repeaters connect multiple segments
• 10BaseT / 100BaseT “fast ethernet”:
– 10/100Mbps, Twisted pair
– Nodes connect to a hub in “star topology”
• Gigabit Ethernet:
– 1Gbps, fibre or copper
– Extending from LAN to MAN
• 10 Gbps Ethernet now!
• High data speed + larger distance +
increasing number of devices per LAN =>
switching
nodes
hub
Twisted Pair Wire Map
• EIA/TIA 568B (UGA Standard)
Standard vs Crossover Cables
Card-to-Hub Wiring
(Standard Cable)
TD+
TDRD+
RD+
RDTD+
RD-
TD-
Card-to-Card (Hub-to-Hub) Wiring
(Crossover Cable)
TD+ (RD+)
TD- (RD-)
RD+ (TD+)
TD+ (RD+)
TD- (RD-)
RD+ (TD+)
RD- (TD-)
RD- (TD-)
Power over Ethernet (PoE)
http://www.nwfusion.com/news/2003/1124infrapoe.html
Ethernet
IP: 10.0.0.10
IP: 10.0.0.11
MAC: 00:00:aa:aa:aa:aa
MAC: 00:00:bb:bb:bb:bb
A
B
C
D
IP: 10.0.0.12
IP: 10.0.0.13
MAC: 00:00:cc:cc:cc:cc
MAC: 00:00:dd:dd:dd:dd
• Most popular LAN technology
nowadays 10Mb/s - 1Gb/s
• Each host has unique 48bit
MAC address (factory assigned)
• Frames sent to MAC addresses
• Broadcasts widely used
• To find destination MAC
address, ARP protocol is used
Ethernet frame
Dest
MAC
Source
MAC
Dest
IP
Source
IP
IP packet
Data
ARP: finding the MAC Address
Host A
ARP Query
Broadcast
Host B
MAC ?
Host B
Host B
IP
ARP Response
Host B
Unicast
MAC
Host B
IP
RFC 826: Address Resolution Protocol, 1982
ARP frame format
IP & Ethernet Multicast Address
Mapping
• IP multicast addresses (class D) range from
224.0.0.1 to 239.255.255.255 and map to
Ethernet destination MAC addresses as
shown below
32-bit Class D IP Address
1110
Low-order 23 bits of multicast
Group ID copied to Enet address
00000001 00000000 01011110 0
48-bit Ethernet Address
Multicast Addresses
• Multicast revises addresses to be protocol
specific: high byte, least bit is “1” if multicast.
Multicast(1)
high
byte
Local(1)/global(0)
administration
48 bit address
• Applications that use multicast
–
–
–
–
Imagecast
AppleTalk zones
One-to-many IP video broadcasting
Service location protocol (SLP)
IGMP Snooping
• Internet Group Management Protocol (IGMP RFC 2236) used to manage IP multicast traffic
• Application wishing to receive traffic for specific
IP multicast address sends out an ICMP join
request (or a leave request to stop receiving
multicast)
• Switches that employ IGMP snooping listen for
IGMP join/leave requests to decide when to send
a specific multicast frame to a port
Switching (same as Bridging)
• Goals
– traffic isolation
– “transparent” operation
– plug-and-play
• Operation
– store and forward Ethernet frames
– examine frame header and selectively forward frame based
on MAC dest address
– when frame is to be forwarded on segment, uses CSMA/CD
to access segment
Switching Tables
0260.8c01.1111
E0:
E0:
E1:
E1:
E0
0260.8c01.2222
0260.8c01.1111
0260.8c01.2222
0260.8c01.3333
0260.8c01.4444
0260.8c01.3333
E1
0260.8c01.4444
Spanning Tree Protocol
X
Y
Segment 1
Broadcast
Segment 2
Spanning tree protocol (IEEE 802.1d)
• Every bridge has bridge-id
– bridge-id = 2-byte priority + 6-byte MAC addr
• Question: MAC address of bridge??
• Every port of bridge has
– port-id = 1-byte priority + 1-byte port-number
– port-cost = inversely proportional to link speed
• Bridge with lowest bridge-id is root bridge
• On each LAN segment, bridge with lowest path cost to
root is designated bridge (use bridge-id and port-id to
break ties)
• A bridge forwards frames through a port only if it is a
designated bridge for that LAN segment
STP terminology
• Port roles:
– Root port (switch port leading to root)
– Designated port (LAN port leading to root)
– Alternate / backup port (anything else)
• Port states:
–
–
–
–
Blocking (no send/rcv, except STP bpdus)
Listening (prepare for learning/forwarding)
Learning (learn MAC addr but no forwarding)
Forwarding (send/rcv frames)
• Can disable STP on port or switch
– All frames are forwarded
– BPDUs?
STP operation
• BPDU carries 4-tuple:
– <root-id, root-cost, bridge-id, port-id>
• Store rcvd and send 4-tuple for each port:
– port with best rcvd 4-tuple is root port
• root bridge has no such port
– if send 4-tuple better than rcv 4-tuple, port is designated
port
– rest of the ports are alternate/backup ports
• Various timers
Spanning tree example
A
B
B3
DP
DP
C
B5
DP
B2
E
RP
D
B7
RP
RP
DP
DP
G
DP
B1
root
DP
F
DP
DP
H
RP
B6
I
B4 DP
DP
J
K
New Spanning Tree Protocol versions
 Implementation of :
•Rapid Spanning Tree Protocol 802.1w (RSTP);
•Per VLAN Spanning Tree 802.1q (PVST +);
•Multiple Spanning Tree 802.1s (MST);
•Load balancing across links;
•BPDU guard;
•Root Guard; and
•Uni-Directional Link Detection (UDLD)
Evolution of Spanning Tree
The following developments in Spanning Tree Protocol are examined:
Per-VLAN Spanning Tree (PVST) is a Cisco-proprietary implementation requiring
ISL trunk encapsulation.
PVST+ provides Layer 2 load balancing for the VLAN on which it runs.
MST (IEEE 802.1s) extends the IEEE 802.1w Rapid Spanning Tree (RST)
algorithm to multiple spanning-trees.
Enhanced PVST + or Multiple Instance of Spanning Tree Protocol (MISTP), a
compromise between PVST+ and MST.
802.1w Rapid Spanning Tree Protocol
 The IEEE 802.1w specification, Rapid Spanning Tree Protocol, provides for
subsecond reconvergence of STP after failure of one of the uplinks in a
bridged environment.
802.1w provides the structure on which the 802.1s features such as
multiple spanning tree operates.
There are only three port states left in RSTP corresponding to the three
possible operational states Learning ,Forwarding and Discarding.
Rapid Transition to Forwarding State is the most important feature
introduced by 802.1w:
• RSTP actively confirms safe port transition to forwarding without relying on
timers;
• There is now a real feedback mechanism that takes place between RSTPcompliant bridges.
•In order to achieve fast convergence on a port, the protocol relies upon two
new variables: edge ports and link type.
Virtual LANs
• LAN (broadcast domain) grows large
• “departments” or “workgroups” not happy with
big broadcast domain
– Security (eavesdropping)
– Bandwidth consumed by flooding/multicasting
• Split LAN into multiple broadcast domains
– Multiple physical LANs?
• Too expensive!
• People move all the time!
• VLAN: logical partition of LAN
Virtual LANs
VLANs: IEEE 802.1q
destination
addr
source
addr
VLAN protocol id
= 0x8100
type
data
FCS
3-bit priority
1-bit CFI
12-bit VLAN id
• “Tagged” Ethernet frames contain VLAN-id
• Switch adds/removes tag when forwarding frames between
trunk and non-trunk ports
• Complications:
– Hosts and legacy switches do not understand VLAN tags
– Tag insertion/removal requires FCS recomputation
– Frame length increases beyond legacy MTU
VLAN Standard: IEEE 802.1q
CFI-Canonical Format Identifier (Ethernet/TokenRing)
The 802.3 (legacy) and 802.1Q Ethernet
frame formats
L2 Tunneling
The default system MTU for traffic on the switch is 1500 bytes. You can configure the switch to support
larger frames by using the system mtu global configuration command. Because the 802.1Q tunneling
feature increases the frame size by 4 bytes when the metro tag is added, you must configure all switches in
the service-provider network to be able to process larger frames by increasing the switch system MTU size
to at least 1504 bytes. The maximum allowable system MTU for Catalyst 3550 Gigabit Ethernet switches is
2000 bytes; the maximum system MTU for Fast Ethernet switches is 1546 bytes.
Some Switches Support Priorities
802.1p Prioritization
• Eight levels of prioritization - p0 (lowest)
through p7 (highest)
• 802.1p example
VLAN/802.1p Switch
FS
Internal Queues:
FS
p7: VS
VS
p0: FS
FS
VS
VS
VS
VS
VS
VS
L2 Switch
Store&Forward vs Cut Through
Switching
• The following diagram depicts the differences
between store-and-forward and cut-through
switching
8 Bytes
6 Bytes
6 Bytes
2 Bytes
46-1500 Bytes
4 Bytes
Preamble
/SFD
Destination
Address
Source
Address
Type/Length
Field
Data and Padding
Frame
Check
Sequence
Cut-through forwards
after destination address
Modified cut-through forwards
after 64 bytes of data
Store-and-forward
forwards after FCS
• Switches should employ store-and-forward
exclusively (cut-through propagates bad packets)
Gigabit Ethernet over Fiber
Wave Division Multiplexing
DWDM 1528 to 1560 nm: erbium doped fiber amplifiers (EDFA)
EDFA every 60km, regeneration every 500km
Erbium doped fiber amplifiers
• A pump laser injects a high intensity pulse of light exciting the erbium
and causing the erbium atoms to release their stored energy.
• The EDFA amplifies all the wavelengths to the same level (gain flatness).
• DWDM 1528 to 1560 nm: EDFA every 60km, regeneration every 500km
Input
Coupler
Isolator
1480 or 980 nm
Pump Laser
Output
Erbium Doped Fiber
StarLight 802.1q VLAN Ids
30-Jul-02 11:11
MREN 6509 (AS 22335)
101 STAR TAP (AS 10764)
100
102 SURFNet (AS 1103)
107
103
104 Abilene (AS 11537)
NWU (AS 103)
105
AMPATH (VLAN 105 only)
128
106
6TAP
108
128 MREN M5 (AS 22335)
109
Argonne (AS 683)
113
110
112
111
117
118 NREN (AS 24)
119
120
121
114 CERN
124
125
126
115
122 ESNet (AS 293)
129
116
NaukaNet
123
130
134
135
131
133
136
132
CA*net-Winnipeg (AS 6509)
147
146
145
144
143
142
141
140
X CA*net-Toronto (AS 6509)
127
OMNINet
137
148
vBNS (AS 145) [Unknown]
138
EVL
Configuration Example
interface GigabitEthernet2/9
description NISN/NASA
mtu 9216
no ip address
speed nonegotiate
switchport
switchport trunk encapsulation dot1q
switchport trunk allowed vlan 210-213,217-226,231,232
switchport mode trunk
switchport nonegotiate
interface GigabitEthernet2/10
description GEMnet
mtu 9216
no ip address
speed nonegotiate
switchport
switchport trunk encapsulation dot1q
switchport trunk allowed vlan 167-169,231
switchport mode trunk
switchport nonegotiate
WKN 20040414