Title: Data/Link Layer Issues
1Data/Link Layer Issues
- Protocol Services
- Topology
- Error Detection Recovery
2Topology vs Geography
Physical Layout How the signal actually travels
Logical Layout "How devices talk to each other"
-or- "How devices hear each other"
3Topologies
4BUS
Every node hears every other node's
transmission directly.
5Ring
Series of unidirectional point-to-point
links without "store forward", usually with a
bypass ability.
6Star
Switching functions all in central node
7Mesh
Each node independently routes over
(bi-directional) point-to-point links.
8IEEE OSI
LLC
2
MAC
1
PHY
LLC Logical Link Control MAC Media Access
Control PHY Physical
9Link/Physical Layer Standards
- Ethernet
- 10BASET, Fast Ethernet, Gigabit Ethernet
- Token Ring
- 4/16MB
- FDDI
- ATM
10Ethernet IEEE 802.3
What the IEEE standard covers- Physical layer
and interface to the link layer. IEEE 802.2 is
the Link layer standard. History-
DEC/Intel/Xerox came up with it, then submitted
to IEEE for standardization. Some changes were
made so Ethernet is not identical to IEEE
802.3 Differences between Ethernet and 802.3
There are some electrical and connector
differences most equipment uses IEEE
802.3. There is difference in the header. DIX
uses TYPE, 802.3 uses LENGTH. SInce the frame is
limited in size, the two coexist. Most people use
the DIX format.
11Ethernet
- Work started back in 1973 by Bob Metcalfe and
David Boggs from Xerox Palo Alto Research Center
(PARC). - He studied the Aloha network and "fixed" the
mathematics. - Experimental Ethernet implemented in 1975.
- Cooperative effort between Digital, Intel, and
Xerox produced Ethernet Version 1.0 in 1980. - This also became known as the Blue Book
specification or DIX standard. Ethernet V2.0
adopted in 1982. - Ethernet was adopted with modifications by the
standards committees IEEE 802.3 and ANSI 8802/3. - Ethernet allows for only connectionless
communication.
12CSMA/CD
"Carrier Sense/Multiple Access with
Collision Detection"
"Driving in Boston"
BUS!
51.2 microseconds
"Many stations Listen before talking listen
while talking if a collision, backoff and try
again"
13Normal Ethernet Operation
14Ethernet Collisions
B
C
Collision
Data transmission for C
Data transmission for A
A
D
15CSMA/CD - A Simple Definition
- A network station wishing to transmit will first
check the cable plant to ensure that no other
station is currently transmitting (CARRIER
SENSE). - The communications medium is one cable,
therefore, it does allow multiple stations access
to it with all being able to transmit and receive
on the same cable (MULTIPLE ACCESS). - Error detection is implemented throughout the use
of a station "listening" while it is transmitting
its data. - Two or more stations transmitting causes a
collision (COLLISION DETECTION) - A jam signal is transmitted to network by the
transmitting stations that detected the
collision, to ensure that all stations know of
the collision. All stations will "backoff" for a
random time. - Detection and retransmission is accomplished in
microseconds.
16Frame/Packet Format
Preamble
SFD
Dst
Src
Type
Data/Pad
FCS
Size 7 1 6
6 2 46-1500
4 (octets) In IEEE 802.3, the Type field is
used as a Length field. Addresses are
generally (3) octets vendor code, (3) octets
device number.
17Ethernet Addressing
Each station recognizes three classes of
addresses. Own address Broadcast address (all
1's) Optionally, one or more multicast
addresses Major reason for broadcast is address
discovery. Multicast addresses are used for
specialized link layer functions.
18Ethernet Cable Names
Name
Fiber
Unshielded Twisted Pair
Thin coaxial
Thick coaxial
RG-8
Wire Type
22 - 26 AWG
62.5/125 micron
RG-58
10BASE5
10BASE2
10BASEF
10BASET
IEEE Name
N/A
Standard Number
IEEE 802.3
IEEE 802.3a
IEEE 802.3i
Other names
Thick net
Thin net
UTP
19Thick Coax Makeup
Center conductor of tin plated solid copper
conductor
Teflon is used for fire code regulations
20Thick Coaxial Connection
Pierce clamp
21Transceivers
- Transmitter/Receiver AUI on one side, media on
the other - Used on all Ethernet networks and is the device
that allows data to flow between the controller
card and the network. - Detects errors on the bus cable plant and reports
them to the station's controller card. - For thick coaxial cable, the transceiver is
external to the controller card and attaches
directly to the thick coaxial cable via a special
cable known as the transceiver cable. - External transceivers have a SQE function that
enables the controller to determine the status of
the transceiver. - Usually has status indicators (LEDs) physically
located on it to indicate the state of the
transceiver (transmitting, receiving, collision,
and power.)
22Thin Coaxial Cable Makeup
Polyethylene foam
Tinned copper wire
Jacket made of PVC or Teflon
EMI braided shielding
23Thin Coaxial Connection
Concatenation of network attachments
Direct connection to card
T connector
BNC connector at each cable end
24Thin Coaxial Connection (cont.)
AUI connector
T connector for connection to cable plant
BNC connector
Interface to computer bus
25UTP Makeup
- UTP was standardized by the IEEE 802.3 committee
in October of 1990. - Standardized by the EIA under TIA 568A.
- UTP for LANs is now classified as
- Category 3 - used for LANs up to 10 Mbps.
- Category 4 - used for LANs up to 16 Mbps.
- Category 5 - used for LANs up to 100 Mbps.
- Cable is made up of 8 strands of 24 AWG wire.
- Only 2 pair are used for single 10BASET
connection.
26Unshielded Twisted Pair
Repeater unit required
- Unshielded twisted
- pair cable
100m max cable run
Straight through pins 1, 2, 3 and 6
Unshielded twisted pair atleast two (2)
twists per foot
RJ-45 connector
RJ-45 Connector
8 pin
8 pin
27Concentrator (Hub) Management
- With the concentration of the wiring into a
common point, network managers can manage the hub
with specialized software. - Network management software resides not only in
the concentrator but on an external workstations
device (a PC, for example). - The workstation can query the concentrator for
information. - Concentrators also allow the control of
individual ports. - This software allows managers to extract
information from each card that is inserted in
the repeater. You could query the hub for
statistics such as - number of packets (bytes),
- number of collisions (single and multiple),
- number of framing errors,
- number of time the particular card de-inserted
itself from the network, - ability to turn on/off any repeater card in the
hub, and - all information is time and date stamped.
- With 10BASET, all information is provided on an
individual-connection basis, giving a manager
information right from the desktop.
28Ethernet Repeaters
- Extend the network by interconnecting multiple
segments - Extend the physical domain of the network
- Governed by the IEEE 802.3c working group
standard. - This governs the electrical specifications of a
repeater. - The physical configurations of a repeater varied
from vendor to vendor. - Some repeaters contain the intelligence to
- detect collisions per cable plant (will not
repeat collision fragments to other cable
plants). - de-insert themselves from a wiring concentrator
(when there are excessive errors on the cable
plant). - submit network management information to a
central controller. - Repeaters have been transformed into wiring
concentrators or hubs - Repeaters can be used to interconnect different
wiring types but not different access methods
(i.e., not Token Ring to Ethernet).
29IEEE802.3 Efficiency
"WARNING Opinion" Utilization
Status 0 - 10 Great! 10 -
40 OK 40 - 60
Performance Problems -- look at it 60
"Utilization"
Signal On
Time
30Token Ring - IEEE 802.5
What the IEEE standard covers History Differen
ces between 802.5 and 802.3
"Physical layer standard (gives link layer
format)"
Essentially an IBM standard 'given' to the
industry"
"Guaranteed response Priorities Controlled
delays"
31Token Ring History
- Presented by IBM in 1982 to IEEE 802 committee.
- First prototype developed in 1983 in Geneva,
Switzerland. - Cabling System was announced in 1984.
- Officially announced in 1985.
- Standardized by IEEE in 1985.
- Only one adopted by the IEEE 802.5 committee.
32Token Ring Technology Summary
- Access method by which network attachments gain
access to the cable plant by acquiring a special
frame called the token. Token is a special
24-bit pattern that continuously circulates the
ring. - Token Ring is a broadcast medium. To receive
data, a destination station performs an address
match. - The destination station merely copies the frame
as it repeats it back to the ring. - When the frame arrives back to the source
station, it strips the frame from the ring and
then releases the token (4 megabit operation
only). - The token is allowed to be released prior to
frame reception on 16-megabit rings. - Token Ring originally ran at 4 Mbps. Upgraded in
1989 to 16 Mbps - Maximum frame size for 4 Mbps is 4472.
- This is based only on the fact a station cannot
hold the token longer than 10 milliseconds. - Maximum frame size for 16 Mbps is 17,800.
33TRN Features
"data rate of 4 or 16Mbps"
Traffic usually (always in 802.5)
unidirectional RAR (802.5) vs RAT (FDDI) for
Token Passing Recovery from lost
token Priorities Frame Structure
"one frame on the net at a time..."
34Controller Attachment to a MAU
The IBM 8228 MAU
Shielded or UTP cable Lobe cables
35Cable Connectors
Hermaphroditic or RJ-45 connectors on MAU
DB-9 connector
MAU
Media filter for UTP only
RJ-11 or RJ-45 connector
Media filter can be on-board
36Multiple MAU Connection
37MAU Operation
Lobe cables
Relays
MAU top view
Ring out
Ring in
MAU bus
All stations are active
38MAU Operation (Inactive Station)
Lobe cables
Relays
Closed
Closed
Closed
MAU top view
Ring out
Ring in
MAU bus
Inactive station
39Token Ring Cable Types
- Type 1
- A shielded data grade cable with two solid wire
twisted pairs. - Available in indoor and outdoor versions.
- Type 2
- A Type 1 indoor cable with four solid twisted
pairs of 24 AWG wire. - Contains four voice grade wires along with four
data grade wires. - Type 3
- Unused existing telephone wire or EIA category 3
wire (4 Mbps operation). - Category 4 is needed for 16 Mbps (speed of the
Token Ring) operation. - Must use a special media filter.
- Type 5
- 100/140 micron fiber cable used for fiber optic
repeater links. - Type 6
- Often used for patch cables.
- Patch cables can be used for MAU-to-MAU
connection or from a wall outlet to a network
attachment.
40Type 3 Media Filter
- Type 3 cable requires a device known as a media
filter. - Its purpose is to filter out any unwanted
signals. - It is a small rectangular device that is usually
part of the UTP cable itself. - It can be a separate device that attaches to the
UTP cable at the end of the cable that attaches
to the controller card. - It can be used on 16- or 4-mb Token Rings.
- It is only used with Type 3 (UTP) cable.
41802.5 Framing
- IEEE 802.5 uses special characters, but does not
use bit stuffing!
Manchester
1 bit
0 bit
Violations!
42Token Ring Frames
1 byte 1 byte
43Token Ring Frame Field Definitions
no preset size
Routing Information Fields
IEEE 802.2
SD
AC
DA
Data
FCS
ED
FS
FC
SA
4 bytes
1 byte
1 byte
1 byte
1 byte
1 byte
6 bytes
6 bytes
lt 18 bytes
DSAP
SSAP
Control
Legend
1 or 2 bytes
1 byte
1 byte
- SD - Starting Delimiter
- AC - Access Control
- FC - Frame Control
- DA - Destination Address
- SA - Source Address
- FCS - Frame Control Sequence
- ED - Ending Delimiter
- FS - Frame Status
44The SD and the AC Fields
45The FC, ED, and FS Fields
Field
Bit 0
Bit 7
FF - indicates a MAC or LLC frame. ZZZZ -
indicates the type of MAC frame.
FC
F F r r Z Z Z Z
I - Intermediate bit
ED
J K 1 J K 1 I E
E - Error bit
A - Address recognized bits
FS
A C r r A C r r
C - Frame copied bits
46Bit Order Transmissionfor Token Ring
- Bit 0 is the first bit transmitted.
- Bit 0 is the left most bit of the byte.
- Unlike Ethernet, the bits in the bytes are not
reversed as they are transmitted. - Example
- 40-00-12 are the first three bytes of a MAC
address. - Translated to binary
- 01000000-00000000-00010010
- As transmitted on a Token Ring
- 01000000-00000000-00010010
- Compared to Ethernet transmission
- 00000010-00000000-01001000
47Token Passing Policies (Defn)
- Multiple Token
- RAT (FDDI) free token is appended to tail of
last packet - Single Token
- ? Token is released upon receipt of leading edge
of own packet - Single Packet
- RAR (802.5)Token is released upon receipt of
trailing edge of own packet
48Token Passing Policies (Usage)
- Multiple Token
- Allows multiple packets on the segment at one
time. Good when packet length is less than ring
latency - Single Token
- More efficient than RAR when packet length is
about the same as ring latency - Single Packet
- Least efficient, but allows controlling station
knowledge of (un)successful transfer before the
token is released (see pg. 224, 1st paragraph)
49Token Passing Policies (Perf.)
- Multiple Token
- Always the best performer, but more complex
- Single Token
- Closer to RAR than RAT
- Single Packet
- Worst performance
- KEY POINT Ratio of ring latency to packet
length, a, is real determiner of performance.
For a ltlt 1, RAR is OK.
50Controller Operation - Phases 0 and 1
- Five-phase initialization
- Phase 0 - Lobe test
- The controller transmits frames between the
controller card and the cable attached between
the controller card and the MAU. - The controller tests to ensure that the lobe
cable can successfully transmit and receive
frames. - Phase 1 - Monitor Check
- Station inserts into the ring (flips the relay in
the MAU) and looks for special frames that are
transmitted by the monitors. - Sets a timer to wait for these frames.
- If the station does not receive any of the
frames, the controller assumes - it is the first ring station on the network,
- there is not an Active Monitor present, or
- inserting into the ring disrupted the ring.
- The controller may initiate the token claim
process.
51Controller Initialization - Phases 2, 3, and 4
- Phase 2 - Duplicate address check.
- Checks to ensure that it can successfully
transmit and receive a frame and to detect other
stations that might have the same MAC address. - The controller transmits a frame to itself.
- If the frame returns with the address recognized
bit set, it notifies one of the monitors and
removes itself from the ring. - Phase 3 - Participation in neighbor notification.
- The station transmits a special frame that will
identify itself to its downstream neighbor. - The station should receive a similar frame for
its upstream neighbor. - Phase 4 - Lan Network Manager Notification
- Notifies LAN Network Manager about its presence
on the ring
52Claim Token Process
- A ring cannot operate without a token circulating
on the ring. - There is only one token per ring.
- The token-claiming process allows one station to
insert the token onto the ring. - This station will be elected as the AM.
- It will purge the ring (ability to transmit a
frame to itself). - After purging the ring, it will insert a new
token on the ring. - The Token-Claim process can be started when the
AM - detects a loss of signal,
- a timer expires and it has not yet received its
AM frame back, or the AM - cannot receive enough of its own Purge Ring MAC
frames. - It can be started when the SM
- detects loss of signal or
- detects expiration of its timer for receiving SM
frames.
53Details of the Claim Token Process
- If there is no token on the ring, all activity
will cease on the ring. - The Active Monitor should be able to recover by
purging the ring and issuing a new Token. - If the Active Monitor cannot recover, the
token-claim process will begin. - Any station will insert its master clock, a
24-bit delay, and start to transmit Token-Claim
frames. - These frames are received by all stations on the
ring. - The station will follow these frames with idle
(clock) signals. - After transmitting the Token Claim frames, the
station starts a timer. - If it does not receive its frames or someone
elses claim frames, it will beacon the ring. - Once the process is started other stations may
participate. - Stations bid for the right to become the AM.
- The station with the highest priority (MAC
address) wins. - That station becomes the AM.
- It will purge the ring and insert a new token.
54Claim Token Process Example
55Token Ring Transmit Mode
- A station that needs to transmit receives the SD
of approaching frame. This station quits
transmitting idles (clock signals). - Checks for priority.
- If the priority in the frame is greater than the
station's priority, then - the station sets reservation bits and awaits new
token. - If the priority in the frame is less than or
equal to the stations priority then - the station changes the T bit in the AC field
from a 0 to a 1, - appends its information to the rest of the frame
and transmits the frame. - If the end of its transmission is reached and it
has not received its current transmission back,
the station - transmits idle characters and awaits current
transmission. - When the station receives its frame back it will
strip the frame and release the token. - The station enters normal repeat mode.
56Token Ring Copy Mode
- The destination Token Ring controller recognizes
its address in the destination field of a
received frame and copies the frame into its
buffer. - If at any time an error is detected, the copy
phase ends and the controller sets the A and E
bits and repeats the frame back to the ring. - If no errors are found, the destination sets the
A and C bits and repeats the frame back to the
ring. - The destination station enters Normal Repeat
mode. - The frame travels on the ring until it reaches
the originator and that station strips the frame
off of the ring and submits the token to the ring.
57Normal Repeat Mode
- A station in normal repeat mode checks current
frames and token for signalling errors. - If any errors are found the station sets the E
bit and repeats the frame back to the ring. - A station in this mode also checks every frame
for its address. - A duplicate address could be found.
- If a duplicate address is found, the station will
transmit a soft error MAC frame to one of the
monitors.
58The Active Monitor (AM)
- Functional address is C00000000001.
- It must be present in order for the ring to
function properly. - The AM is the kingpin of the ring.
- The AM
- tracks lost tokens and ensures that only one
token exists on a single ring. - monitors frames and priority tokens that
circulate the ring more than once. - initiates neighbor notification,
- provides a latency buffer to recover the clock
signal and so that at least 24 bits (the size of
the token) can be transmitted on the ring, and - supplies the master clocking .
59Token Recovery
- Monitor Station
- 1 station becomes responsible for monitoring the
token for token loss or token busy - Time Outs
- Token time out (Beaconing)
- No monitor (Claim frames (highest addr wins)
60Options for Token Ring
- For 16 megabit rings, early token release allows
a ring station to release the token before
receiving its original frame back. - It is based on the ring length
- A station will not release the token when it is
still transmitting its frame and it has started
to receive its frame back. - Allows greater use of Token Ring bandwidth.
- Token Ring operates at 4 and 16 Mbps.
- 4 and 16 Mbps controllers are not allowed on the
same ring. - Ring will beacon when this condition occurs.
- To have 4 and 16 Mbps ring interoperate, you must
use a data forwarding device such as a bridge or
a router. - IBM is currently experimental with a new Token
Ring controller which allow it to operate between
52 - 100 Mbps.
61Data Link Layer
Uses 'bit pipe' Physical Layer to send
packets Packet Formats - Generic Framing
(Layer 1), Addresses and control information
(layer 2), and data (info from layer 3 and
up) Point-to-Point vs Broadcast - Key idea is
that not all packet formats are alike. One needs
to look at particluar technologies to see what is
needed.
62Data Link Services
- Unacknowledged Connectionless Service
- Most LANs
- Upper layers handle error recovery
- Acknowledged Connectionless Service
- Odd duck. Example?
- Connection-oriented Service
- Reliable Delivery ...
63Link Protocols
Used to provide reliability. Basic idea can be
used at any layer ABP SRP GoBack
N Windowing Flow Control
Don't need to know details at this time, but know
general operation and that they provide assured
delivery.
64Performance
- Overhead vs Frame Length
- Error rate (bit error vs block error)
- Physical Layer
- distance
- propagation delay
65Error Control
Error Detection - Methods Parity, Checksum, CRC
-- generically Frame Check Sequences Error
Correction - The basic idea is to add redundant
information so that the receiver can deocde the
message even if some (specified) number of bits
are damaged (e.g., Hamming codes) Error
Recovery includes error correction but also
includes actions taken to get a message
retransmitted
66Connection Oriented Services
- Two modes of operation
- Operational
- Non-operational
- Operational mode incorporates three functions
- Link establishment.
- A source station sends a frame to a destination
station requesting a connection. - The destination station may accept or reject the
connection request. - Information transfer.
- Allows information to be transferred after a
connection is set up and the required handshaking
has taken place. - Reliable information is transferred between the
two stations. - Link termination.
- Either side of the connection may terminate the
connection at any time.
67IEEE OSI again
LLC
2
MAC
1
PHY
LLC Logical Link Control MAC Media Access
Control PHY Physical
68IEEE 802.2 Fields
Bit 0
I/G D D D D D D D D
C/R S S S S S S S
Length of the Information field is access method
dependent
SSAP address
DSAP address
Control
Information
1 byte
1 byte
1 or 2 bytes
Source address
Length field
Destination address
IEEE 802.2 field
CRC
69SAP Types
- E0 - Novell NetWare
- F0 - NetBIOS
- 06 - TCP/IP
- 42 - Spanning Tree BPDU
- FF - Global SAP
- F4 - IBM Network Management
- 7F - ISO 802.2
- 00 - NULL LSAP
- F8, FC - Remote Program Load
- 04, 05, 08, 0C - SNA
- AA - SNAP
- 80 - XNS
- FE - OSI
70SubNetwork Access Protocol (SNAP)
- Most common implementation of LLC1 is from a
subsection of the IEEE 802.2 standard known as
SNAP. - At the time of IEEE 802.2s introduction, most
network protocols were designed to use the
Ethernet packet format. - SNAP allows for the migration of the standard
network protocols to the IEEE 802.2 format. - Supported by TCP/IP, NetWare, OSI, AppleTalk, and
many other protocols. - The second purpose for the SNAP protocol is to
allow those protocols that do not support the
IEEE 802 standard to be able to traverse IEEE 802
LANs. - SNAP uses a reserved SAP AA (for both the DSAP
and SSAP). - It uses the unnumbered frame format control
field equal to 03. - Actual SNAP header consumes 5 bytes
- Three bytes for the Organizationally Unique
Identifier (OUI) field, and - Two bytes for an Ethernet Type field.
71Protocol Discriminator
SNAP header
Length field
Source address
Destination address
SSAP
Control
Data
Pad
CRC-32
DSAP
AA
AA
03
Type field
OUI
Protocol discriminator
00-00-00
08-00
3 bytes
2 bytes
72Verification
- Finite State Machines
- Estelle Other Languages
- Petri Nets
- Blind Faith (or, code it in C...)
73Naming Conventions
and Confusion
74Naming Conventions cont
75Intro to ATM
- Asynchronous Transfer Mode
- Text References
- Sect 2.6
- Sect 3.6.3
- Sect 5.6
- Sect 6.5
76ATM Background
- Outgrowth of TELCO transition to integrated
services - Only real gt100Mbit standard
- Offers multiservice (voice video data) potential
- Switched architecture familiar to TELCOs, not to
high speed data networks
77What is ATM?
Note Tanenbaum considers this more a network
layer technology.
78ATM - A layered standard
AAL - ATM Adaptation Layer Assembles
and disassembles broadband servicesinto
a stream of cells Each cell has a
header that contains routing information ATM -
Asynchronous Transfer Mode Switches the
cells around the network based on the routing
information in the header Physical Layer
Provides the physical transportation of
cells across the network (Note CCITT
reference model, p. 63)
79ATM - A Switched Architecture
- Cells (small, fixed length packets) are switched
in a connection-oriented manner but not using
circuits like todays voice.
Switch
Switch
Edge Device
Edge Device
80What is ATM Switching?
- Why small cells?
- (3264)/248 5 header bytes
- Mixed Traffic
- Packet (random)vs Circuit (TDM) Switching
- Q.2931
- SVC, PVC
81Physical Layer Options
- SONET (US)/ SDH (Europe)
- SMDS
- DQDB
- Speeds from DS3 on up! (45Mbs to Gbps)
- OC-3c gt 155.52Mbps gt 149.76Mbps
- optical carrier
- 3rd level in heirarchy
- full duplex (two strands of fiber)
- Also OC-12c (622Mbps), OC-48c (2048Mbps)
- Look at the interesting way to frame cells
82ATM Adaptation Layer(AAL)
Classes of Service 1, 2, 3/4, 5 1 circuit
emulation 2 variable bit rate service 3/4
connection oriented data service 5
connectionless data service SAR - Segmentation
and Reassembly Convergence Sublayer the
miscellaneous category
83ATM Cell
ATM cells are constant size packets of 53 bytes
size. -- 48 bytes payload, 5 bytes
header/overhead.
VPI - Virtual Path ID VCI - Virtual Channel
ID Type - Payload type (internal) Res -
reserved CLP- Cell loss priority HEC- Header
Error Control
84VCI/VPI Operation
A Virtual Channel exists between two
switching points
A Virtual Path contains 'bundles' of VCs
85ATM Switch Architecture
- Crossbar
- Banyan
- TDM busses
- Buffering
- Input
- Output
- Both?
86ATM Protocols
- UNI, NNI
- Services
- LAN Stuff
87ATM Services
- CBR
- VBR (RT, NRT)
- UBR
- ABR
88ATM Quality of Service
- QoS A contract
- Traffic Descriptors
- Cell Rate Options (pg 462)
- Traffic Shaping
- Traffic Policing
89ATM Congestion Control
- Admission Policy
- Reservation System
- Rate Based Control
- Other
90ATM Flow Control
The leaky bucket algorithm CLP in ATM
header Frame Relay comparisons
91Routing
- IISP (Interim Inter-switch Signaling Protocol)
- PNNI (Private Network-Network Interface)
- Phase 1
- Phase 2
92IISP
- Interim
- Allowed multi-vendor interoperability before
completion of NNI - Signaling
- Routing via manually configured NSAP prefixes
93PNNI
- Topology abstraction
- Peer group(group of nodes)
- One switch elected Peer Group Leader
- All nodes in group have identical view of group
- Hierarchy of logical groups
- Up to 105 levels of hierarchy
94PNNI Routing
NSAP Domain
A12
A2
B
A11
A13
View from A117 at A11
95Sequence of Events
- A117 -gt B25
- Forward to switch (A11)
- Switch knows topology of A1 group
- B reachable by A2 - A2 reachable by either A12 or
A13 - DTL (Designated Transit List)
- A12A2B
- A22A23B
- B2
96ATM LAN Stuff
- LAN Link Layer Domain
- ELANs VLANs
- LANE MPOA
- LECS, LES, BUS
97LANE v1
- LAN Emulation
- No QoS (Quality of Service) Support
- Uses AAL5 signaling
- optimized for data transport
- entire cell payload available for user data
- LEC - LAN Emulation Client
- LAN Emulation Service
- LECS - LAN Emulation Configuration Server
- LES - LAN Emulation Server
- BUS - Broadcast and Unknown Server
- STP (Spanning Tree Protocol) supported
98LEC - LAN Emulation Client
- Software process on any ATM-connected LAN switch,
router, PC, or workstation - Layer 2 process
- Prior knowledge of certain parameters
- LECs ATM address
- LAN type to be emulated
- maximum data frame size
- any route descriptors (for SR bridging)
- whether it is willing to proxy (respond to
LE-ARP) - LAN name - SNMPv2 display string
99LECS - LAN Emulation Configuration Server
- One per administrative domain
- Gives identity of ELAN (Emulated LAN)
- Returns ATM address of LES, type of LAN emulated,
and maximum PDU size of ELAN - Controls which physical LANs are combined to form
VLANs (Virtual LAN) - LECS address known via ILMI or its well-known
NSAP address
100LES- LAN Emulation Server
- Adds LECs to ELAN
- Assigns LECID to joining LEC
- Table of address information of LEC
- MAC address
- proxy for MAC address
- Token Ring route descriptors
- LECs can communicate directly with each other
only when they are connected to the same LES - Multiple LESs on the same physical ATM LAN
- Answers LE-ARP requests from LECs
101BUS- Broadcast and Unknown Server
- During address resolution LEC forwards all frames
to the BUS - floods frames to all LECs
- after address resolved flush protocol used to
guarantee order of cells - All multicast and broadcast traffic sent through
BUS - Traffic limited to 10 frames/second
- Intelligent BUS
- resolve destinations
- CLS- connectionless server
102LANE Setup
103Connections
- All SVC (switched virtual circuits)
- SVCs required
- LECs and LECS
- LES and LECS
- Control Direct - LECs and LES
- pt-mpt Control Distribute - LES to LECs
- Multicast Send - LECs and BUS
- pt-mpt Multicast Forward - BUS to LECs
- Data Direct - LEC and LEC
- PVC (permanent virtual circuit) possible to
connect LEC and LECS
104Virtual Channel Connections
LANE Server (LES)
Broadcast and Unknown Server (BUS)
Control Direct VCC
Control Direct VCC
Multicast Send VCC
Multicast Send VCC
LANE Client (LEC)
LANE Client (LEC)
LANE Client (LEC)
LANE Client (LEC)
Control Distribute VCC
Multicast Forward VCC
LAN Switch
Data Direct VCC
LAN Switch
ATM Host
ATM Host
Configuration Direct VCC
Configuration Direct VCC
LANE Configuration Server (LECS)
105NHRP
- Next Hop Resolution Protocol
- Grew out of ATMARP
- Only IP
- Allows shortcut routes (pt-pt)
- direct VCCs across ATM network
- Address resolution across multiple IP networks
- If network unknown, request forwarded to other
NHSs (Next-hop Server) - NHS with knowledge will forward response to
source router - Router must have ability to bypass default route
106RSVP
- Resource Reservation Protocol
- Provides QoS (Quality of Service) guarantees
- Operates in simplex
- each direction has separate reservation
- maps well to ATM (two individual VCCs)
- Built on IP, but no data transport built-in
- Only if resources available and does not conflict
with policy - Flowspec (bandwidth and delay) and filterspec
(type of packets) transmitted downstream - hop by hop
107MPOA
- Multiprotocol over ATM
- EDFG (Edge Device Functional Groups)
- existing LAN segments via LAN switches
- AHFG (ATM-attached Host Functional Groups)
- ATM-connected host
- Layer 3
- Only supports IP for now
- Uses LANE for Layer 2 forwarding within a single
Layer 3 subnet - Adaptation of NHRP to provide connectivity
between hosts in different subnets
108MPOA Operation
109Competing Technologies
- Fast Ethernet
- 100BASE-TX, 100BASE-FX,100BASE-T4, 100BASE-VG
- FDDI, FDDI- II
- HPPI
- Gigabit Ethernet (IEEE 802.3z)
110ATM Issues
- SONET/SDH duplication of services
- ATM overhead
- ATM granularity and bandwidth management
- ATM connectionless service
- End point synchronization
- Flow Control !!! (bandwidth allocation,
correlated traffic) - ATM Forum
111Internetworking
- Bridges
- Transparent bridges
- Source Routing - Transparent Bridges
- Routers (Network Layer)
- Brouters
3
2
2
2
1
1
1
1
112Why Bridges
- Isolation of Physical Layer Effects
- Bandwidth Multiplication
- Security or Traffic Isolation
113Segmenting Traffic
114Transparent Bridges
- Interconnect multiple cable segments to allow for
extension of a network. - Can be used to interconnect different access
methods (Ethernet to Token Ring) and different
physical layers. - Operate at the data link layer.
- They are protocol transparent.
- They are designed to operate regardless of the
upper-layer protocol. - They operate on the source and destination
address in the MAC header.
115T-L-F Bridges
- Bridges only forward traffic destined for other
cable segments. - They operate transparently to any stations that
are active on the network. - Packet formats and software drivers on the
workstations remain the same. - Bridges do not have to be programmed with the
addresses of all the devices on the network.
116Learning, Filtering, and Forwarding
117Filtering - An Example
118Forwarding - An Example
119Forwarding Beyond One Bridge
120Loops
- Complexity of bridging arises when two or more
bridges interconnect the same two cable segments. - This is called providing redundancy or providing
a loop. - There are problems with this type of design
including - duplicate packets,
- broadcast packets, and
- unknown destination packets.
121Duplicate Packets
122Broadcasts
123Unknown Destination Address
124Spanning Tree Algorithm
- Bridged networks must allow for redundancy. Only
one path should be enabled to any destination on
the network. - STA is a protocol unto itself. Dont confuse it
with the transparent bridge protocol. IEEE
802.1d - In an active STA topology certain bridges are
allowed to forward packets. - Other bridges will participate in the STA but do
not forward packets. - These are backup bridges that dynamically become
available. - Bridges that do not forward packets are placed in
blocking mode. - These bridges still participate in the spanning
tree protocol.
125Source Routing Bridges
- Developed as a bridge protocol for Token Ring
LANs. - Source routing gained popularity due to IBMs
support of it. - It is easy to install a source route network.
- It is not easy to grow a source route network
into a large network. - Invented due to technical limitations of the
source route chip set.. Early source route chip
sets could not be set for promiscuous mode. - Source routing was also invented to allow two
non-routing protocols to be placed on a LAN
NetBIOS and SNA. - Source Routing does not build forwarding tables
based on MAC addresses. - Most of the intelligence for this algorithm is
found in the network stations. - Each frame carries complete route information
with it.
126Source Routing Features
- Source routing requires split intelligence to be
carried in the node and the bridge. - All frames contain routing information, which
does produce more overhead. - Uses STA to configure which bridges will forward
single route broadcast frames. - All paths are active which legally allows loops
to be designed. - Provided a routing solution for those protocols
that could not be routed (NetBIOS). - Easy to follow ring/MAC address for
troubleshooting.
127Source Routing Features (cont.)
- Source Routing originated as an alternative to
transparent bridging - Originally, Token Ring could not be placed in
promiscuous mode ( requirement for transparent
bridging) and therefore an alternative model was
created - Allowed for SNA and NetBIOS traffic an attempt to
enjoy the benefits of routing - As a data link layer implementation.
128Source Routing Overview
- Each separate ring is assigned a unique ring
number, assigned on the source route bridge port
and not on the ring station. - Each bridge is assigned a bridge number. There is
a single number for the whole bridge, no matter
how many ports it has. - End stations try to find destination ring
stations by broadcasting special discovery
frames. - A frame will contain source route information
based on one bit in the source address. - A source route frame may not cross more than
seven bridges. - At the eighth bridge, the frame is discarded.
129Source Routing Example
MAU
MAU
2
Find a station off ring
Bridge 5
Node 1
Node 2
Bridge 6
1
Find a station on the local ring
Bridge 7
Ring 4
Ring 3
130Routing Information Field
131The Route Designator
Bridge 1
Discovery frame
Ring B
Ring A
RC
RD1 RD2
RC
Token Frame Header
Token Frame Trailer
Token Frame Header
Token Frame Trailer
Routing Control
Routing Control
00B1 00A0
Routing Information Field
Routing Information Field
132Source Route Frame Types
- Four types of Source Route frames
- Single Route Explorer (SRE)
- Also known as Spanning Tree Explorers (STE)
- So named by the IEEE 802.5 working group
- All Routes Explorer (ARE)
- Specifically Routed Frame (SRF)
- Single Route Explorer with a specific route
return.
133Token Ring to Ethernet Conversion
134Ethernet to Token Ring Conversion
Copy and bit reverse
Ethernet frame
Type
FCS
DA
SA
Info
Preamble
Copy
SD
AC
FC
DA
SA
RIF
DSAP
SSAP
CTRL
Type
Info
FCS
ED
FS
OUI
Insert
SNAP header
Token Ring frame
135Token Ring to IEEE 802.3 Conversion
136IEEE 802.3 to Token Ring Conversion
Copy and bit reverse
IEEE 802.3 frame
FCS
DA
SA
Info
Preamble
SFD
Length
DSAP
SSAP
PAD
CTRL
Cut Insert
Copy
SD
AC
FC
DA
SA
RIF
DSAP
SSAP
CTRL
Info
FCS
ED
FS
Token Ring frame