Title: Chapter 5
1Chapter 5 Protocol Architectures
2Protocol ArchitectureThe IEEE Protocol Reference
Model
- Introduction
- At the lower layers of the Local Area Networking
world there are two protocols models that
predominate the IEEE 802 model and the OSI
reference model - Comparison of the OSI reference model to the IEEE
802 reference model
3Protocol ArchitectureIntro to the IEEE Protocol
Reference Model
- Physical Layer functions
- Encoding/decoding of signals
- Synchronization
- Bit transmission/reception
- Data Link Layer functions
- Provide one or more addressable service access
points (SAPs) - On transmission create a frame containing data,
addressing, and control/error-detection fields - On reception disassemble frame and check validity
of addressing and error detection - Control access to the physical transmission medium
4Protocol ArchitectureIntro to the IEEE Protocol
Reference Model
- In the IEEE 802 reference model the data link
layer functions are split into two sublayers
the Medium Access Control sublayer and the
Logical Link Control sublayer - This separation allows generic higher level
functions to be grouped into a common Logical
Link Control that can be used with different
physical layer specific MAC layers - The Logical Link Control sublayer provides the
addressing and is responsible for managing access
to a shared medium a function not specified for
the data link layer in the OSI model - The MAC handles transmission and reception of
frames and controlling physical access to the
medium
5Protocol ArchitectureLogical Link Control
- Principles of Logical Link Control
- Concerned with the transmission of a link-level
PDU between two stations with no intermediate
switching nodes - Where the LLC sublayer fits into the protocol
stack Figure 5.2 - Different from most other link control protocols
for two primary reasons - It must support a shared, multi-access medium
- It doesnt have to be concerned with many medium
specific functions handled by the MAC sublayer
6Protocol ArchitectureLogical Link Control
Principles
- Four fundamental services are provided to upper
layers - Connectionless Service
- Connection-oriented Service
- Multiplexing of logical connections
- Multicasting broadcasting
- Addressing in the frames is important for all LLC
services!
7Protocol ArchitectureLogical Link Control
Addressing
- As mentioned earlier, the process of transferring
data between networked stations can be broken
into two steps - Getting the data to the destination station
- Once at the station, getting the data to the
destination process within the station - Such a separation in the process leads to a
two-level addressing scheme - The MAC address uniquely identifies a physical
interface from a LAN to the station multiple
interfaces are OK for a station but they have
unique addresses - The LLC address uniquely identifies a LLC user or
process executing on the station
8Protocol ArchitectureLogical Link Control
Addressing
- In addition, group addresses are important
- Broadcast addresses refers to all interfaces
within a certain context - Multicast addresses refer to a subset of
interfaces within some context - In the context of a group address the interface
considers the data as addressed to itself - Possible combinations of individual and group MAC
LLC addressing Figure 5.5 - In many cases well-known addresses are used for
specific functions (network control)
9Protocol ArchitectureLogical Link Control
Addressing
- An example LLC addressing scenario
10Protocol ArchitectureMedium Access Control
- MAC Techniques
- The key parameters of any MAC technique are where
how - Where is control centralized or distributed?
- In a centralized scheme a station must receive
permission from a controller before transmitting
while in a decentralized scheme the stations
collectively follow a procedure to determine who
is allowed to transmit - Centralized schemes allow greater control, allows
the use of simple station logic, and avoids
ambiguous states that may occur with distributed
coordination - Centralized schemes can have a single point of
failure and also become a performance bottleneck
11Protocol ArchitectureMedium Access Control
Techniques
- How is the access control synchronous or
asynchronous? - Synchronous schemes can be very inefficient in
data transmission - Bandwidth allocated even if it is not used
- Three basic asynchronous schemes
- Round robin
- Reservation
- Contention
12Protocol ArchitectureAsynchronous MAC Techniques
- Round robin each station in turn is given the
opportunity to transmit - Can be used in a centralized (polling) or
decentralized system - Most efficient when many stations transmit
streaming traffic - Reservation each station reserves in advance an
opportunity to transmit - Most efficient with streaming traffic
- Can be used in a centralized or decentralized
control environment - Contention a station must win the medium to
transmit - Best for bursty traffic
- Simple to implement and very efficient with light
loads system throughput can collapse under heavy
load - A strictly distributed scheme most often found in
LAN products
13Protocol ArchitectureMedium Access Control Frame
Format
- The exact frame format varies between specific
MAC protocols, Figure 5.6 shows a very general
format - Control field contains any fields needed to
control the operation of the MAC protocol or
implement some of it functionality (e.g.
acknowledgements, priority levels) - Destination MAC address
- Source MAC address
- Payload usually the LLC Protocol Data Unit
encapsulated within the MAC frame - CRC (Cyclic Redundancy Check)
14Protocol ArchitectureMAC Frame Format
- CRC (Cyclic Redundancy Check) a special code
mathematically generated from the rest of the
frame which can detect and sometimes correct
errors - The sender generates the CRC and inserts it in
the field - Upon reception the receiver also generates the
CRC and compares it the CRC in the received frame - If they are the same the frame was received
without error, otherwise there is at least one
bit error in the frame - With the IEEE 802 LAN protocol architecture,
error detection is the responsibility of the MAC
sublayer error tracking and recovery is the
responsibility of the LLC sublayer
15Protocol ArchitectureMAC Frame Format
- Examination of Figure 5.7 encapsulation in the
TCP/IP and IEEE 802 protocol stack to put the MAC
and LLC layers in context
16Protocol ArchitectureBridges and Routers
- Introduction
- It is very common for organizations to require
interconnection of LANs - Two systems used for this purpose bridges
routers - Bridges
- Bridges are typically designed to connect
together LANs that use the same MAC protocols
(this keeps them simple) - Advantages to using bridges over a single large
LAN - Reliability minimize single points of failure
- Performance increases aggregate throughput
- Security partitions sensitive data
- Geography may be physically impossible to put
everyone on the same LAN
17Protocol ArchitectureBridges
- Design characteristics of bridges
- Bridges make no modifications to the control data
and to end stations they believe they are on one
large LAN - Frames must have addressing and routing
intelligence to know which interface to transmit
a frame on - Bridges may connect more than two LANs
18Protocol ArchitectureBridges
- Protocol Architecture diagram for a bridge
Figure 5.9a
19Protocol ArchitectureRouters
- Routers are general-purpose devices used to
connect dissimilar LANs by operating as a network
layer relay - Routers must deal with the following issues
- Different Addressing formats
- Different maximum (and minimum) frame sizes
- Different physical interfaces
- Different levels of reliability
- Other differences in functionality, including
priority and quality of service mechanisms - To use routers all systems on the connected
networks that wish to communicate must share a
common network layer - In the TCP/IP protocol stack IP is the common
glue - IP is a best effort service it imposes no
reliability requirements on the lower protocol
layers
20Protocol ArchitectureRouters
- Protocol Architecture diagram for a router Fig.
5.9b - It is possible and indeed very common to find
mixed networks of bridges (switches) and routers
21Protocol ArchitectureOther Networking Devices
- Other devices besides routers and bridges can be
used for network interconnection - Another possible but less common method for
connecting together networks is to use a gateway - Gateways usually operate at the application layer
allowing completely different network protocol
stacks to interoperate - Gateways are used in security applications to
provide a high degree of isolation - Hubs are essentially multiport repeaters with
very little intelligence all that is necessary
is the proper propagation of the collision
detection signals
22Protocol ArchitectureOther Networking Devices
- As mentioned earlier, a LAN switch is similar to
a bridge but has additional capabilities - LAN switches usually perform frame forwarding in
hardware instead of software for maximum
performance - LAN switches can forward multiple frames
simultaneously bridges can only forward one at a
time - LAN switches have two modes of forwarding they
can perform both cut-through and
store-and-forward forwarding - LAN switches are rapidly replacing bridges in
most networks
23Protocol ArchitectureAppendix The IEEE 802
Standards
- Introduction
- A widely known and followed set of LAN standards
has allowed networking to proliferate by
fostering lower equipment prices promoting
interoperability - The IEEE 802 committee was established to develop
a set of LAN protocols these protocols have
since been adopted by ANSI and ISO to foster
truly international standardization - The committee is focused on Data Link Physical
layer standards for LANs
24Protocol ArchitectureAppendix The IEEE 802
Standards
- The IEEE 802 has broken down responsibility for
standards development into functional areas as
follows - 802.1 Higher Layer LAN Working Group
- 802.2 Logical Link Control Working Group
(inactive) - 802.3 Ethernet Working Group
- 802.4 Token Bus Working Group (currently
inactive) - 802.5 Token Ring Working Group
- 802.6 MAN Working Group (DQDB technology -
inactive) - 802.7 Broadband Technical Advisory Group
(inactive) - 802.8 Fiber Optic Technical Advisory Group
- 802.9 Isochronous LAN Working Group
- 802.10 Security Working Group
- 802.11 Wireless LAN Working Group
- 802.12 Demand Priority Working Group
(100VG-AnyLAN) - 802.14 Cable Modem Working Group
- 802.15 Wireless Personal Area Network Working
Group - 802.16 Broadband Wireless Access Study Group
25Protocol ArchitectureAppendix The Cyclic
Redundancy Check (CRC)
- An error detection mechanism is important for
data link layer protocols to ensure that reliable
communication is possible by notifying higher
layers when they must retransmit data - Commonly accomplished by adding a frame check
sequence to the data frame this is used as an
integrity check by the receiver - One of the most useful frame check sequences is
the Cyclic Redundancy Check (CRC)
26Protocol ArchitectureAppendix The Cyclic
Redundancy Check (CRC)
- How the CRC Works
- Based on a common binary arithmetic principles
- Sender Receiver agree on a common divisor
polynomial C(x) - What is transmitted is the n-bit message plus a
set of k check bits the total nk bits are
exactly divisible by C(x) - Heres the process the sender uses
- Multiply M(x) by xk (add k zeros on to M(x)) to
create M(x) - Divide M(x) by C(x) and find the remainder
- Subtract the remainder R(x) from M(x) forming
S(x) - S(x) is actually the n-bit message followed by
the k-bit CRC - The receiver of the message verifies message
integrity by - Dividing S(x) by C(x)
- If the division equals zero there is no
detectable error in the transmission - A non-zero result indicates errors
27Protocol ArchitectureAppendix The Cyclic
Redundancy Check (CRC)
- How the CRC Works
- Careful choice of the divisor polynomial
determines what errors can be caught - Undetected errors are bit errors that caused the
received message to be evenly divided by the
divisor polynomial C(x) - By choosing a good prime number of sufficient
length almost all errors can be detected with
very little overhead Table 5.1 - All of the IEEE 802 MAC frames use a 32 bit CRC
(as known as CRC-32) - C(x) 100000100110000010001110110110111 for
CRC-32
28LAN Addresses and ARP
- IP address drives the packet to destination
network - LAN (or MAC or Physical) address drives the
packet to the destination nodes LAN interface
card (adapter card) on the local LAN - 48 bit MAC address (for most LANs) burned in
the adapter ROM - Why not use IP addresses?
- Portability
- Update ROM on reboot
- How about no MAC add.?
- Host receives every packet
- Overhead to host processor
29- Address allocation
- MAC address allocation administered by IEEE
- A manufacturer buys a portion of the address
space (to assure uniqueness) in chunks of 224
addresses - First 24 bits are fixed
- Manufacturer can use last 24 bits to produce NICs
with unique add. - Analogy
- (a) MAC address like student ID number
- (b) IP address like postal address
- MAC flat address ? portability
- IP hierarchical address ? NOT portable (need
mobile IP) - Broadcast LAN address 1111.1111
- Each IP node (Host, Router) on the LAN has ARP
module and Table - ARP Table IP/MAC address mappings for some LAN
nodes - lt IP address MAC address TTLgt
- lt .. gt
- TTL (Time To Live) timer, typically 20 min
30(No Transcript)
31- ARP Routing Packet Within a LAN
- Host A wants to send packet to destination IP
addr XYZ on same LAN - Source Host first checks own ARP Table for IP
addr XYZ - If XYZ not in the ARP Table, ARP module
broadcasts an ARP packet - lt XYZ, MAC (?) gt
- ALL nodes on the LAN accept and inspect the ARP
packet - Node XYZ responds with unicast ARP pkt carrying
own MAC addr - lt XYZ, MAC (XYZ) gt
- MAC address cached in ARP Table
32- ARP Routing Packet to Another LAN
- Say, route packet from source IP address
lt111.111.111.111gt to destination address
lt222.222.222.222gt - In routing table at source Host, find router
111.111.111.110 - In ARP table at source, find MAC address
E6-E9-00-17-BB-4B, etc
33Hubs
- Hubs are physical layer devices
- Operate on bits rather than frames
- Broadcast each incoming bit to all other
interfaces - Hubs are same as repeaters with additional
network management functionality - Multi-tier hub design
- Individual LANs connect to a 10BeseT hub via
point-to-point connections - All hubs connected to a backbone hub via
point-to-point connections - Entire topology is called a LAN
- Individual LANs ? LAN segments
- All LAN segments belong to the same collision
domain
34Backbone Hub
- Benefits of LAN segments connected via backbone
hub - Inter-domain communication among hosts
- Extends maximum distance using hubs as repeaters
- Multi-tier approach provides graceful degradation
- Backbone hub can disconnect a malfunctioning
segment - Limitations
- Independent collision domains are transformed to
one large collision domain - Overall bandwidth is limited by the bandwidth of
each LAN - Example 3 LANs each with 10 Mbps ? total
bandwidth lt 10 Mbps - If different departments use different LAN
technologies they cannot be connected through a
backbone hub - Restrictions on number of hosts, maximum distance
between any two hosts, and maximum allowable
number of tiers
35Bridge
- Bridge is a layer 2 device
- Examines frames rather than simply broadcasting
bits - Uses layer 2 destination addresses for forwarding
- Separate collision domains for each connected LAN
- Example
- Three LANs connected through a bridge
36Bridge
- Bridges overcome some of the limitations of a
hub - Permit inter-LAN communication while preserving
isolated collision domains - Bridges can interconnect different LAN
technologies - For example, 10 and 100 Mbps Ethernets
- No limit on the size of the LANs connected
through bridges - Functions of a bridge
- FilteringWhether a frame should be forwarded to
another interface or dropped - Forwarding
- Ability to determine a target interface
- Implemented through a bridge table
- A bridge table entry contains LAN address of a
node, corresponding bridge interface, and time
at which the entry was placed in the table
37Bridge
- Self learning
- A bridge table is populated automatically ? no
protocol needed - Table is initially empty
- Bridge forwards copies of a frame whose
destination address does not have an entry in the
table to all output interfaces to transmit to all
LAN segments using CSMA/CD - For each arriving frame, bridge stores
- Source address field
- Interface from where it arrived from
- Current time
- Aging time after which entry for a source are
removed if no frames are received with that
source address - Bridges are plug-and-play devices
- No involvement from system administrator to
configure it - Transparent bridges
38Bridge
- Problem with hierarchical LANs connected through
bridges - Failure of a bridge near the top of hierarchy
causes all LAN segments below it to be
unreachable - Not a fault-tolerant design
- Solution
- Allow multiple, redundant paths between LAN
segments - Greatly improves fault tolerance
- Problem frames can cycle and multiply within
connected LANs - Cycling and multiplying problem is solved through
a spanning tree protocol
39 Bridge Spanning Tree Protocol
Bridges determine a spanning tree Communicate
over LANs Result is a subset of original
topology that has no loops Using spanning tree,
bridges virtually disconnect appropriate
interfaces Avoids cycles and multiple copies of
frames Spanning tree algorithm is re-run when one
of the links in spanning tree fails
40Bridges vs. Routers
- Similarities
- Both are store-and-forward devices
- Both can isolate the collision domains
- Are interchangeable devices
- Differences
- Bridges operate at layer 2 while routers operate
at layer 3 - Bridges are plug-and-play devices whereas routers
generally require configuration - Different efficiency and computational
requirements
41Ethernet Switches
- Layer 2 (frame) forwarding, filtering using LAN
addresses - Switching
- Assume each host connected through a pair of
links - A-to-B and A-to-B simultaneously,
- No collisions
- No CSMA/CD needed
- Often individual hosts, star-connected into
switch - Ethernet, but no collisions!
42An Institutional Network Using Hubs, Ethernet
Switches, and a Router
- Switch and bridge has similar trade-offs
- Both will result in broadcast and flooding storms
- Router is needed at the edge to prevent these
storms from spilling outside
43Cut-Through Switching
- An alternative to store-and-forward switching
- For any packet switch router, bridge, or
Ethernet switch - Efficiently utilizes empty output buffers
- Cut-through switching
- Frame forwarded from input to output port without
awaiting for assembly of entire frame - Start forwarding packet before all of it has
arrived at input - Initial part must contain the destination
information to start forwarding - Slight reduction in latency
- Delay in gathering entire packet L/R where L is
packet length and R is transmission rate - This delay can be eliminated/reduced only when
output buffer becomes empty before all of the
packet has arrived at input
44Cut-Through Switching (Contd)
- Cut-through over store-and-forward
- Limited advantage
- Results in .12 msec (for 100 Mbps Ethernet) to
1.2 msec (for 10 Mbps Ethernet) reduction for max
Ethernet packet sizes only when output buffers
are empty