CIS4445 DATA COMMUNICATIONS

1
CIS-4445 DATA COMMUNICATIONS

• Chapter 7
• Wide Area Networks (WANs)

2
HOMEWORK - CHAPTER 7
Review Questions 1, 3, 4, 6, 7, 8, 15, 23, 26,
28
Exercises 1, 2, 3, 4, 8, 12, 13, 14, 15
3
Introduction

• Motivation for WANs
• Wide Area Network Routing
• Public Data Networks The X Series
• Internet Protocols
• Transport Protocols

4
Generalized Wide Area Network
X
Z
W
A
S
D
B
C
T
Y
5
Motivation for WANs

• Grew out of a need to connect many networks
• Connections often required protocol converters to
  translate from one protocol to another because
  the networks were different
• Cover large area
• Many sub-networks/hosts
• LAN protocols not appropriate
• Collision detection not appropriate distances
  too far
• Token passing not appropriate token takes long
  time to arrive back.

6
Motivation for WANs

• More-complex routing
• Many ways to get somewhere
• Need to handle node/link failure
• Need to handle congestion
• Need to buffer significant amounts of data while
  protocol decides what to do with packets
• Management of the WAN becomes an issue
• Generally not going to have one organization with
  control/responsibility for entire network

7
Motivation for WANs

• Sub-networks don't always use same protocols (may
  differ at any OSI level)
• Level 1 (repeater)
• level 2 (bridge)
• Level 3 (router)
• Determines which route to follow
• Level 7 (gateway)
• Entire protocol different
• e.g. compression schemes, encoding, encryption,
  connection protocols, etc.

8
Introduction

• Motivation for WANs
• Wide Area Network Routing
• Public Data Networks The X Series
• Internet Protocols
• Transport Protocols

9
Wide Area Network Routing

• Network protocols have to determine how to get
  the information from Point A to Point B
• A good starting point is to look at how routing
  was accomplished in smaller networks using
  Bridges at Layer 2
• Depends on many factors, e.g.
• Cost (money) of sending over link
• Time required (link speed)
• Congestion (current traffic over link)
• Not always simple shortest-path
• Need to think about bringing in higher layers of
  the OSI model to handle the tougher connection
  problems
• Chapter 6 - layers 1 and 2 only ones used

10
Generalized WAN w/ Connection Costs
D
B
5
2
4
A
3
7
F
4
1
7
2
C
6
E
11
Partial Routing Tables for Nodes

• Use similar approach as used in Chapter 6 with
• bridge routing

• Develop routing tables for each node in the WAN
  based on cost
• Not going to have a lookup table for every node
  - too much

Node A
Node B
Node E
12
Types of Routing

• Need to develop routing algorithms that will work
  for a WAN
• Similar to using Bridges at layer 2, there are
  several basic types of routing algorithms which
  we need to discuss
• Centralized Routing
• Distributed Routing
• Static Routing
• Adaptive Routing
• Plus combinations!

13
Centralized Routing

• One node obtains connection/cost info
• Decides on routes
• Broadcast routes to other nodes
• Use a Routing matrix

Destination Node
Source Node
14
Centralized Routing

• Advantages
• Simple routing method because one location
  assumes routing control
• Disadvantages
• Failure of the central location or any links
  connected to it has severe effect on providing
  routing information to network nodes

15
Distributed Routing

• Each node determines/maintains own table
• Starts with knowing only its nearest neighbors
• Obtain costs from neighbors to other nodes
• Share routing information with other nodes to
  gather knowledge of the wider network

D
A
2
7
2
1
E
2
4
B
C
16
Distributed Routing

• Advantages
• Failure of a node or link has little effect in
  providing accurate routing information
• Disadvantages
• Exchange of information is more complex
• May take longer for a node to learn of conditions
  in remote locations

17
Static Routing

• Tables fixed once established
• Assumes the cheapest path is not dependent on
  time
• Valid assumption if conditions don't change much
• Major upgrades to a node in the network
• Loss of a network node

18
Static Routing

• Advantages
• Simple method because nodes do not have to
  execute routing algorithms repeatedly
• Disadvantages
• Insensitive to changing conditions
• What was a good route may turn into a very bad one

19
Adaptive Routing

• Tables adjust to accommodate conditions
• Necessary if conditions change often
• Problems
• Packet may never get to destination (sent back
  and forth as conditions change)
• Difficult to implement efficiently takes time
  for info to propagate to all nodes (may be
  out-of-date).

20
Adaptive Routing

• Advantages
• Provides the most current information regarding
  link costs
• Disadvantages
• High overhead because nodes must maintain current
  information
• Transmitting information regarding changing
  conditions adds to network traffic

21
Dijkstra's Algorithm

• Also know as
• The shortest path algorithm
• Forward search algorithm
• Static "centralized"
• Can be made adaptive if executed often
• Must know cost on all links
• Iterative process which builds a set of nodes one
  at a time
• "Shortest path" known for each node in set.

22
Dijkstra's Algorithm - Process

• Initialize
• set S A (where A is the starting node)
• for all other nodes x
• PRIOR(x)A if directly connected to A, else
  "undefined".
• COST(x)cost of link AgtX (or infinity if none)

23
Dijkstra's Algorithm - Process

• Loop (until no nodes left)
• W x there's a link from x to a node in S,
  but x is not in S
• choose y from W with minimum COST
• S S union y
• for all x not in S, update COST
• COST(x) minimum of COST(x) and (COST(y)cost of
  link x to y)
• if cost changed, PRIOR(x) is y

24
Dijkstra's Algorithm - Process

• PRIOR function determines path back to start
• Each node can exec with itself as start node
• Or centralize and broadcast results.

25
Adding Nodes using Dijkstra's Algorithm

• Snodes to which cheapest route to A is known

• Wnodes connected to a node in S via a direct
  link

S
W
Direct Link
Put node X in S
X
Route
A
Is there a cheaper route to V through X?
V
Cost (V) represents cost of the current route
from A to V (if it exists)
26
Dijkstra's Algorithm - Example
27
Values Defined by Dijkstra
28
Cost Comparisons for Dijkstras Algorithm
Cost Comparisons After Adding C to S
Cost Comparisons After Adding B to S
29
Routing in Wide Area Networks

• For networks that are very large, either in terms
  of the number of hosts or the distances between
  hosts, the "shared medium" approaches used in
  LANs just don't scale.
• No matter how efficient the protocol for sharing
  the medium may be, having all message go to all
  stations involves too much unnecessary
  transmission of information when the number of
  stations is very large.
• We have seen that the efficiency of most medium
  sharing protocols decreases as the physical size
  of the network increases

30
Routing in Wide Area Networks

• In such networks, it is better to depend on
  intermediate nodes to forward packets selectively
  along paths leading to their destinations. This
  is similar to the work performed by bridges on a
  LAN except
• We are no longer concerned that the process be
  transparent to sending stations or to network
  administrators, so...
• We would rather use multiple (i.e. cyclic) paths
  in the network to maximize performance and
  improve reliability rather than depend on a
  spanning tree.

31
Routing in Wide Area Networks

• To make all this work, when machine A wants to
  talk to machine B, somehow the network has to
  find a path from A to B.
• There are several major characteristics of the
  process of finding such paths that vary in
  significant ways from network to network

32
Routing in Wide Area Networks

• Characteristics of routing to consider
• Static Routing originally used in small networks
• Dynamic Routing routing protocols that adjust to
  changes in the state of network links
• Distributed Routing distribute the task of
  computing routes among all the nodes of the
  network
• Datagrams vs. Virtual Circuits ie packet by
  packet vs message by message
• Hop-by-hop routing vs. source routing Source
  only knows initial node vs source figuring out
  the entire route

33
Routing in Wide Area Networks

• Alternatives of hop-by-hop routing and datagram
  based networks are more reasonable than they
  might seem
• With hop-by-hop routing, the work done for each
  packet is simply to look its destination up in a
  destination/next-step table
• Even if a connection had been established first,
  the switching node would have to lookup the
  packet's virtual circuit ID to figure out the
  next step
• When routing is done hop-by-hop, the switches are
  really depending on big tables to make routing
  decisions.
• The problem isn't what you do while forwarding a
  packet but rather how those tables get build

34
Routing in Wide Area Networks

• Weve already talked about Dijkstras Algorithm
  for determining a least cost path from A to B
• Now lets examine a few other algorithms

35
Bellman-Ford Algorithm

• Backward Search Algorithm
• Each node A knows its immediate neighbors and the
  cost to send to them.
• For any node Z in the network, X is the neighbor
  to send frames destined to Z if X minimizes the
  following
• COSTAgtX COSTXgtZ
• i.e. for each neighbor Y, calculate
• COSTAgtY COSTYgtZ
• If Y does not know how to get to Z,
• COSTYgtZ infinity (i.e. large)

36
Bellman-Ford Algorithm

• Let X be the Y which produces the lowest cost.
• Periodically each station sends its table of
  costs to its neighbors
• Neighbors use these to adjust their own routing
  tables
• Over time, the algorithm will converge to a
  shortest path solution
• Slow reaction to change (depends on topology)
• Rely on neighbors for shorter paths
• Invalidate routing table entries periodically.
  When?

37
Routing in Wide Area Networks

• Weve discussed Dijkstras and the Bellman-Ford
  Algorithms
• Although not obvious, Dijkstras algorithm will
  converge to a shortest path much more rapidly
  than Bellman-Ford
• Bellman-Ford is more adaptable to a distributed
  environment
• Dijkstras algorithm is superior if we can find
  some way to make Dijkstras algorithm work better
  in a distributed environment
• all we really need is to find a way to send each
  node's link state information to all other nodes

38
Routing in Wide Area Networks

• We cannot depend on routes based on shortest
  paths to deliver the link-state packets required
  by Dijkstra's algorithm to determine these
  shortest paths
• If nothing else, think of the initialization
  problem - How do you distribute the link state
  packets when the network is first started and
  nodes have no idea how to route anything?
• Similar to the start-up problem is the problem of
  how the network will recover if a damaged link
  whose failure had partitioned the network into
  disjoint components becomes operational again.
• If the load on the current shortest path rises
  suddenly, then to reroute packets to use less
  loaded paths a network that used "shortest paths"
  would have to depend on congested paths to
  deliver the information needed to route around
  the congestion.

39
Routing in Wide Area Networks

• Use Link State Routing to provide information
  about all nodes in the network to each node
• Basically use a form of the flooding algorithm we
  talked about in chapter 6 when we talked about
  routing with transparent bridges and the spanning
  tree algorithm
• Use counters to control the number of link state
  packets in the network
• Run Dijkstra's algorithm periodically to update
  routing tables based on recent link state packets

40
Routing in Wide Area Networks

• So far what we have discussed will work at least
  in theory for any network
• However, a major consideration is not the cost of
  the route but merely the ability to get the
  information there in a timely manner
• This is especially true for very large networks
  or systems of networks
• Need to take a different approach
• The Internet solution - Hierarchical Routing!

41
Hierarchical Routing

• Large WAN
• Impossible for each node to store entire routing
  table
• Too much info to send between nodes
• Nodes divided into domains
• Routing within domain accomplished by Interior
  Routing Protocols like those weve already
  discussed
• Routers connect domains using exterior routing
  protocols

42
Hierarchical Routing

• Destination address specifies node (host), domain
  and sub-domains.
• Router only forwards frames destined for another
  domain.
• e.g. postal addresses

43
Hierarchical Routing
Designated Nodes
A
B
X
Y
Domain 2
Domain 1
W
Subdomain 1
Subdomain 2
Domain 3
44
Hierarchical Arrangement of Domains
Root
Domain 1
Domain 2
Domain 3
A
X
Other nodes
Other nodes
B
Y
Subdomain 1
Subdomain 2
W
Z
C
K
Other nodes
Other nodes
45
Hierarchical Routing

• Internet uses hierarchical routing
• 32-bit address (4 8-bit numbers)
• 2 parts (2-level hierarchy)
• IP address of site (network)
• IP address of host (machine)
• Class AC addresses

46
Hierarchical Routing

• Names (user_at_host.dept. .domain)
• no geographical significance
• for users only, not for routing
• DNS (Domain Name Server) protocol translates
  between Name IP number
• Address depletion becoming a problem
• Too few addresses for number of machine on
  Internet
• Extending the standard
• Reusing addresses

47
Routing Problems

• Congestion
• Too much data at link cannot send them quickly
  enough.
• Frames delayed network unresponsive
• Solutions to problems with congestion
• Do nothing (wait for it to go away)
• Eliminate waiting frames
• Use flow control
• Allocate buffers for each virtual circuit
• Choke packets

48
Summary of Routing Strategies
49
Introduction

• Motivation for WANs
• Wide Area Network Routing
• Public Data Networks The X Series
• Internet Protocols
• Transport Protocols

50
X Series Protocols

• Four common protocols X.25, X.3, X.28, and X.29
• All deal with getting data through a public data
  network
• Deals with transmitting packets through a myriad
  of switches - ie packet switching networks

51
Packet-Switched Network
A
Two basic modes 1 - Virtual Circuits 2 -
Datagram Service
5
4
B
3
1
2
1
Packet Switched Network
4
5
C
3
2
D
Switching logic in the cloud makes routing
decisions similar to those we just discussed
52
Virtual Circuit Connections
A
C
Vc1
Vc3
Vc5
Y
X
Vc3
Vc1
Vc2
B
D
53
Routing Tables using Virtual Circuits
54
Datagram Service
4
5
A
Y
3
2
1
D
X
Packet Switched Network
55
Comparison of Virtual Circuits and Datagrams
Virtual Circuit
Datagram

• Helps prevent congestion. Deterministic use of
  resources.
• If circuit active too long, current path may not
  be the best one
• Routing decision is made just once
• Packets always arrive in order
• Node failure breaks the link - loss of data

• Unexpected packets make congestion control more
  difficult
• Nodes route each packet using the most current
  information
• Separate routing decisions for each packet
• Packets can arrive out of order
• A failed node can be routed around

56
X.25 Public Data Network Interface
DTE
Network defined protocol
X.25 Protocol
DCE
DCE
DTE
X.25 Protocol
Packet Data Network
57
X.25 Interface Standard

• Defines a synchronous transmission analogous to
  the lowest three layers of the OSI model
• Network layer receives the user data and puts it
  into an X.25 packet
• The X.25 packet is sent to the data link layer
  where it is embedded in a LAPB frame (link access
  protocol - balanced)
• The physical layer then transmits the frame using
  the X.21 protocol

58
X.25 Protocol Layers
Packet layer
Network
Network
LAPB
Data Link
Data Link
Physical
Physical
X.21 or X.21bis
59
X.25 Packet Formats
1
1 or 2
variable
1
Flags, logical group
logical channel
control
DATA
Data Packet
1
1 or 2
variable
1
Flags, logical group
logical channel
Packet Type
Other Information
Control Packet
60
X.25 Packet Packaging
X.25 Packet
LAPB Frame
61
Triple-X Standards

• Used to connect non-DTE devices to
  packet-switched networks
• X.3 - Packet Assembler and Disassembler
• accepts input from a dumb terminal and constructs
  X.25 packets
• X.28 - Provides the interface standard between
  the X.3 and the dumb terminal
• What commands/keystrokes mean
• X.29 - PAD-host interface
• Defines the communication protocol between the
  PAD and the remote host

62
Triple-X Protocols
X.29 Protocol
host
X.25 Protocol
DCE
Non-X.25 device
Network defined protocol
X.3 PAD
X.28 Protocol
Packet Data Network
63
Introduction

• Motivation for WANs
• Wide Area Network Routing
• Public Data Networks The X Series
• Internet Protocols
• Transport Protocols

64
Internet Protocols

• With all the different approaches to networking
  we have considered, it should be clear that one
  of the problems facing the effective use of
  networks is simply that there are too many kinds
  of networks.
• To provide the sort of global communication
  possible through the Internet, we need to provide
  mechanisms that enable machines connected to
  networks based on distinct technologies and
  protocols to communicate.

65
Internet Protocols

• The Internet is based on a different approach to
  interconnecting dissimilar networks
• Rather than simply translating frames of one
  existing protocol format to another, the Internet
  is based on defining a new "virtual network"
  protocol -- the iP (or internet protocol)
• Given the common language represented by the iP,
  one only needs 2N converters. N that convert from
  the existing protocols to iP and N that convert
  from iP to the existing protocols.
• To interconnect without overlaying a protocol
  such as IP we would need N(N-1) different
  protocol converters

66
Internet Protocols
DNS
SMTP
FTP
TELNET
TCP
UDP
IP and ICMP
Layer Three
Lower layer protocols
67
IP Transmitting Packets over Different Networks
B running TCP/IP
A running TCP/IP
User to User connection
TCP segment
TCP segment
Data
Data
IP packet
IP packet
Router running IP
Router running IP
TCP
TCP
Ethernet
TR
X.25
X.25
Ethernet frame
IP
IP
IP
IP
IP
IP
Token ring frame
Token Ring LAN
X.25 PDN
Ethernet LAN
68
Internet Protocol Packet
Packet Length
Type of service
Version
Header length
Identification
Flags
Fragment Offset
Time to Live
Protocol
Checksum
Header
Source IP address
Destination IP address
Options
Message Data
69
Internet Domains
Domain
Meaning

• Com
• edu
• int
• gov
• mil
• net
• org
• country code

• Commercial institution
• Educational institution
• International organizations
• Government agency
• Military
• Network service providers
• Nonprofit organizations
• jp for Japan, etc

70
Internet Address Classifications
71
Domain Name System

• A distributed database that contains copies of
  text addresses and their associated 32-bit
  address
• All information about addresses is spread across
  the internet among a collection of DNS servers
• When a host computer needs an address
  translation, it calls a DNS server to look up the
  internet address
• Each DNS server responsible for a ZONE in the DNS
  hierarchy
• How then does a host know how to find the IP
  address? And get the information physically there?

72
Finding IP addresses

• Remember that internet routing is hierarchical.
• Within a physical network we depend on the
  underlying networks delivery mechanism (broadcast
  or switched).
• We only count on IP routing and routers to get a
  packet to some machine (a router) actually
  attached to the same network as the destination
  machine.
• In order to perform hierarchical routing,
  machines in the network need a way to quickly
  determine whether a packet is destined for a
  machine in the current network or requires
  forwarding through IP routers
• IP solves this by using an address format that is
  tied to the underlying networks

73
Finding IP Addresses

• Each machine's IP address has two sub-parts.
• Network Part
• Host Part

74
Finding IP Addresses

• Network part
• The first portion of a host's address identifies
  the network to which the host is attached. All
  machines attached to a given physical network
  must share this component of their IP addresses
• The network part of an IP address is assigned by
  a "central authority". Before you can connect to
  the Internet, you must contact this authority to
  get a network number
• InterNIC!!

75
Finding IP Addresses

• Host part
• Machine's within a network are identified by the
  second portion of the IP address. No two machines
  within a given hardware network may have the same
  host portion of their IP addresses
• Once a network number is assigned by the central
  authority, those running a local network can
  independently assign host numbers to machines
  within their network

76
Finding an IP address
Root
edu
com
org
mit
uwgb
uwm
brooks
microsoft
acm
ieee
gbmsol
gbraxa
When a host needs an address translation it sends
a request to the local name server. If the local
server does not know it, the request is sent to
one of the DNS servers at the top level. The top
level DNS server then directs the address request
to the appropriate name server
77
Finding an IP address

• Now that we know how to find an IP address
• We can route a packet through the Internet - is
  that all?
• We still have to resolve that IP address into a
  physical address that contains the hardware
  protocol address - NOT the IP address
• This is normally done by the destination network

78
Finding the Hardware Address

• We still need to find the appropriate hardware
  protocol address for the destination host (or
  router)
• most real implementations of TCP/IP depend on a
  protocol named ARP (Address Resolution Protocol)
  to build a look-up table on demand.
• ARP (usually) depends on the broadcast capability
  of LANs

79
Finding the Hardware Address

• When a machine's IP software needs to find the
  hardware address corresponding to a local IP
  address, it first looks in its (initially empty)
  table/cache of IP-to-hardware address pairs
• If it can't find a match, it broadcasts a
  hardware protocol packet asking if anyone knows
  the hardware address for the given IP address
• If a machine sees such a packet containing a
  request for its own address it responds with a
  packet containing its hardware address
• This hardware address then gets added to the cache

80
Summary of a Routers Actions

• Receive an IP packet and extract the IP address
• If the packets source route option is marked,
  route the packet accordingly
• Else, determine the network number contained in
  the IP address
• Does it match a network to which the router is
  connected?

81
Summary of a Routers Actions

• If yes, determine the physical address of the
  destination (either by cache lookup or dynamic
  binding) and send a frame containing the IP
  packet to that destination
• If no, find the network number in the routing
  table and forward the packet to the specified
  router
• If for some reason the network is not in the
  routing table, forward the packet to a default
  router

82
IP Summary

• What IP provides is called "Best Effort deliver".
  
• IP doesn't deliberately try to avoid delivering
  packets
• It just doesn't do anything to ensure they get
  delivered
• While IP does not attempt to correct errors, it
  does report many of them

83
ICMP

• All implementations of IP are expected to
  implement ICMP (Internet Control Message
  Protocol) which is used to report unusual
  conditions
• ICMP is implemented by placing ICMP messages in
  IP packets and then asking IP to deliver them
• Obviously, if IP is really working badly, ICMP
  will be an ineffective way to report errors
• ICMP messages provide ways to
• Report when there is no known way to reach a host
• Report when a packet exceeds its time to live
  (probably because of a routing loop)
• Test to see if a host is responsive

84
Introduction

• Motivation for WANs
• Wide Area Network Routing
• Public Data Networks The X Series
• Internet Protocols
• Transport Protocols

85
Transport Protocols
DNS
SMTP
FTP
TELNET
TCP
UDP
Layer Four
IP and ICMP
Lower layer protocols
86
Transport Protocols

• The physical, data-link and network layers of the
  OSI protocol hierarchy provide the ability to
  deliver messages from one machine to another
  through complex networks of networks
• We really need more than that - the ability to
  deliver a message from one process to another
• Typical server machine is likely to be running
  several processes intended to provide distinct
  network services.
• When a packet arrives at such a machine, the
  network software needs some way to determine
  which server process the packet should ultimately
  be delivered to
• If it is possible to run two clients on a machine
  simultaneously (Netscape and NCSA Telnet), then
  the network software has to somehow figure out
  which client each incoming packet is destined
  for.

87
Transport Protocols

• Providing process-to-process communication
  through a network is the role associated with the
  "transport" level in the OSI model
• The most widely used protocol in the Internet
  Protocol suite, TCP (Transmission Control
  Protocol), is an example of a transport level
  protocol.
• Lets start our discussion of transport protocols
  with a simpler, less well known member of the
  Internet protocol suite, UDP (User Datagram
  Protocol).

88
UDP

• The beauty of UDP is that it does so little.
• Basically, almost the only difference between UDP
  and IP is the use of something called port
  numbers
• Remember, the big difference between the
  transport layer and the network layer is the
  desire to provide process-to-process
  communication
• To do this, we need some form of
  process-to-process addressing
• It would be dangerous to use packet addresses
  formed by joining an IP address to a process
  number because process numbering schemes vary
  from operating system to operating system

89
UDP

• UDP introduces the notion of an addressable set
  of "ports" associated with each machine and
  allows messages to be sent and received only
  through these ports
• Like using extensions in a office telephone
  system
• IP address is the basic phone number
• Port address is the extension to the person you
  want to reach

90
UDP

• Before a process running on a machine can send or
  receive UDP messages, it must ask the TCP/IP
  software on its machine to allocate a UDP port
  for its use
• A process can either ask for a port with a
  particular number or any free port
• On systems with security mechanisms, certain port
  numbers (the low ones) are reserved for
  privileged processes
• When a packet is sent, the destination address
  must include both the IP address for the
  destination machine and a destination port
  number.
• When a packet is delivered, the receiving process
  is given the sender's IP address and the source
  port number

91
UDP

• To deliver UDP packets, the network again depends
  on the magic of encapsulation
• The format of a UDP packet is

Source Port
Destination Port
Check sum
Length
Packet Data

• The packet only contains port numbers (no IP
  addresses)
• To deliver a UDP packet, the UPD software places
  the entire UDP packet in the data field of an IP
  packet
• IP packet will have the source and destination IP
  addresses in its header (and which will of course
  end up as the data field of some hardware
  protocol packet)

92
UDP Summary

• UDP provides layer 4 services for DNS
• UDP has little overhead
• UDP provides no acknowledgement of errors
• UDP does perform error detection and if an error
  occurs, discards the data

93
TCP

• A major function of TCP is to ensure reliable
  delivery
• Why do we have to deal with reliability again
  when we dealt with it at the data link layer?
• If you accept that we must deal with it again,
  why not take care of it at the network layer - in
  IP?
• Good questions!!

94
TCP

• First, realize that even if low level protocols
  are designed to be reliable, network layer
  protocols may need to do additional work to
  preserve this reliability
• If an intermediate router crashed after receiving
  and acknowledging a message but before forwarding
  it, the message might still be lost
• IP, of course, is designed to function on a wide
  range of underlying networks. So, it is best to
  assume some of them might not provide reliability
  guarantees
• IP does not provide any guarantees of reliable
  delivery
• IP is still one step away from an end-to-end
  connection

95
TCP

• If we really want confirmation of a message's
  delivery, we will want to know that it got to the
  target process not just the target machine
• UDP, like IP, does not guarantee delivery
• The UDP checksum is optional
• UDP is intended for end-to-end applications where
  reliability is not a primary concern
• Within the Internet Protocol Suite, however,
  there is a reliable transport layer named TCP

96
TCP

• TCP shares some of UDP's features
• It is implemented by encapsulating TCP packets
  (actually called TCP segments) in IP packets
• It depends on port numbers (that are interpreted
  independently from UDP port numbers).

97
TCP

• TCP differs from UDP in several ways
• It guarantees reliable, ordered delivery of data.
  
• It is byte-stream oriented.
• That is, it does not offer to deliver packets but
  bytes.
• For example, even though a sending process
  invokes TCP 5 times sending 120 bytes of data
  each time, the receiving TCP may deliver 3 units
  of 200 bytes each to the process at the other
  end.
• Basically, the bytes sent will be delivered in
  the order sent but not necessarily in the same
  units.

98
TCP Segment
Destination Port
Source Port
Sequence Number
Acknowledgement Number
Header
Flags
Header length
window
Urgent Pointer
checksum
Options
Message Data
99
TCP

• Destination Port identifies the application to
  which the segment is sent
• Source Port specifies the application sending
  the segment
• Sequence Number The sequence number of the
  starting byte in data
• Acknowledgement Number Contains the byte
  sequence number the receiving TCP entity expects
  to see

100
TCP

• Header Length Specifies the size of the TCP
  header as a multiple of 4 bytes
• Flags Specifies when other fields contain
  meaningful data
• Window Tells receiving TCP entity how many more
  data bytes it can send beyond those already
  acknowledged
• Checksum 16-bit transport layer error detection
• Urgent Pointer Points to the first byte
  following the urgent data

101
TCP Connection Management
A
B
t1
Send TCP Segment with SYN 1 and sequence field
x
Send TCP Segment with SYN 1, acknowledgement
field x1, and sequence field y
t2
Send TCP Segment with acknowledgement field y1
t3
Send Data TCP Segments. Include sequence numbers
beginning with y1 and acknowledgement numbers
beginning with x1.
Send Data TCP Segments. Include sequence numbers
beginning with x1 and acknowledgement numbers
beginning with y1.
time
time
102
TCP Connection Management
A
B
Initial sequence number 100
Initial sequence number 700
Data, s101, a701,...
Data, s201, a701,...
Wait for more credit
t1
First segment removed from buffer also finds 100
more bytes of buffer space
t2
Data, s701, a301, c0
Data, s801, a301, c200
Data, s301, a901,...
Data, s401, a901,...
Wait for more credit
t3
time
time
103
TCP Congestion Control

• Also need to provide some way of controlling
  congestion
• Use a congestion window to limit the data flow to
  less than the ideal credit limit normally allowed
• Congestion window is based on a timeout
• Reduce the congestion window by half on each
  successive timeout

104
End Chapter 7
105
(No Transcript)
106
(No Transcript)