Module A Updated January 2009 More on TCP/IP Module A Updated January 2009 Raymond Panko’s Business Data Networks and Telecommunications, 7th edition May only be used by adopters of the book

Multiplexing

Multiplexing IP packets can carry different things in their data fields TCP segments UDP datagrams ICMP supervisory messages (later) RIP messages (later) IP Data Field IP Header

Stream of Arriving or Outgoing IP Packets Multiplexing We say that IP can multiplex (mix) different types of traffic in a stream of IP packets Single IP Packet Carrying UDP Datagram UDP IP-H TCP IP-H UDP IP-H ICMP IP-H Stream of Arriving or Outgoing IP Packets

Multiplexing IP process must pass contents of arriving IP packets to the correct process for subsequent handling TCP UDP UDP IP-H IP ICMP Arriving Packets IP Process

Multiplexing IP process must also accept messages from multiple processes and multiplex them on an outgoing stream TCP UDP IP-H UDP IP ICMP Outgoing Packets IP Process

Multiplexing Need a Way for Receiving IP Process to Know What Is in the Data Field So it can pass the contents to the appropriate process IP Data Field IP Header

Multiplexing IP Header has an 8-bit Protocol field Identifies the contents of the data field 1=ICMP, 8=TCP, 17=UDP, etc. Version (4) Hdr Len (4) TOS (8) Total Length in bytes (16) Indication (16 bits) Flags (3) Fragment Offset (13) Time to Live (8) Protocol (8) Header Checksum (16) Source IP Address Destination IP Address

Multiplexing Other Messages Have Analogous Fields TCP and UDP Identify contents of data field TCP and UDP Have Port number fields Identify the application process (80=HTTP) Source Port # (16) Destination Port # (16) Sequence Number (32 bits) Acknowledgement Number (32 bits) Hdr Len (4) Reserved (6) Flags (6) Window Size (16)

Multiplexing Other Messages Have Analogous Fields PPP Identify contents of data field PPP Protocol field identifies contents of information field as IP, IPX, a supervisory message, etc. Flag Addr Ctrl Prot Info CRC Flag

More on TCP Acknowledgements Sequence Number Field Initial Sequence Number Acknowledgement Number Field

TCP TCP is Reliable IP packets carrying TCP segments may arrive out of order TCP must put the TCP segments in order 5 3 4 2 1

TCP TCP is Reliable Each correct TCP segment is acknowledged by the receiver TCP Segment Source Transport Process Destination Transport Process ACK

TCP Segment Each TCP segment sent by a side must have a sequence number Simplest: 1,2,3,4,5,6,7 To detect lost or out-of-sequence messages TCP uses a more complex approach 3? 1 4 2 5

TCP Sequence Numbers TCP header has a 32-bit sequence number field Source Port # (16) Destination Port # (16) Sequence Number (32 bits) Acknowledgement Number (32 bits) Hdr Len (4) Reserved (6) Flags (6) Window Size (16) TCP Checksum (16) Urgent Pointer (16) Options (if any) PAD Data Field

with Initial Sequence Number (79) TCP Sequence Numbers Initial Sequence Number is randomly selected by the sender; Say, 79 Sent in the sequence number field of the first TCP segment TCP Header 79 TCP Data Field Sequence Number Field with Initial Sequence Number (79)

TCP Sequence Numbers Data octets in data fields of all segments in a connection are viewed as a long string TCP Segment 1 79 TCP Segment 2 80 81 82 TCP Segment 3 83 84 ISN 3 Octets in Data Field 2 Octets in Data Field

TCP Sequence Numbers Supervisory segments, which contain a header but no data, are treated as carrying a single octet of data TCP seg 1 898 899 TCP seg 2 900 TCP seg 3 901 902 Carries data Supervisory segment Carries data

TCP Sequence Numbers Sequence number field gets the value of the first octet in the data field TCP 1 79 TCP 2 80 81 82 TCP 3 83 84 79 is SeqNum Field Value 80 is SeqNum Field Value 83 is SeqNum Field Value

TCP Acknowledgements Acknowledgement must indicate which TCP segment is being acknowledged TCP Segment Source TCP Process Destination TCP Process ACK

TCP Acknowledgements TCP header contains a 32-bit Acknowledgement Number field to designate the TCP segment being acknowledged Source Port # (16) Destination Port # (16) Sequence Number (32 bits) Acknowledgement Number (32 bits) Hdr Len (4) Reserved (6) Flags (6) Window Size (16) TCP Checksum (16) Urgent Pointer (16) Options (if any) PAD Data Field

TCP Acknowledgment Numbers Acknowledgement Number field contains the next byte expected—the last byte of the segment being acknowledged, plus one TCP 1 79 TCP 2 80 81 82 TCP 3 83 84 79+1=80 is AckNum Field Value 82+1=83 is AckNum Value 84+1=85 is AckNum Value

TCP Acknowledgement Number Quiz: A TCP segment contains the following data octets 567, 568, 569, 570, 571, 572, 573, 574 What will be in the sequence number field of the TCP segment delivering the data? What will be in the acknowledgement number field of the TCP segment acknowledging the TCP segment that delivers these octets?

TCP Flow Control Flow Control One TCP process transmits too fast Other TCP process is overwhelmed Receiver must control transmission rate This is flow control Too Much Data TCP Process TCP Process Flow Control Message

TCP Flow Control A TCP segment has a Window Size field Used in acknowledgements Source Port # (16) Destination Port # (16) Sequence Number (32 bits) Acknowledgement Number (32 bits) Hdr Len (4) Reserved (6) Flags (6) Window Size (16) TCP Checksum (16) Urgent Pointer (16) Options (if any) PAD Data Field

Acknowledgement with Window Size Field TCP Flow Control A TCP segment has a Window Size field Tell how many more octets the sender can send beyond the segment being acknowledged Data TCP Process TCP Process Acknowledgement with Window Size Field

TCP Flow Control Example TCP segment contained octets 45-89 Acknowledgement number for TCP segment acknowledging the segment is 90 If Window Size field value is 50, then Sender may send through octet 140 Must then stop unless the window has been extended in another acknowledgement

TCP Flow Control Each Acknowledgement extends the window of octets that may be sent Called a sliding window protocol 1-44 45-79 80-419 420-630 May send through 480 400 1-44 45-79 80-419 420-630 May send through 920 500

Application Layer Message TCP Fragmentation TCP Segments have maximum data field sizes (Size limit details are discussed later) What if an application layer message is too large? Application Layer Message TCP Data Field Max TCP Header

TCP Fragmentation Application layer message must be fragmented Broken into several pieces Delivered in separate TCP segments App Frag 1 App Frag 2 App Frag 3 TCP Data Field Max TCP Header

TCP Fragmentation Note that, in TCP fragmentation, the TCP segment is NOT fragmented The application layer message is fragmented App Frag 1 App Frag 2 App Frag 3 TCP Data Field Max TCP Header

TCP Fragmentation Transport layer process on the source host does the fragmentation Application layer on the source host is not involved Transparent to the application layer Application Application Message Transport TCP Segment TCP Segment Internet

TCP Fragmentation Transport layer process on the destination host does the reassembly Application layer on the destination host is not involved; Gets original application layer message Application Application Message TCP Segment TCP Segment Transport Internet

TCP Fragmentation What is the maximum TCP data field size? Complex Maximum Segment Size (MSS) Maximum size of a TCP segment’s data field NOT maximum size of the segment as its name would suggest!!!

TCP Fragmentation MSS Default is 536 octets Maximum IP packet size any network must support is 576 octets Larger IP packets MAY be fragmented IP and TCP headers are 20 octets each if there are no options This gives the default MSS of 536 Smaller if there are options in the IP or TCP header

TCP Fragmentation MSS Default is 536 octets Suppose the application layer process is 1,000 octets long Two TCP segments will be needed to send the data The first can send the first 536 octets The second can carry the remaining 464 octets of the application layer message

TCP Fragmentation Each side MAY announce a larger MSS An option usually used in the initial SYN message it sends to the other If announces MSS of 2,048, this many octets of data may be sent in each TCP segments 536 is only the default—the value to use if no other value is specified by the other side

More on Internet Layer Processes

Mask Operations Masks were introduced in Chapter 3 IP addresses alone do not tell you the size of their network or subnet parts Network Mask Has 1s in the network part Has 0s in the remaining bits Subnet Mask Has 1s in the network plus subnet parts

Mask Operations Based on Logical AND Example Both must be true for the result to be true Example 1010101010 Data 1111100000 Mask 1010100000 Result

Mask Operations Based on Logical AND Example If mask bit is 1, get back original data If mask bit is 0, get back zero Example 1010101010 Data 1111100000 Mask 1010100000 Result

Mask Operations IP packet arrives at a router Router sees destination IP address 11111111 01000000 10101010 00000000 Compares to each router forwarding table row Address Part in First Entry 11111111 01000000 00000000 00000000 Mask in First Entry 11111111 11100000 00000000 00000000

Mask the IP destination Address Mask Operations Mask the IP destination Address 11111111 01000000 10101010 00000000 (IP address) 11111111 11100000 00000000 00000000 (mask) 11111111 01000000 00000000 00000000 (result) Compare Result with First Entry Address part 11111111 01000000 00000000 00000000 (address part) The Entry is a Match!

Mask Operations Recap Read destination IP address of incoming IP packet For each entry in the router forwarding table Read the mask (prefix) Mask the incoming IP address Compare the result with the entry’s IP address part Do they match or not?

Mask Operations Simple for Computers Computers have circuitry to AND to numbers Computers have circuitry to COMPARE two numbers to see if they are equal or not Very computer-friendly, so used on routers Difficult for people, unfortunately

IPv6 Current version of the Internet Protocol is Version 4 (v4) Earlier versions were not implemented The next version will be Version 6 (v6) No v5 was implemented Informally called IPng (Next Generation) IPv6 is Already Defined Continuing improvements in v4 may delay its adoption

IPv6 IPv6 will raise the size of the internet address from 32 bits to 128 bits Now running out of IP addresses Will solve the problem But current work-arounds are delaying the need for IPv6 addresses

IPv6 Improved Security Improved Quality of Service (QoS) But, through IPsec, v4 is being upgraded in security as well Improved Quality of Service (QoS) But under IETF Differentiated Services (diffserv) initiative, IPv4 is being upgraded in this area as well

IPv6 Extension Headers IPv4 Headers are complex IPv6 basic header is simple Extension headers for options Basic Header Extension Header 1 Extension Header 2

IPv6 Extension Headers Basic header has 8-bit Next Header field Identifies first extension header or says that payload follows Basic Header NH Extension Header 1 Extension Header 2

IPv6 Extension Headers Each extension header also has 8-bit Next Header field Identifies next extension header or says that payload follows Basic Header Extension Header 1 NH Extension Header 2

IPv6 Extension Headers Next header field is an elegant way to allow options Easy to add new extension headers for new needs Basic Header Extension Header 1 NH Extension Header 2

IP Fragmentation

MTU Maximum Transmission Unit (MTU) Largest IP packet a network will accept Arriving IP packet may be larger MTU IP Packet

IP Fragmentation If IP packet is longer than the MTU, the router breaks packet into smaller packets Called IP fragments Fragments are still IP packets Earlier in Mod A, fragmentation in TCP MTU IP Packet 3 2 1 IP Packets Fragmentation

IP Fragmentation What Is Fragmented? Only the original data field New headers are created MTU IP Packet 3 2 1 IP Packets Fragmentation

IP Fragmentation What Does the Fragmentation? The router Not the subnet MTU IP Packet 3 2 1 IP Packets Fragmentation

Multiple Fragmentations Original packet may be fragmented multiple times along its route Source Host Internet Process Destination Host Internet Process Fragmentation

Defragmentation Internet layer process on destination host defragments, restoring the original packet IP Defragmentation only occurs once Source Host Internet Process Destination Host Internet Process Defragmentation

Fragmentation and IP Fields More Fragments field (1 bit) 1 if more fragments 0 if not Source host internet process sets to 0 If router fragments, sets More Fragments field in last fragment to 0 In all other fragments, sets to 1 1 1 Original IP Packet Fragments

Total Length in bytes (16) Identification Field IP packet has a 16-bit Identification field Source host internet process places a number in the Identification field Different for each original IP packet Version (4) Hdr Len (4) TOS (8) Total Length in bytes (16) Indication (16 bits) Flags (3) Fragment Offset (13) Time to Live (8) Protocol (8) Header Checksum (16)

Identification Field IP packet has a 16-bit Identification field If router fragments, places the original Identification field value in the Identification field of each fragment 47 47 47 47 Original IP Packet Fragments

Identification Field Purpose Allows receiving host’s internet layer process know what fragments belong to each original packet Works even if an IP packet is fragmented several times 47 47 47 47 Original IP Packet Fragments

Total Length in bytes (16) Fragment Offset Field Fragment offset field (13 bits) is used to reorder fragments with the same Identification field Contains the data field’s starting point (in octets) from the start of the data field in the original IP packet Version (4) Hdr Len (4) TOS (8) Total Length in bytes (16) Indication (16 bits) Flags (3) Fragment Offset (13)

Fragment Offset Field Receiving host’s internet layer process assembles fragments in order of increasing fragment offset field value This works even if fragments arrive out of order! Works even if fragmentation occurs multiple times Fragment Offset Field 730 212

Fragmentation: Recap IP Fragmentation Data field of a large IP packet is fragmented The fragments are sent into a series of smaller IP packets fitting a network’s MTU Fragmentation is done by routers Fragmentation may be done multiple times along the route

Defragmentation: Recap IP Defragmentation Defragmentation (reassembly) is done once, by the destination host’s internet layer process

Defragmentation: Recap All IP packets resulting from the fragmentation of the same original IP packet have the same Identification field value Destination host internet process orders all IP packets from the same original on the basis of their Fragment Offset field values More Fragments field tells whether there are no more fragments coming

Dynamic Routing Protocols Why Dynamic Routing Protocols? Each router acts independently, based on information in its router forwarding table Dynamic routing protocols allow routers to share information in their router forwarding tables Router Forwarding Table Data

Routing Information Protocol (RIP) Routing Information protocol (RIP) is the simplest dynamic routing protocol Each router broadcasts its entire routing table frequently Broadcasting makes RIP unsuitable for large networks Routing Table

Routing Information Protocol (RIP) RIP is the simplest dynamic routing protocol Broadcasts go to hosts as well as to routers RIP interrupts hosts frequently, slowing them down; Unsuitable for large networks Routing Table

Routing Information Protocol (RIP) RIP Is Limited RIP routing table has a field to indicate the number of router hops to a distant host The RIP maximum is 15 hops Farther networks are ignored Unsuitable for very large networks Hop Hop

Routing Information Protocol Is a Distance Vector Protocol “New York” starts, announces itself with an RIP broadcast “Chicago” learns that New York is one hop away Passes this on in its broadcasts New York Chicago NY is 1 Dallas 1 hop

Routing Information Protocol Learning Routing Information “Dallas” receives broadcast from Chicago Already knows “Chicago” is one hop from Dallas So New York must be two hops from Dallas Places this information in its routing table New York Chicago NY is 1 Dallas 1 hop 1 hop NY is 2

Routing Information Protocol Slow Convergence Convergence is getting correct routing tables after a failure in a router or link RIP converges very slowly May take minutes During that time, many packets may be lost

Routing Information Protocol Encapsulation Carried in data field of UDP datagram Port number is 520 UDP is unreliable, so RIP messages do not always get through A single lost RIP message does little or no harm UDP Data Field RIP Message UDP Header

OSPF Routing Protocol Link State Protocol Link is connection between two routers OSPF routing table stores more information about each link than just its hop count: cost, reliability, etc. Each router knows the state of every link Allows OSPF routers to optimize routing based on these variables Link

OSPF Routers Network Is Divided into Areas Each area has a designated router Area Designated Router

OSPF Routers When a router senses a link state change Sends this information to the designated router Area Designated Router Notice of Link State Change

OSPF Routers Designed Router Notifies all Routers Within its area Area Designated Router Notice of Link State Change

OSPF Routers Efficient Only routers are informed (not hosts) Usually only updates are transmitted, not whole tables Designated Router Area Notice of Link State Change

OSPF Fast Convergence When a failure occurs, a router transmits the notice to the designated router Designated router sends the information back out to other routers immediately

OSPF Encapsulation Carried in data field of IP packet Protocol value is 89 IP is unreliable, so OSPF messages do not always get through A single lost OSPF message does little or no harm IP Data Field OSPF Message IP Header

Selecting RIP or OSPF Within a network you control, it is your choice Your network is an autonomous system Select RIP or OSPF based on your needs Interior routing protocol

Selecting RIP or OSPF RIP is fine for small networks Easy to implement 15 hops is not a problem Broadcasting, interrupting hosts are not too important

Selecting RIP or OSPF OSPF is Scalable Works with networks of any size Management complexities are worth the cost in large networks

Border Gateway Protocol (BGP) To connect different autonomous systems Must standardize cross-system routing information exchanges BGP is most popular today Gateway is the old name for router Exterior routing protocol Autonomous System Autonomous System BGP

Border Gateway Protocol (BGP) Distance vector approach Number of hops to a distant system is stored in the router forwarding table Normally only sends updates Autonomous System Autonomous System BGP

Border Gateway Protocol (BGP) Encapsulation BGP uses TCP for delivery Reliable TCP is only for one-to-one connections If have several external routers, must establish a TCP and BGP connection to each BGP Message TCP Header IP Header

Address Resolution Protocol (ARP)

Internet and Data Link Layer Addresses Each host and router on a subnet needs a data link layer address to specify its address on its switched network This address appears in the data link layer frame sent on a subnet For instance, 48-bit 802.3 MAC layer frame addresses for LANs Subnet DA DL Frame for Subnet

Addresses Each host and router also needs an IP address at the internet layer to designate its position in the overall Internet Switched networks in an internet are called subnets 128.171.17.13 Subnet Subnet Subnet

Internet and Data Link Addresses Serve Different Purposes IP address To guide delivery to destination host across the Internet (across multiple networks) Switched Network Address To guide delivery between two hosts, two routers, and a host and router within a single switched network Same LAN, Frame Relay network, etc.

Analogy In company, each person has a company-wide ID number (like IP address) In company, person also has a local office number in a building Paychecks are made out to ID numbers For delivery, also need to know office number

Address Resolution Problem Router knows that destination host is on its subnet based on the IP address of an arriving packet Does not know the destination host’s subnet address, so cannot deliver the packet across the subnet Destination Host 128.171.17.13 Subnet subnet address?

Address Resolution Protocol (ARP) Router creates an ARP Request message to be sent to all hosts on the subnet Address resolution protocol message asks “Who has IP address 128.171.17.13?” Passes ARP request to data link layer process for delivery Subnet ARP Request

Address Resolution Protocol (ARP) Data link process of router broadcasts the ARP request message to all hosts on the subnet On a LAN, MAC address of 48 ones tells all stations to pay attention to the frame Subnet ARP Request

Address Resolution Protocol (ARP) Host with IP address 128.171.17.13 responds Internet process creates an ARP response message Contains the destination host’s subnet address (48-bit MAC address on a LAN) ARP Response Subnet

Address Resolution Protocol (ARP) Router delivers the IP packet to the destination host Places the IP packet in the subnet frame Puts the destination host’s subnet address in the destination address field of the frame Deliver IP Packet within a subnet frame Subnet

Address Resolution Protocol ARP Requests and Responses are sent between the internet layer processes on the router and the destination host ARP Request Router Destination Host Internet Process Internet Process ARP Response

Address Resolution Protocol However, the data link processes deliver these ARP packets Router broadcasts the ARP Request Destination host sends ARP response to the subnet source address found in the broadcast frame Router Destination Host Internet Process Internet Process Broadcast ARP Request Data Link Process Data Link Process Direct ARP Response

IP Address Classes How large is the network part in an IP address? Today we use network masks to tell Originally, IP had address classes with fixed numbers of bits in the network part Class A: 8 bits (24 bits in local part) Class B: 16 bits (16 bits in local part) Class C: 24 bits (8 bits in local part)

Class A IP Address IP address begins with 0 7 remaining bits in network part Only 128 possible Class A networks 24 bits in local part Over 16 million hosts per Class A network! All Class A network parts are assigned or reserved

Class B IP Address IP address begins with 10 (1st zero in 2nd position) 14 remaining bits in network part Over 16,000 possible Class B networks 16 bits in local part Over 65,000 possible hosts A good trade-off between number of networks and hosts per network Most have been assigned

Class C IP Address IP address begins with 110 (1st zero in 3d position) 21 more bits in network part Over 2 million possible Class C networks! 8 bits in local part Only 256 possible hosts per Class C network! Unpopular, because large firms must have several

Class D IP Address IP address begins with 1110 Used for multicasting, not defining networks Sending message to group of hosts Not just to one (unicasting) Not ALL hosts (broadcasting) Say to send a videoconference stream to a group of receivers

Class D IP Address All hosts in a multicast group listen for this multicast address as well as for their specific own host IP address In Group Accept Packets to Multicast Address Not in Group Reject In Group Accept

Multicasting Traditionally, unicasting and broadcasting Multicasting Unicasting: send to one host Broadcasting: send to ALL hosts Multicasting Send to SOME hosts Hosts on a list 500 stations viewing a video course 50 computers getting software upgrades Standards exist and are improving Not widely implemented yet

Why Multicasting Do not need to send an IP packet to each host Routers split when needed Reduces traffic Multiple Packets Single Packet

Mobile IP IP addresses are associated with fixed physical locations Mobile IP is needed for notebooks, other portable equipment Computer still gets a permanent IP address When travels, also gets a temporary IP address at its location This is linked dynamically to its permanent IP address

MultiProtocol Label Switching (MPLS) When a packet arrives, routers must consider all possible routes, then select the best one This is extremely expensive MPLS adds a tag to each packet An MPLS router examines the tag and passes the packet back out This is fast and inexpensive

MultiProtocol Label Switching (MPLS) This is similar to virtual circuits in ATM In addition, packets with similar priority or other characteristics can be given the same label and so be handled in the same way

MultiProtocol Label Switching (MPLS) MPLS Table for Q Tagged Packet Label 110 … Port 1 … NHR R … 110 Q Routing Decision R S

Copyright © 2009 Pearson Education, Inc. Publishing as Prentice Hall All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher. Printed in the United States of America. Copyright © 2009 Pearson Education, Inc.   Publishing as Prentice Hall