Spring 2010 CS 332 1

Chapter 4: Internetworking

Spring 2010 CS 332 2

Assumptions

• Data pipe from every machine to every other machine.– Need not be single link.

– Pipe can lose or corrupt messages.

• Sender/receiver may be on different physical networks, using different technology

• So what info do we need to build a single “logical” network (either reliable or unreliable)?

Spring 2010 CS 332 3

Issues• Getting various technologies to work with one

another (i.e. creating a single “network” from many heterogeneous systems).– Problem magnified since packet may need to traverse

several different networks (and network technologies), each with their own addressing schemes, service models, media access protocols, etc.

• Scale: It’s the big issue– How can you find an efficient path through a network

with millions (and perhaps billions eventually) of nodes?

– How do you provide addressing for a network with this many nodes?

Spring 2010 CS 332 4

Internetwork

• Arbitrary collection of possibly heterogeneous networks interconnected to provide host-to-host packet delivery service.

• Network: Directly connected or switched network that uses a single technology (i.e. ATM, 802.5, Ethernet).– Could be many physical networks creating a single

logical network.

Spring 2010 CS 332 5

Internetwork

• Internet—THE internetwork.– Runs the Internet Protocol (IP or Kahn-Cerf)

– Interesting because it has faced the problems of scale

– Experimental versions 1977 – 1981

– IPv4 first deployed in 1981

• internet—abstract internetwork

Spring 2010 CS 332 6

IP is a big deal

• Vint Cerf and Bob Kahn with Pres. Bush at 2006 ceremony where they received the Presidential Medal of Freedom for their work on IP.

White House News & Policies photo

Spring 2010 CS 332 7

IP Internet

• Concatenation of NetworksNote Hn denotes host,Rn denotes router.

Spring 2010 CS 332 8

IP Internet

• Protocol Stack

R1

ETH FDDI

IPIP

ETH

TCP R2

FDDI PPP

IP

R3

PPP ETH

IP

H1

IP

ETH

TCP

H8

Spring 2010 CS 332 9

The Internet

Outline Best Effort Service ModelGlobal Addressing Scheme

Spring 2010 CS 332 10

Service Model

• Connectionless (datagram-based)– So each packet must be “self-contained”

• Best-effort delivery (unreliable service)– packets are lost– packets are delivered out of order– duplicate copies of a packet are delivered (?!)– packets can be delayed for a long time

Spring 2010 CS 332 11

Why?!

• Best Effort service model is simple as it gets – intentionally!– If you provide best effort service over a network

technology that provides reliable delivery, you’re fine

– Providing reliable delivery over an unreliable network means extra functionality in the routers

– Keeping the routers as simple as possible was an IP design goal. (Why?)

• Note: IP today runs over many technologies that were not in existence when IP was invented!

Spring 2010 CS 332 12

IP Datagram Format

V ersion HLen TOS Length

Ident Flags Offset

TTL Protocol Checksum

SourceAddr

DestinationAddr

Options (variable)Pad

(variable)

0 4 8 16 19 31

Data

In 32 bit wordsIn bytes

Note: fieldsaligned on32 bit boundaries

Spring 2010 CS 332 13

Fields

• Version: note placement at front of packet (why?)• Header Length: in 32 bit words (20 bytes when no

options)• Type of service: later• Length: of entire packet in bytes (note max of

65,535 bytes because of 16 bit length field)• Ident, flags, offset all deal with fragmentation• Time to live: first seconds, but evolved to be hop

count

Spring 2010 CS 332 14

Fields• Protocol: demux key specifying higher level protocol that

gets datagram

• Checksum: take IP header as sequence of 16 bit words, add them using ones complement, take ones complement of result. – Relatively easy to calculate in software

– Not as strong error detection as CRC

– Bad packets discarded by router (potential bad dest. addr.)

• Src, dest address: pretty clear (and these are unique!)

• Options: rare, but complete IP implementation must handle them all! Presence determined by header length field

Spring 2010 CS 332 15

Fragmentation and Reassembly

• Each network has some MTU (why not uniform?)– Why not some uniform standard?– What is a reasonable choice for a given host?

• Strategy– fragment when necessary (MTU < Datagram length)– try to avoid fragmentation at source host– re-fragmentation is possible – fragments are self-contained datagrams– delay reassembly until destination host– do not recover from lost fragments

Spring 2010 CS 332 16

Fragmentation and Reassembly Header fields used in F &R (bits in parens)• Ident field (16): chosen by sending host, intended

to be unique among all datagrams that might be received at this dest from this source over reasonable time period.– All fragments keep this same ident value

• Offset (13): specifies 8 byte chunks of data (Why? And why not fragment #?)

• Flags: M is “more” flag

Spring 2010 CS 332 17

Example

H1 R1 R2 R3 H8

ETH IP (1400) FDDI IP (1400) PPP IP (512)

PPP IP (376)

PPP IP (512)

ETH IP (512)

ETH IP (376)

ETH IP (512)

Ident = x Offset = 0

Start of header

0

Rest of header

1400 data bytes

Ident = x Offset = 0

Start of header

1

Rest of header

512 data bytes

Ident = x Offset = 512

Start of header

1

Rest of header

512 data bytes

Ident = x Offset = 1024

Start of header

0

Rest of header

376 data bytes

MTU 532 bytes

Note: fragmentation can occurat multiple hops!

Spring 2010 CS 332 18

Global Addresses

• Properties– globally unique (don’t want anyone with my phone #)

• Why not just use Ethernet address?!

– hierarchical: network + host (really interface)

• Dot Notation– 10.3.2.4– 128.96.33.81– 192.12.69.77

Network Host

7 24

0A:

Network Host

14 16

1 0B:

Network Host

21 8

1 1 0C:

Spring 2010 CS 332 19

IP Internet Note Hn denotes host,Rn denotes router.

Routers need twoIP addresses.All hosts on same

network have samenetwork part ofIP address

Spring 2010 CS 332 20

Terminology

• Routing Mechanism: How a router selects the link over which to forward a packet

• Routing Protocol: Policies that determine what is placed in the routing tables.

These are not the same thing!

Spring 2010 CS 332 21

Datagram Forwarding • Strategy

– every datagram contains destination’s address– if directly connected to destination network, then forward to host– if not directly connected to destination network, then forward to

some router– forwarding table maps network number into next hop– each host has a default router– each router maintains a forwarding table

• Example (R2) Network Number Next Hop 1 R3 2 R1 3 interface 1 4 interface 0

Spring 2010 CS 332 22

Recall:

R2

R1

H4

H5

H3H2H1

Network 2 (Ethernet)

Network 1 (Ethernet)

H6

Network 3 (FDDI)

Network 4(point-to-point)

H7 R3 H8

Spring 2010 CS 332 23

Pseudocode

if (networknum dest = networknum my interface)

deliver packet over that interface

else

if (networknum in my routing table)

deliver packet to next hop router

else

deliver packet to default router

Spring 2010 CS 332 24

Address Translation • Map IP addresses into physical addresses

– destination host– next hop router– Why not just broadcast it? (E.g. if physical network is

Ethernet).

• Techniques– encode physical address in host part of IP address– table-based

• ARP– table of IP to physical address bindings– broadcast request if IP address not in table– target machine responds with its physical address– table entries are discarded if not refreshed

Spring 2010 CS 332 25

ARP Details • Request Format

– HardwareType: type of physical network (e.g., Ethernet)– ProtocolType: type of higher layer protocol (e.g., IP)– HLEN & PLEN: length of physical and protocol addresses

• Stands for “hardware address length” and “protocol address length”

– Operation: request or response – Source/Target-Physical/Protocol addresses

• Notes– table entries timeout in about 15 minutes (why?)– update table with source when you are the target (why?)– "refresh" table if already have an entry (i.e. reset timeout)– do not refresh table entries upon reference

Spring 2010 CS 332 26

ARP Packet Format

TargetHardwareAddr (bytes 2 – 5)

TargetProtocolAddr (bytes 0 – 3)

SourceProtocolAddr (bytes 2 – 3)

Hardware type = 1 ProtocolType = 0x0800

SourceHardwareAddr (bytes 4 – 5)

TargetHardwareAddr (bytes 0 – 1)

SourceProtocolAddr (bytes 0 – 1)

HLen = 48 PLen = 32 Operation

SourceHardwareAddr (bytes 0 – 3)

0 8 16 31

Spring 2010 CS 332 27

Dynamic Host Configuration Protocol (DHCP)

• Manually configuring IP information can be hard– Large networks

– Configuration process error prone• Every host needs correct network number

• No two hosts can have same IP address

• DHCP automates process– Network management has to scale, too – not just

network operation

Spring 2010 CS 332 28

DHCP (continued)

• At least one DHCP server per administrative domain– Centralized repository for host configuration info

• Info can be sent to hosts at boot or connection time.

• Can also be used to maintain pool of available addresses assigned on demand

• Method– Send DHCPDISCOVER msg to 255.255.255.255.

– Response is DHCP Offer message – also broadcast (why?)

– Host chooses one of offers and sends Reply and gets Ack

– Host "leases" IP address for a period of time – can renew

– Relay agents

Spring 2010 CS 332 29

Internet Control Message Protocol (ICMP)

• Communicates error messages and other conditions that require attention

• ICMP messages are acted on by either the IP layer or higher layers (TCP or UDP).

• Transmitted within IP datagrams

type checksumcode

Contents depends on type and code

0 1587 16 31

Spring 2010 CS 332 30

ICMP (cont.)

• 15 different values for the type field, then several codes for each of the types

• Checksum computed same as for IP packet• Contains first 8 bytes of IP datagram that

generated the message so sender can ID• Complete specification of protocol is RFC 792

(Postel)• Another good source is TCP/IP Illustrated, Vol. 1,

Ch. 6

Spring 2010 CS 332 31

Types of ICMP Messages

• Echo (ping)• Redirect (from router to source host)• Destination unreachable (protocol, port, or host)• TTL exceeded (so datagrams don’t cycle forever)• Checksum failed • Reassembly failed• Cannot fragment

Spring 2010 CS 332 32

Virtual Private Networks (VPNs)

• Goal: simulate private network of dedicated links on a public (shared) network

• Easy on circuit-switched infrastructure• Not so easy on IP-based internetwork• Strategy: routers create an IP "tunnel" for VPN

traffic.

Spring 2010 CS 332 33

IP Tunneling

• Router R1 at one end of tunnel is provided with IP address of router R2 at the other end.

• R1 is directly connected to the network where the host that requested the tunnel lives.

• R2 is directly connected to the network where the destination host lives.

• All VPN traffic is encapsulated as IP packets from R1 to R2 (this is not the norm)

• R2 strips tunnel header and forwards packet to ultimate recipient.

Spring 2010 CS 332 34

Tunnel implementation

NetworkNum NextHop

1 Interface 0

2 Virtual interface 0

Default Interface 1

Routing tablefor router R1:

Spring 2010 CS 332 35

What is it good for?

• Security – tunnel plus encryption• Using routers that have enhanced capabilities, e.g.

multicast• Tunneling other protocols across IP• Short-circuiting normal routing – useful for

mobility