The Network Layer

1
The Network Layer
2
Responsibilities

• Addresses
• Routing
• Fragmentation and reassembly

3
Network layer addresses

• IP address structure (v4)
• Class A 0xxxxxxxyyyyyyyyyyyyyyyyyyyyyyyy
• Class B 10xxxxxxxxxxxxxxyyyyyyyyyyyyyyyy
• Class C 110xxxxxxxxxxxxxxxxxxxxxyyyyyyyy
  
• Multicast1110xxxxxxxxxxxxxxxxxxxxyyyyyyyy
• Reserved 1111xxxxxxxxxxxxxxxxxxxxyyyyyyyy
  

4
IP v4 Class A

• Class A 0xxxxxxxyyyyyyyyyyyyyyyyyyyyyyyy
• 27 networks
• each with up to 224 hosts attached
• Not quite. Addresses of all 0 or all 1 are
  special cases and not permitted for general use

5
IP v4 Class B, C

• Class B 10xxxxxxxxxxxxxxyyyyyyyyyyyyyyyy
• 214 networks
• each with up to 216 hosts
• - again, not quite.
• Class C 110xxxxxxxxxxxxxxxxxxxxxyyyyyyyy
• 221 networks
• each with up to 28 hosts (approximately)

6
Non unique addresses

• Growth of the Internet has placed demands on the
  address space not anticipated originally. There
  are more machines than addresses available.
• Some machines do not need a unique address,
  because they do not communicate over the
  Internet.
• Addresses are set aside to be used as desired for
  those machines
• 10.0.0.0 - 10.255.255.255 (10/8 prefix) Start
  with 10, use 8 bits
• 172.16.0.0 - 172.31.255.255 (172.16/12 prefix)
• 192.168.0.0 - 192.168.255.255 (192.168/16 prefix)

Ref. RFC 1918
7
Network and Host addresses

• The network address identifies a network
  comprised of multiple computers and other
  devices.
• Routers deal with network addresses.
• Once the transmission reaches the right network,
  the local network protocols deal with delivery to
  the correct machine.
• The host address identifies a particular
  machine-to-network connection.

8
Subnets

• Once upon a time, 254 hosts per network seemed
  pretty reasonable
• That was before PCs
• Class C networks are not large enough for most
  kinds of organizations
• Multiple Class C networks in a single
  organization imposes management overhead

9
Subnets (2)

• Subnetting allows an organization to subdivide a
  network internally. The internal networks
  continue to look like a single network from
  outside the organization
• Take some bits from the host part of the IP
  address and make them part of the network part
  for internal routing

10
The subnet mask

• Allows the routers to know how many bits are part
  of the network address and how many are part of
  the host address
• Example a class B network is subnetted so that 5
  bits of the host address are part of the network
  address

1
10011001
1101000
00011111
00000110
Network
Host
2
Network
Host
3
1 the 32 bit address 2 network/host
division without subnetting 3 network/host
division with subnetting
Mask tells which bits to consider part of the
network address 1 in each net address position
0 elsewhere. Mask for the example is 11111111
11111111 11111000 00000000 Dotted decimal
representation 255 255
248 0
11
Subnets (3)

• The old network addresses for some of our
  machines
• Tiger 153.104. 7.161
• wild 153.104. 1. 10
• renoir 153.104. 7.174
• camille 153.104. 7. 1
• tanner 153.104. 7.178
• hawk 153.104. 8. 50
• cassel 153.104. 7.181
• smurfs 153.104.24. 32

What class network?
153 10011001 Class B network
Any indication of subnetting?
12
Current subnetting

• Mendel (CSC)
• Within VUs 153.104
• IP address range start 200.1
• IP address range end 203.254
• What is the subnet mask?

255.255.252.0
13
Subnetting and DSL

• Some DSL providers offer static IP addresses in
  groups of 8 (really?)
• What does that mean in terms of IP subnetting?
• One possibility
• A class C network is divided among a group of
  subscribers. Each gets a subnet mask that allows
  8 addresses.
• Addresses 000 and 111 are not legal IP addresses,
  though.

14
IP v6

• 128 bit addresses
• written as 8 parts, separated by
• each part is 6 bits, expressed in hex
• (no more dotted decimal)
• Notes
• space reserved for other address schemes
• place to imbed the local link address
• multicast, anycast, no broadcast

15
IP V6 packet layout
Flow Label
Version
Priority
Next Header
Hop Limit
Payload Length
Source Address
Destination Address
Version 6 Flow label connect packets from
the same source Payload Packet size in bytes
Next header Next layer up connection
(Protocol) Hop Limit time to live in hops
16
Routing - Link State overview

• Each routing node obtains
• the information concerning the immediate
  neighbors of each other node in the network
• Once this information is available, the node
  constructs a graphical representation of the
  internet

17
Routing - Link State details

• Enter self into table
• Enter data from immediate neighbors
• mark this data tentative (T)
• For each node marked T in the table, examine the
  connection information about that node and enter
  it into the table.
• Consider T nodes in order of cost to get there,
  least costly first
• Previously unknown nodes are added
• Previously known nodes are examined to see if a
  better route is found

18
Routing - Distance Vector

• Each router node knows about itself
• the distance to itself 0
• first entry in the routing table
• Each router knows about its directly connected
  neighbors
• the distance to a direct neighbor 1
• next set of entries
• Exchanging information with neighbors extends the
  diameter of the known universe to each router

19
Our sample network
20
A special problem
What happens if we apply the link state protocol
to the following special situation
C
A
B
1. Determine the routing table entries for each
router A, B, C
2. Assume the connection between B and C is
broken
3. Show the steps by which A, B revise their
tables
This is the counting to infinity problem
21
Border Gateway Protocol

• See http//www.cisco.com/univercd/cc/td/doc/cisint
  wk/ics/icsbgp4.htm
• For complete information on BGP
• BGP is a link state protocol
• BGP is run between autonomous systems, rather
  than within autonomous systems
• Instead of using a cost metric, the BGP messages
  contain an entire route to the destination

22
Routing within the VU domain
153.104.0.249
Connection to our service provider
153.104.0.254
153.104.0.18
153.104.0.19
Internal routers
...
153.104.0.1
153.104.200.1
153.104.203.1
153.104.202.1
153.104.201.1
How would you fill in the missing numbers?
23
Routing from Renoir out
1 153.104.200.1 (153.104.200.1) 0.825 ms 0.631
ms 0.590 ms 2 153.104.0.1 (153.104.0.1) 1.024
ms 0.724 ms 0.701 ms 3 153.104.0.254
(153.104.0.254) 1.053 ms 1.382 ms 1.801 ms 4
207.68.14.11 (207.68.14.11) 6.086 ms 9.067 ms
6.155 ms 5 205.171.38.85 (205.171.38.85) 8.062
ms 10.089 ms 12.455 ms 6 nyc-core-03.inet.qwes
t.net (205.171.17.121) 11.345 ms 10.354 ms
10.395 ms 7 nyc-core-01.inet.qwest.net
(205.171.17.82) 10.308 ms 17.639 ms 8
wdc-core-02.inet.qwest.net (205.171.5.235)
19.174 ms 16.058 ms 17.888 ms 9
wdc-core-03.inet.qwest.net (205.171.24.6) 20.636
ms 20.425 ms 21.594 ms 10 hou-core-01.inet.qwes
t.net (205.171.5.187) 36.128 ms 43.064 ms
44.321 ms 11 hou-edge-07.inet.qwest.net
(205.171.23.14) 37.849 ms 41.555 ms 41.659
ms 12 205.171.36.154 (205.171.36.154) 52.102 ms
50.555 ms 52.055 ms 13 192.12.10.60
(192.12.10.60) 49.084 ms 49.554 ms 46.130
ms 14 ser9-msfc1.gw.utexas.edu (128.83.2.9)
50.420 ms 50.396 ms 46.334 ms 15 128.83.37.18
(128.83.37.18) 49.908 ms 57.542 ms 50.448
ms 16 cs.utexas.edu (128.83.139.9) 50.164 ms
46.581 ms
24

• traceroute Christie.netlab.csc.villanova.edu
• traceroute to Christie.netlab.csc.villanova.edu
  (153.104.203.200), 30 hops max, 38 byte packets
• 1 pm40.iwaynet.net (198.30.105.210) 117.453 ms
  109.666 ms 119.863 ms
• 2 icg-gw.iwaynet.net (198.30.105.193) 119.719
  ms 109.765 ms 139.856 ms
• 3 oeb7-sl0-0-0c10.columbus.oar.net
  (199.18.98.37) 129.763 ms 118.785 ms 109.832
  ms
• 4 oeb9-atm1-0.columbus.oar.net (199.18.202.19)
  119.748 ms 129.768 ms 109.871 ms
• 5 208.46.62.49 (208.46.62.49) 139.748 ms
  139.751 ms 149.855 ms
• 6 chi-core-03.inet.qwest.net (205.171.20.33)
  129.769 ms 129.782 ms 159.867 ms
• 7 chi-core-02.inet.qwest.net (205.171.20.29)
  159.762 ms 139.801 ms 119.864 ms
• 8 nyc-core-02.inet.qwest.net (205.171.5.249)
  149.749 ms 139.759 ms 149.839 ms
• 9 205.171.17.118 (205.171.17.118) 139.753 ms
  169.741 ms 159.854 ms
• 10 205.171.38.62 (205.171.38.62) 149.753 ms
  159.793 ms 205.171.38.86 (205.171.38.86) 159.861
  ms
• 11 207.68.14.50 (207.68.14.50) 159.701 ms
  629.814 ms
• 12 153.104.0.249 (153.104.0.249) 179.816 ms
  159.723 ms 199.836 ms
• 13 153.104.0.18 (153.104.0.18) 169.751 ms
  159.807 ms 169.850 ms
• 14 153.104.0.18 (153.104.0.18) 1339.845 ms
  !H
• 15 153.104.0.18 (153.104.0.18) 1889.932 ms
  !H
• 16 153.104.0.18 (153.104.0.18) 1869.955 ms
  !H
• 17

Routing to Christie - attempt when netlab was
disconnected
25
Routing - scale

• How big is a routing table?
• Assume the current IP v4 address scheme
• Assume that subnets are internal and not the
  problem of internet routers
• What is the potential load on a router?

26
Classless Inter-Domain Routing

• First pass at hierarchical routing in the
  Internet
• Assign addresses in clumps that are not dependent
  on the old Class A, B, C scheme.
• Much more flexible in the allocation of space and
  able to serve the needs of users more efficiently.

27
CIDR address assignments
CIDR Block Prefix Equivalent Class
C of Host Addresses
/27 1/8th of a Class C 32
hosts /26
1/4th of a Class C 64 hosts
/25 1/2 of a Class C
128 hosts
/24 1 Class C 256 hosts
/23 2
Class C 512 hosts
/22 4 Class C
1,024 hosts
/21 8 Class C 2,048
hosts /20
16 Class C 4,096 hosts
/19 32 Class C
8,192 hosts
/18 64 Class C
16,384 hosts
/17 128 Class C 32,768 hosts
/16
256 Class C 65,536 hosts ( 1 Class B)
/15
512 Class C 131,072 hosts
/14 1,024 Class C
262,144 hosts
/13 2,048 Class C 524,288 hosts
CIDR Block Prefix Equivalent
Class C of Host Addresses
/27 1/8th of a Class C 32 hosts
/26 1/4th of a Class C 64
hosts /25 1/2 of a Class C
128 hosts /24
1 Class C 256
hosts /23 2 Class C
512 hosts /22
4 Class C 1,024 hosts
/21 8 Class C
2,048 hosts /20 16
Class C 4,096 hosts
/19 32 Class C 8,192
hosts /18 64 Class C
16,384 hosts /17
128 Class C 32,768
hosts /16
256 Class C 65,536 hosts ( 1 Class
B) /15
512 Class C 131,072 hosts
/14 1,024 Class C
262,144 hosts /13
2,048 Class C 524,288 hosts
28
A case
Currently, big blocks of addresses are assigned
to the large Internet Service Providers (ISPs)
who then re-allocate portions of their address
blocks to their customers. For example, Pacific
Bell Internet has been assigned a CIDR
address block with a prefix of /15 (equivalent to
512 Class C addresses or 131,072 host addresses)
and typically assigns its customers CIDR
addresses with prefixes ranging from /27 to /19.
These customers, who may be smaller ISPs
themselves, in turn re-allocate portions of their
address block to their users and/or customers.
However, in the global routing tables all
these different networks and hosts can be
represented by the single Pacific Bell Internet
route entry. In this way, the growth in the
number of routing table entries at each
level in the network hierarchy has been
significantly reduced. Currently, the global
routing tables have approximately 35,000 entries.
Ref http//public.pacbell.net/dedicated/cidr.htm
l
29
Network Address Translation

• Primary source for information RFC1631
• Goal Stand between the local network
  environment and the rest of the Internet

Local network environment
Router
The Internet
IP address
30
Why use NAT

• Non unique addresses on the internal network work
  fine for communication that does not involve the
  global Internet.
• To provide communication between a machine with a
  non unique address and the global Internet, the
  address must be translated into a globally unique
  address.

31
How dynamic NAT works
192.168.0.1
153.104.203.220
153.104.203.220
192.168.0.2
153.104.203.220
192.168.0.3
153.104.203.220
Internal network has non-unique IP addresses
NAT box has an address translation table and a
set of assigned IP addresses that can be used in
the Internet
32
Internal host external connection
192.168.0.1
153.104.203.220
192.168.0.3
153.104.203.220
192.168.0.2
153.104.203.220
192.168.0.3
153.104.203.220
Internal host requests connection to an external
host
NAT associates the internal address with a
globally unique address and makes the connection
with the external host
33
Response from external host
192.168.0.1
153.104.203.220
192.168.0.3
153.104.203.220
192.168.0.2
153.104.203.220
192.168.0.3
153.104.203.220
Response from external host connected to the
right internal host
Once there has been an exchange of messages, the
table has the mapping needed and further
communications are just passed through.
34
Overloading
192.168.0.1
153.104.203.220
192.168.0.3
Port 2000
Port 23
Port 2001
192.168.0.2
Port 2002
192.168.0.2/25
Port 2003
192.168.0.3
Port 2004
When there are not as many unique IP addresses as
internal hosts who may want to access external
hosts, add the use of port numbers in the table
35
Variable length subnet masks

• Originally, subnet masks were of a fixed length
• Clearly inefficient for an organization that has
  logical subnets of varying size
• Recent revisions of the routing protocol
  implementations allow variable length subnet masks

36
Fragmentation, reassembly

• Routers connect networks
• pass messages from one network to another
• Network characteristics are not all the same
• maximum packet size varies
• Routers must break up large packets to allow them
  to go into networks where the maximum allowed
  size is smaller
• Question Where to reassemble?

37
Reassembly question

• Should a router join packets to make larger ones
  when a fragmented transmission is leaving a
  network?
• Large packets require fewer routing decisions
  they are more efficient
• Reassembly and then later fragmentation are time
  consuming these activities should be minimized.

38
Network layer summary

• Addressing
• current most common is IP v4
• subnetting adds flexibility to network sizes
• Routing
• Link State and Distance Vector
• Fragmentation/Reassembly
• dealing with the restrictions of individual
  networks.