Advanced IP Addressing Management

This lecture will cover the following points:
IPv4
Classful Addressing
Subnetting
IP Addressing Crisis
Classless Interdomain Routing CIDR
Route Aggregation and Supernetting
Supernetting and Address Allocation
Variable Length Subnet Masking VLSM
Classless and Classful Routing Protocols
Route Summarisation Overview
Route Flapping
Private IP Addresses
Discontiguous Subnets
Network Address Translation
IP Unnumbered
DHCP Overview
DHCP Operation
Configuring a DHCP Pool
Easy IP
Configuring IP Helper Addresses
IP Helper Address Example
IP Address Issues & Solutions
IPv6 Address Format
Summary

IPv4
IPv4 is a 32-bit number that describes the network and host identity of a particular interface on a network. Often seen as dotted decimal. Networks today are becoming classless in terms of IP addressing schemes and ignoring the older convention of classful addressing. This is to overcome the wastage of IP addresses by the classful system and also to help keep router lookup tables shorter by employing address aggregation wherever possible. this is helping to keep IPv4 from running out completely before IPv6 is introduced globally.

Classful Addressing
There are 5 main classes of addresses, A, B, C, D, and E. A, B and C are used fopr general purpose addressing. Class D addresses (224.0.0.0 to 239.255.255.255) are used for multicasting e.g. EIGRP, OSPF, RIPv2 multicasts (see multicast addresses).  Class E addresses (240.0.0.0 to 255.255.255.254) are used for experimental purposes, not for host addressing.

Class A addresses  have over 16 million possible host addresses and need to be subnetted to be of any use.
Class B addresses have 65534 addresses and also will need to be subnetted.
Class C addresses may need to be subnetted to suit the organisation.

Subnet Masking
See RFC 950. Express the dotted decimal IP address as a binary number. work out which part of the address relates to the network and which part relates to the host. Write the subnet mask directly below the binary version of the IP address.Use 1s for the network and 0s for the host identity. This will give the subnet mask for that host. Convert back to dotted decimal.

If a subnet mask is specified on a router, it forces the router to ignore the classful addressing scheme.

IP Addressing Crisis
Mathematically it can be shown that 50% of IPv4's address space is taken up by class A addresses, class B 25%, class C 12.5%, classes D&E 12.5%. Class A occupies addresses 1.0.0.0 to 126.0.0.0 giving 126 organisations the chance to have 16 million plus addresses - not very likely. Class B offers 16384 companies the chance to have 65534 addresses. Class C offers more addresses but each is limited to 254 hosts.
The IETF published two significant findings in 1992:
1.  Exhaustion of IPs - class B almost gone
2.  Class C will overload Internet core routers if completely allocated.
The solution was to retool IPv4 while waiting for IPv6 to be designed and implemented. IPv6 addresses have 128 bits. One significant solution for the interim was to introduce PAT - port address translation otherwise known as NAT with Overloading.

CIDR
CIDR was introduced in 1994 - see RFCs 1517, 1518, 1519, 1520. It improved the scaleability of IPv4 by:
 Route Aggregation and Supernetting
Supernetting is the same as route aggregation. For instance a contiguous set of IP addresses in a routing table such as:
155.24.0.0 /16
155.25.0.0 /16
155.26.0.0 /16
155.27.0.0 /16
155.28.0.0 /16
155.29.0.0 /16
155.30.0.0 /16
155.31.0.0 /16
Can be reduced to one entry of 155.24.0.0 /13

Work this out by writing the addresses in binary one above the other as shown:
10011011.00011000.00000000.00000000
10011011.00011001.00000000.00000000
10011011.00011010.00000000.00000000
10011011.00011011.00000000.00000000
10011011.00011100.00000000.00000000
10011011.00011101.00000000.00000000
10011011.00011110.00000000.00000000
10011011.00011111.00000000.00000000

Note that the first 13 bits are identical and only the last (32 - 13) 19 bits differ. Therefore we can express the whole set of addresses as:
10011011.00011000.00000000.00000000 /13

or 155.24.0.0 /13

The usefulness of this is that an ISP can allocate a contiguous address space to a company and then aggregate all of these addresses into one address to advertise to the Internet.

Supernetting and Address Allocation
Suppose a company requires 400 host addresses, a single class B address would waste over 16000 addresses. This could be solved by using 2 class C addresses, giving 508 host addresses. Assume that the company needs to route between the two networks. Internet routers would require 2 routing table entries for the company. The solution is to go classless and use supernetting. Here we would use CIDR and request the address block from the ISP. The ISP takes on the burden of managing its address space. Internet routers keep summary routes (supernet routes) to ISP networks. It is then the responsibility of the ISP to keep the routes to its own customers. This keeps Internet routing tables much smaller.

Example
207.21.54.0 /24  is 11001111.00010101.00110110.00000000
207.21.55.0 /24  is 11001111.00010101.00110111.00000000

As in the previous example, the first 23 (highlighted) bits are the same. This gives  one supernetted address of 207.21.54.0 /23


In the above figure, the ISP has a block of 256 class C addresses and allocates 2 of them to the company as shown. The ISP will use its router to hold details of its customers and takes the responsibility of routing packets to the correct destination. The ISP will only advertise its set of addresses 207.21.0.0 /16 back to the Internet.

From this example it can be seen that CIDR gives efficient allocation of addresses leading to a reduced number of routing table entries.

VLSM
This is sometimes known as subnetting a subnet. Suppose we have a company with 6 departments. We could divide a class C address up to suit this. Consider borrowing 3 bits from the host section of a class C address. /27 creates 8 subnets. This would waste addresses on serial links as these only require 2 addresses so we have 3 choices to prevent this:
  1. Use VLSM with a subnet mask of /30. This produces subnets with 4 addresses each. First and last are unusable being the network or wire address and the broadcast address, leaving 2 host addresses
  2. Use private addressing
  3. Use IP unnumbered
If we are splitting subnets, use 'no auto-summary' command with RIPv2 and EIGRP to ensure that subnet masks are sent along in the packets.

Classless and Classful Routing Protocols
In order to make use of classless addressing, it is necessary to send the subnet mask along with routing updates. The following table shows the protocols that are suited to classless addressing.
Classful
 Classless
RIPv1
 RIPv2
IGRP
 EIGRP
EGP
 OSPF
BGP3
 ISIS

BGP4

RIPv1 carries no subnet mask, has no authentication and broadcasts its updates on 255.255.255.255.
RIPv2 (RFC 1723) carries subnet mask, has authentication and multicasts its updates on 224.0.0.9.

To implement this use the command

router(config)#router rip
router(config)#version 2

Route Summarisation Overview
CIDR and VLSM promote route aggregation (summarisation). To make best use of this use EIGRP or OSPF as the routing protocol.

These techniques will help keep routing table entries down upstream of these networks. The above figure illustrates that the entire addressing scheme of the 3 rightmost routers can be advertised back to the Internet as one single address of 192.168.48.0 /20.

Route Flapping
This is a fault when  a router interface repeatedly goes up and down. Without summarisation, a network e.g. 192.168.56.0 /24 can flap and would burden its upstream router with updates. The route would repeatedly enter and then be removed from the routing table. this route would then be reported as reachable then unreachable to other routers further upstream.

Summarisation keeps flapping isolated and localised.

Private IP Addresses
(RFC 1918) These are sets of addresses that are not routed (by policy) on the Internet backbone and are discarded.

10.0.0.0 to 10.255.255.255
172.16.0.0 to 172.31.255.255
192.168.0.0 to 192.168.255.255

These addresses can be used on router serial interfaces on the Internet because they are neither source nor destination addresses. Packets can travel through links having private addresses. Use of private addresses saves using up globally unique addresses.
disadvantage: Cannot use ICMP, SNMP or TELNET for contacting these addresses. It is necessary to contact unique LAN addresses.

Discontiguous Subnets
This occurs when subnets from one major network are separated by a completely different network or subnet.


RIPv1 and IGRP cannot deal with this situation. Updates sent from B to A have no subnet information so B is seen as a route to 207.21.24.0 /24 but A is already connected to 207.21.24.0 /27 so rejects the route.

RIPv2 and EIGRP both perform automatic summarisation on classful boundaries. This is not always desirable especially in discontiguous subnets. Use the following command to disable automatic summarisation for RIPv1 and EIGRP

Router(config-router)# no auto-summary

If private addresses are used on Internet-connected networks, packets and routing updates should be filtered to avoid leaking RFC 1918 addresses between ASs. If two adjacent systems used addresses from 192.168.0.0 /16, routers would get confused with updates from both systems.

Network Address Translation
See RFC 1631. This technology swaps one IP address for another across a NAT device e.g. a UNIX system or a router. This is used on stub networks only. NAT can be static when an inside resource e.g. web server needs to be accessed from outside or dynamic where addresses for inside hosts are allocated automatically. This works on a 1 to 1 basis.

Port Address Translation, PAT allows many inside local addresses to map to one global IP address. This can also be known as Many to 1 NAT or Overloading. TCP and UDP port numbers are mapped to keep track of inside local addresses.

Inside local addresses are stored with the same port and a global IP address is chosen plus a new spare port. These can be re-mapped when a response is received.

Using IP Unnumbered
VLSM and private addresses cannot be supported by RIPv1 nor IGRP. IOS allows IP unnumbered and this can be used with classful routing.To operate, the IP address of another interface or a loopback interface is borrowed.

To implement:

Give E0 an IP number. Now go to int S1 & type:

ip unnumbered e0

Then do the same operation on the other end of the serial link.

This only works if:
 Drawbacks:
 The LAN address is 'borrowed' and applied to the WAN interface.

DHCP Overview
DHCP is used to configure desktop clients. IOS offers EasyIP, a combination of DHCP and NAT.
DHCP leases IP addresses to a client. On expiry, client makes another request and usually gets the same number. NT/ 2000/ 2003 server or UNIX/ LINUX are used for DHCP but IOS offers DHCP. The IOS version gives a default address lease life of 24 hours. These addresses are chosen form a specified pool of addresses.

The DHCP server in IOS can also supply:
 DHCP allows the ability to specify MAC addresses of clients and assign the same number each time (in the case of a server).
DHCP uses UDP ports 67 and 68 (the same ports as BOOTP) see RFC 951.

DHCP Operation

At boot, the client broadcasts a DHCPDISCOVER on UDP 67 and any DHCP server listening will reply with a DHCPOFFER on UDP 68. The client replies to the first DHCPOFFER that it receives. The DHCPREQUEST is broadcast to let all other DHCP servers know which offer was accepted by the client. The DHCP server then checks the availability of the given address with a default of 2 pings to that address. Finally a DHCPACK is sent back to the client by the DHCP server that issued the address.

DHCP has 2 major mechanisms:
1.  A protocol to deliver configuration parameters to hosts
2.  A mechanism to allocate addresses to hosts

There are 3 main methods to assign IP addresses to hosts:
1.  Automatic - Permanent and assigned by DHCP
2.  Dynamic - DHCP leases an address
3.  Manual - Administrator defines an address for the client - DHCP just delivers it

To turn on DHCP on a router use the command:

dhcp service

To get to DHCP pool configuration mode, use:

ip dhcp pool {name}  - where name defines the pool purpose e.g.engineering

Manual bindings are possible. IP addresses are mapped to MAC addresses of hosts in the DHCP database. Manual bindings are stored in NVRAM.

Configuring a DHCP Pool
DHCP is enabled by default. It can be toggled with

[no] service dhcp

Set up a DHCP pool and give it a meaningful name.
Specify the network
Return to global config and exclude addresses if the need arises (for servers etc).

Example

ip dhcp pool engineering
network 192.168.1.0 255.255.255.0
exit
ip dhcp excluded-address 192.168.1.1 192.168.1.10  (for the use of servers etc)

Set up DNS and NBNS and Default Gateway

ip dhcp pool engineering
dns-server {address 1 address 2 ...}
netbios-name-server {address 1 address 2 ...}
default router {address}
domain-name {domain} e.g. eng  lease [days[hours[minutes]| infinite]

Note that up to 8 dns and netbios server addresses can be specified

Useful commands to use with DHCP
The following show  commands can be useful when troubleshooting DHCP:

show ip dhcp binding [address]   gives a list of bindings
show ip dhcp conflict [address]   gives a list of address conflicts
show ip dhcp database [url]        shows recent activity
show ip dhcp server statistics     gives detailed statistics for DHCP server and sent and received messages

Easy IP
This uses DHCP itself to obtain a single WAN address from the ISP. Uses NAT Overload on router inside private LANs. Allows for 'plug-and-play' routing.

Using IP Helper Addresses
A router does not forward broadcasts by default. Broadcasts are used to locate TACACS+ (an authentication server), DHCP, DNS, TFTP servers. This means one of these servers per subnet unless IOS Helper Address is enabled.

Use the command:

ip helper-address

This relays broadcast messages for key UDP services across the router. Can be used to send a unicast to a certain IP address OR forward a directed broadcast to a specific network or subnet.

This technology increases packet duplication which reduces router and network performance in networks with redundancy.

UDP flooding should be used here if 2 routers connect to the same network. 1 is used for UDP broadcasts, 2 is used for UDP unicasts.
UDP flooding is more efficient than IP Helper.

Configuring IP Helper Addresses
Identify the router interface that will receive UDP broadcasts e.g.

int f0/0
ip helper-address [address]

Now specify the UDP broadcasts that will be forwarded

ip forward-protocol udp 53

This allows DNS broadcasts to be forwarded.

There are 8 ports that can be specified:
Time                                 37
TACACS+                        49
DNS                                 53
BOOTP & DHCP Server   67
BOOTP & DHCP Client    68
TFTP                               69
NBNS                              137
Netbios Datagram Service  138

If a service is required that is not in the list, use the following (e.g. for UDP on port 517)

ip forward-protocol udp 517

We can also subtract a service by using the no form of the above command, e.g.

no ip forward-protocol udp 37   (stops time broadcast forwarding)

IP Helper Address Example

A unicast helper-address will not be useful if Host A wants more than one service from the server farm.
A directed broadcast to 172.24.1.255 is required.

To get to the TACACS+ server, use ip helper address 172.16.1.2.

Check int E3 for Directed Broadcast Forwarding - it may be disabled. If so:

int E3
ip directed-broadcast

IP Address Issues & Solutions
There are several problems associated with IPv4
depletion of addresses especially class B
Growth of routing tables on the Internet

Solutions that have been proposed are:
Subnetting 1985
VLSM 1987
CIDR 1993

But none of these can save IPv4. Mobile devices and household appliances will cause the available addresses to run out.

Solution is IPv6 with 128-bit addresses. IPv6 will address the issues of address depletion, QoS, address autoconfiguration, authentication and security.

As long as IPv4 is viable, IPv6 will be kept away as it requires new hardware, software and administration methods.

See http://www.ipv6.org

IPv6 Address Format
See RFC 1884 & RFC 2373
These addresses will be written in Hex e.g.
1080:0000:0000:0000:0008:0800:200C:417A

Instances of repeated 0s can be removed, but BEWARE, only one set of repeated 0 fields may be omitted otherwise the original address cannot be recovered.

1080:0:0:0:0008:0800:200C:417A can be shortened to:

1080:: 8:800:200C:417A

IPv6 addresses are 128 bit identifiers for interfaces and sets of interfaces, not nodes.

There are 3 types of addresses:
Unicast - sent to ONE interface
Multicast - defines a set of interfaces to send to
Anycast - delivers to the first or nearest interface in an anycast group

There is a hierarchy in IPv6 addresses
Public Topology - a collection of providers that supply Internet connectivity
Site Topology - local to an organisation that does not provide connectivity to nodes outside itself
Interface Identifier - level specific to a node interface
bits
 3
 13
 8
 24
 16
 64
Name
 FP
 TLAID
 RES
 NLAID
 SLAID
 Interface ID

FP defines the type of address, UNICAST, MULTICAST, ANYCAST (e.g. 001 is UNICAST)
TLAID - Top Level Aggregation ID - identifies the authority responsible for address at the highest level of the routing hierarchy
RES - reserved for growth of TLAID or NLAID as growth continues
NLAID - Next Level Aggregation ID - identifies ISPs
SLAID - Site Level Aggregation ID - used by an organisation to create an addressing hierarchy and identify subnetworks
Interface ID - identifies individual interfaces on a link. Derived from IEEE EUI-64.

IPv6 also offers RFC 1918 addresses for local use only.

Summary

Subnet Masks, VLSM, Private Addressing and NAT technologies have been introduced to extend the life of IPv4 as long as possible before IPv6 has to be introduced.

Aggregation of addresses can help to simplify upstream router tables. NAT and DHCP are offered by IOS so that these services may be provided without a standalone server. NAT gives the ability to swap IP addresses at a boundary so that private addresses may be used within a stub network.

IOS also provides the mechanism of IP Helper Address to forward UDP broadcasts to servers on different subnets.

IPv6 addresses are 128 bits and contain fields to identify the type of traffic and Top, Next and Site Level IDs.