Network Layer

Layer 3 is responsible for routing packets to their destination and for Quality of Service (as opposed to Transport Layer, which is responsible for specifying port numbers and data reliability).

The IP (internet protocol) is the best known Layer 3 protocol, which functions as a connection-less protocol with no acknowledgement (thus the use of Layer 4 TCP connections, for example). Other Layer 3 protocols include ICMP and IPSec.

IP Addressing

IP addressing is a logical addressing scheme implemented at Layer 3; we use IP addressing to partition networks into smaller “subnets”. This can improve the performance and security of the system - for example, having all servers of one kind in a single subnets means that data won’t be exiting that subnet, and will also allow for easier access control into the subnet.

IP addresses differ from MAC addresses because:

1. IP are Layer 2, MAC are layer 2
2. IP uses logical addressing, MAC uses one big flat scheme (and thus no separation is possible)

The IP Header

Header data	Description
4-bit version	IPv4 vs IPv6
4-bit hdr length	Describes the length of the header for some optional info
Type of Service	Used for QoS information
16-bit total length	Total length of packet in bytes
16-bit ID	Used for fragment info.

The 8-bit TTL is a counter that gets -=1 every time it goes through a router. If the TTL == 0, then the packet is dropped in order to avoid routing loops.

The 8-bit protocol defines the Layer 4 protocol.

Note

The 32-bit destination IP address can also be a FQDN, which gets resolved into IP address via DNS

Unicast, Multicast, and Broadcast

IP traffic has three primary types: unicast, multicast, broadcast:

Traffic type	Destination(s)
Unicast	Single destination host
Multicast	Multiple interested hosts
Broadcast	All hosts on subnet

Unicast

Here traffic simply goes from a sender to a single host. Note that we can also have unicast traffic to multiple hosts: it will send separate copies of the same data to multiple hosts.

In other words, this is separate transmissions of the same data - assuming each packet is 1 MB, we would require 3 MB to send the same data to three separate hosts.

Broadcast

The sender sends one message to a switch. The message then gets flooded to all other hosts in the same subnet (meaning it will not go past routers - they do not forward broadcasts, since routers would forward all traffic from the internet, for example).

Multicast

This sends a single copy of data that gets sent to multiple hosts. This is different from unicast to various hosts because the sender does not need to create multiple transmissions.

Note that multicast does traverse subnets. Multicast requires that the receiving hosts request the data.

Decimal to Binary

When adding a column to the left in binary, the value is always multiplied by 2. Thus the number we are trying to convert to binary would have to be fit in a table such as:

32	16	8	4	2	1
Thus for example, if we want to convert the number 236 into binary we fill out a table that begins with the first column value higher than our conversion number (so in this case, our binary table starts at 256). We then see our numbers fits in the table, subtract the column value and move over (the numbers in brackets are the result of subtracting the column value from the number):

256	128	64	32	16	8	4	2	1
0(236)	1(108)	1(44)	1(12)	0(12)	1(4)	1(0)	0	0
Thus 236 in binary is 011101100.

Note that when doing this conversion there should always be 0 left over after the last bit is assigned. We can also add up all the column values where we have a 1 and it should add up to our original number: 128+64+32+8+4 = 236.

IPv4 Addresses

An IPv4 address is 32 bits long written as 4 octets in decimal format where each octet is 8 bits long. When written in decimal, each of the octets has a value between 0 and 255 (or a total of 256 values).

Note

To show IP addresses on IOS we can use the show ip interface brief command from an enable prompt.

To convert an IPv4 address to binary simply apply the Decimal to Binary process to each octet. For example, 192.168.10.15 converts to 11000000.10101000.00001010.00001111

IP Address Classes

When looking at the host portion of a network, the bigger the Subnet masks (that is, the more bits of the address it occupies), the fewer amount of hosts we can have. For example:

• Subnet mask = /8, we have 24 bits available for hosts
• Subnet mask = /24, we have 8 bits available for hosts

Note

The IP addresses in the IP class scheme include various ranges, but there are also reserved Private Addresses in each class. Private addresses are valid for host assignment, but not routable on public internet (and thus should have no internet access). Private ranges are the following:

1. Class A: 10.0.0.0 to 10.255.255.255
2. Class B: 172 .16.0.0 to 172.31.255.255
3. Class C: 192.168.0.0 to 192.168.255.255

Class A

Network addresses always start with 0, and default subnet mask is /8 (first octet). Valid network addresses from 1.0.0.0 to 126.0.0.0/8

These IP addresses are assigned to networks with very large numbers of hosts. To accomplish this, the higher order bit (that is, the first bit in the address) is always set to 0, with a default subnet mask of /8.

Because the high order bit is set to 0, the highest value that can be set in the first IPv4 octet is 127 (that is, 64 + 32 + 16 + 8 + 4 + 2 + 1). However, 127.0.0.0 is a reserved address, and thus the highest network address available is 126.0.0.0.

What this means is that we can have up to 126 networks and up to 16,777,214 hosts per network, since the host addresses range from 1.0.0.1 to 126.255.255.254 (note that 126.255.255.255 is the broadcast address for that network).

Reserved Class A Addresses

• 0.0.0.0/8 signifies “this network”

  • Thus, 0.0.0.1 to 0.255.255.255 are not valid host addresses
  • This also means that there are 16 million valid addresses that could be used to signify “this network”

• 127.0.0.0/8 is reserved as the loopback address for testing local computer

  • Thus, 127.0.0.1 to 127.255.255.255 are not valid host addresses either
  • Note that the “loopback” address therefore contains 16 million valid network addresses

Class B

Network addresses always start with 10 and the default subnet mask is /16 (second octet). Valid network addresses from 128.0.0.0 to 191.255.0.0/16.

Class B addresses were originally assigned to medium/large networks. This allows for 16,384 networks and 65,534 hosts per network.

Class C

Network addresses always start with 110 and default subnet mask is /24 (third octet). Valid network addresses from 192.0.0.0 to 223.255.255.0/24.

Class C addresses were originally assigned to small networks. This allows for 2,097,152 networks and 254 hosts per networks.

This is a reasonable size that could be allocated for a real world network, or subnetted into smaller subnets.

Class D

Class D addresses are reserved for IP multicast addresses. Their high-order bits always start with 1110 and thus compose addresses from 224.0.0.0 to 239.255.255.255. These are not allocated to hosts and have no default subnet mask.

Note

The range of 224 to 239 is derived entirely from the first octet. 224 = 11100011 while 239 = 11101111 (the zero is required because Class D addresses always start with 1110)

The way multicast works is that the sender will still send out a package with a SRC and DST header (as it does in unicast, normal traffic). However, the destination header will be set to an IP address in the multicast range (for example, 239.0.0.1). This will then be repeated by the switch and received by hosts that are “interested” in the data.

This is a lot like tuning into a radio station. The interesting destination hosts run an application that defines an interest in data from the 239.0.0.1 address, so they subscribe (note that this would have to be configured in the routers). This is useful to save on bandwidth.

Class E

Class E addresses are “experimental and reserved for future use”. They have high-order bits set to 1111, and have neither hosts nor default subnets. They range from 240.0.0.0 to 255.255.255.255 (although the 4x255 address is the broadcast address for this network, and is the only pre-defined ).

Note

255.255.255.255 is the broadcast address for this network. Meaning that whatever network the host is on, messaging 4x255 will broadcast the message.

Link to original