D-Link DES-3326 - Manual (Page 50)

DES-3326 Layer 3 Fast Ethernet Switch User’s Guide
Because these functions are required by a large number of applications, they are grouped into a single
protocol, rather than being the part of the specifications for just sending e-mail. TCP is then a library of
routines that application software can use when reliable network communications are required.
IP is also a library of routines, but with a more general set of functions. IP handles the routing of
packets from the source to the destination. This may require the packets to traverse many different
networks. IP can route packets through the necessary gateways and provides the functions required for
any user on one network to communicate with any user on another connected network.
The communication interface between TCP and IP is relatively simple. When IP received a packet, it
does not know how this packet is related to others it has sent (or received) or even which connection
the packet is part of. IP only knows the address of the source and the destination of the packet, and it
makes its best effort to deliver the packet to its destination.
The information required for IP to do its job is contained in a series of octets added to the beginning of
the packet called headers. A header contains a few octets of data added to the packet by the protocol in
order to keep track of it.
Other protocols on other network devices can add and extract their own headers to and from packets as
they cross networks. This is analogous to putting data into an envelope and sending the envelope to a
higher-level protocol, and having the higher-level protocol put the entire envelope into it’s own, larger
envelope. This process is referred to as encapsulation.
Many levels of encapsulation are required for a packet to cross the Internet.
Packet Headers
TCP
Most data transmissions are much longer that a single packet. The data must then be divided up
among a series of packets. These packets must be transmitted, received and then reassembled into the
original data. TCP handles these functions.
TCP must know how large a packet the network can process. To do this, the TCP protocols at each end
of a connection state how large a packet they can handle and the smaller of the two is selected.
The TCP header contains at least 20 octets. The source and destination TCP port numbers are the most
important fields. These specify the connection between two TCP protocols on two network devices.
The header also contains a sequence number that is used to ensure the packets are received in the
correct order. The packets are not numbered, but rather the octets the packets contain are. If there are
100 octets of data in each packet, the first packet is numbered 0, the second 100, the third 200, etc.
To insure that the data in a packet is received uncorrupted, TCP adds the binary value of all the octets
in the packet and writes the sum in the checksum field. The receiving TCP recalculates the checksum
and if the numbers are different, the packet is dropped.
50