D-Link DES-3326 - Manual (Page 56)

DES-3326 Layer 3 Fast Ethernet Switch User’s Guide
For IP to work, the local system must be attached to a network. It is safe to assume that any system on
this network can send packets to any other system, but when packets must cross other networks to
reach a destination on a remote network, these packets must be handled by gateways (also called
routers).
Gateways connect a network with one or more other networks. Gateways can be a computer with two
network interfaces or a specialized device with multiple network interfaces. The device is designed to
forward packets from one network to another.
IP routing is based on the network address of the destination IP address. Each computer has a table of
network addresses. For each network address, a corresponding gateway is listed. This is the gateway to
use to communicate with that network. The gateway does not have to be directly connected to the
remote network, it simply needs to be the first place to go on the way to the remote network.
Before a local computer sends a packet, it first determines whether the destination address is on the
local network. If it is, the packet can be sent directly to the remote device. If it is not, the local computer
looks for the network address of the destination and the corresponding gateway address. The packet is
then sent to the gateway leading to the remote network. There is often only one gateway on a network.
A single gateway is usually defined as a default gateway, if that gateway connects the local network to a
backbone network or to the Internet. This default gateway is also used whenever no specific route is
found for a packet, or when there are several gateways on a network.
Local computers can use default gateways, but the gateways themselves need a more complete routing
table to be able to forward packets correctly. A protocol is required for the gateways to be able to
communicate between themselves and to keep their routing tables updated.
Packet Fragmentation and Reassembly
TCP/IP can be used with many different types of networks, but not all network types can handle the
same length packets.
When IP is transmitting large files, large packets are much more efficient than small ones. It is
preferable to use the largest possible packet size, but still be able to cross networks that require smaller
packets.
To do this, IP can ‘negotiate’ packet size between the local and remote ends of a connection. When an IP
connection is first made, the IPs at both ends of the connection state the largest packet they can
handle. The smaller of the two is selected.
When an IP connection crosses multiple networks, it is possible that one of the intermediate networks
has a smaller packet size limit than the local or remote network. IP is not able to determine the
maximum packet size across all of the networks that may make up the route for a connection. IP has,
therefore, a method to divide packets into multiple, smaller packets to cross such networks. This
division of large packets into smaller packets is referred to as fragmentation.
A field in the TCP header indicates that a packet has been fragmented, and other information aids in
the reassembly of the packets into the original data.
Gateways that connect networks of different packet size limits split the large packets into smaller ones
and forward the smaller packets on their attached networks.
56