Introduction to Packet Fragmentation

Each network imposes some maximum size on its packets. The network designers are not free to choose any maximum packet size they wish as there are various factors like hardware, operating system, protocols, compliance with some (inter)national standard, desire to reduce error induced transmissions to some level and desire to prevent one packet from occupying the channel too long.
Packets larger than the allowable MTU (Maximum Transmission Unit) must be divided into multiple smaller packets, or fragments, to enable them to traverse the network.
If a packet that is about to be sent (for eg : over an Ethernet link) is bigger than that, the router which is about to send the packet over that link will fragment the packet i.e. the router will split the packet up into smaller messages (known as fragments) that are each small enough to be transmitted over the link. When the fragments arrive at their destination (the computer to which they are being sent), that computer can reassemble the fragments to recover the original message - assuming none of the messages are lost in transit.

How can be fragmentation avoided ?
If the option of "don't fragment" is set ON in IP version 4, and the router wants to send the packet over a link for which the packet is too large, the router will not send the packet at all. Instead, the router will send a message back to the sender of the packet that was too large. The sending computer can then respond to this by sending out smaller packets. This is known as "path MTU discovery".

Strategies for recombining fragments :
- Transparent Fragmentation : When an oversized packet arrives at gateway, the gateway breaks it into smaller fragments, each fragment is addressed to same exit gateway, where pieces are recombined. In this way passage through the small packet network has been made transparent.
Benefits : It maximizes bandwidth on higher links and deterministic fragmentation unlikely.
Drawbacks : Packets may be reassembled/fragmented, gateways more complex, performance gains bounded because the max TU will be the MTU of the first hop. Plus, the IP layer at the destination may still have to perform reassembly if the last link had a smaller MTU than the first link. Only use on links with unusually small MTUs.



- Non transparent Fragmentation : This strategy includes refraining the recombining of fragmented packets at intermediate gateway. Once a packet is fragmented, each fragment is treated as an original packet. All fragments are passed through the exit gateway. Recombination exists only at destination host.
Benefits : Multiple exit gateways can now be used and higher performance can be achieved.
Drawbacks : Overhead increases. Also, it requires every host to be able to do reassembly.