An example NL responsibility - packet fragmentation and reassembly

Messages may be divided into smaller data units, termed packets, before transmission. These packets may traverse the network independently until they reach the destination node.

This requires fragmenting or segmenting the messages into packets at the source node and then reassembling the packets into messages at the destination.

Furthermore, each packet's header must indicate that it is part of some bigger message.

An important consideration for fragmentation is determining the 'optimal' packet size.

Compared to variable-sized packets, fixed size packets offer a number of advantages:

• Throughout the network, each node's buffer size may be fixed.
• It is simpler to prevent congestion at a destination node, since the destination may accurately estimate the number of buffers to pre-allocate.
• Fixed-sized packets result in simpler memory allocation schemes - typically 'once-only' allocation can be used instead of constantly using 'random-sized' fully dynamic allocation.

The primary disadvantage with fixed-sized packets is under-utilization of memory when the message size slightly exceeds an integral multiple of the packet size (for example, consider the very short last packet in the figure).