The introduction of gigabit (1 Gb/s) Ethernet has allowed the introduction of “Jumbo”
frames with more than 1500 data bytes. Clearly the overhead involved by limiting the
number of bytes to 1500 is considerable, and using jumbo frames makes more efficient
use of the network. However, the following should be noted:
• Many real world networks only operate at 100Mb/s at local level and these cannot
accept jumbo frames.
• Many gigabit networks are not configured to accept jumbo frames. Only networks so
configured can use them.
• The maximum data size of a jumbo frame is 9000 bytes and typically jumbo frames
carry 8000 data bytes.
• While network efficiency goes up with larger frames, the effect of a lost frame or
packet is more serious.
Nodes, Switchers, and RoutersAn Ethernet Local Area Network (LAN) will have a number of nodes and, in principle, all
nodes can communicate with each other. The principle used is “Carrier Sense, Multiple
Access/Collision Detection” or CSMA/CD. This means that when not transmitting, all
nodes are listening. When a node transmits, no other node attempts transmission.
However, signal speed limits mean that a “collision” is possible. Such collisions are
detected and the competing parties “back off” for another attempt. The principle works
well for small networks, but introduces inefficiency in networks with high traffic.
Networks are constrained by the use of various switching and routing devices.
• An Ethernet hub simply allows nodes to be connected together and CSMA/CD
applies.
• An Ethernet switch intelligently routes internode traffic (nodes only receive traffic
addressed to them). This reduces or eliminates bus contention at the local level. A
switch can also allow a node to operate duplex communication (simultaneous transmit
and receive).
• An Ethernet bridge is a two port switch used for segmenting networks or joining
dissimilar media.
• An Ethernet router connects multiple networks and connects to networks of other
types.
Routers and switches use routing tables to determine how traffic is directed. These can be
dynamic, in the sense that they are generated as needed by examining the traffic. They
can also be static, imposing strict rules about how traffic is directed.
This factor is of great importance with respect to transmitting image data over networks.
In practice, unless a network is specifically programmed to carry image data, it is likely that
the data will be blocked at the first router it encounters.
ProtocolsFor communication to work over networks, there must be some formality about how
communication is done in order to ensure interoperability between different systems.
The ISO proposed a model for this in its Open Systems Interconnection (OSI) model. It
defines seven different “layers” for any intercommunication protocol, starting at the
bottom with a “physical” layer (which might be Ethernet, wireless or some standard serial
communications method) going to the top which is the “application” layer relating to the
actual task in hand.
VN-Matrix 200 Series • Understanding Network Performance 98