Network
Architecture
A network's architecture
generally defines its overall structure, including its topology, physical
media, and channel access method. The following is a brief summary of the
more popular architectures used in networking to day.
Ethernet
Developed by Xerox, Ethernet is
the most popular network architecture today. It has many advantages,
including ease of installation and lower costs. Ethernet is generally less
expensive than most other architectures. Another reason that it is so
popular is that it can support the use of many different media types
(cable). Ethernet uses a channel access protocol called CSMA/CD (Carrier
Sense Multiple Access with Collision detection). Simply put, this protocol
oversees the transmission of data across the wire. If a machine is
transmitting, it is not possible for another machine to transmit at the same
time. It must wait till the medium is free. If it does transmit, the data
sets will collide, causing a garbled signal. So the role of Carrier Sense is
to be able to detect if the wire is available. Multiple Access permits
multiple machines to share the wire, while Collision Detection takes care of
any collisions that do occur, and provide the machines involved another
chance at transmitting their various data again. Ethernet networks run on a
bus topology, or more accurately, a star-bus, which physically is a star,
but logically operates like a bus, i.e. a broadcast system. Ethernet
operates at speeds of 10Mbps and newer standards support 100Mbps, these are
the two categories that divide Ethernet, based on transmission speeds and
media use.
10 Mbps Standards
There are four major
implementations of 10Mps Ethernet:
· 10Base5: Ethernet using thick
coaxial cable with a maximum segment length of 500mtrs
· 10Base2: Ethernet using thin
coaxial cable with a maximum segment length of 185mtrs
· 10Base-T: Ethernet over
unshielded twisted-pair (UTP) cable with a maximum cable length of 100mtrs
· 10Base-F: Ethernet over
fibre-optic cable with a maximum cable length of 2000mtrs 100 Mbps Standards
· 100VG-AnyLan: Emerging
Architecture that is a mixture of Token Ring and Ethernet. Uses Fiber and
UTP. Cable lengths of 100, 150 and 2000 meters.
· 100Base-T: (fast Ethernet)
This standard has three sub
categories.
o 100Base-T4: Four-pair Cat 3, 4
or 5 UTP; Max cable length 100mtrs.
o 100Base-TX: Two-pair Cat 5 UTP.
Max cable length 100mtrs.
o 100Base-FX: Two -strand
fibre-optic cable. Max cable length 2000mtrs
Token Ring
The Token Ring architecture was
developed by IBM in the mid-1980's, providing users with fast, reliable
transport. Token Ring is so called, because an empty data frame continually
circulates the network, and any node wishing to transmit, would seize the
token, and put its data onto the network with the address of the intended
node. By using the token passing channel access method, token ring networks
ensure that all computers get equal time on the network.
As the frame circulates the
network, each node examines the address field of the frame in order to
determine whether or not the frame is intended for it. If not, the node
allows the frame to pass to the next node. This process continues until the
intended node receives the frame, at which point the node takes a copy of
the data and releases the frame back onto the network. When it returns to
the originating node, the frame is removed and a new frame is generated. The
new empty frame is left to circulate the network until a node waiting to
transmit seizes it.
Unlike Ethernet, there are no
collisions, so data seldom has to be re-sent. Because all computers on the
network have equal access to the token, traffic is consistent and token ring
handles increases in network size gracefully.
The newest versions of token
ring operate at speeds of up to 16Mbps. Because collisions never occur,
token ring can handle larger packets sizes than Ethernet. This allows large
blocks of data to be transferred. Token ring networks run on a ring
topology, or a star-ring topology, i.e. a physical star configuration, but
the hub device logically operates like a ring were the token is circulated
from port to port in an infinite loop.
Fibre Distribute Data Interface
(FDDI)
This architecture is installed
in high-demand networks, and is a very reliable solution. FDDI uses the
token-passing channel access method while using dual rings for fault
tolerance. That is to say, if one cable breaks, the other is used to work
around the problem, thus keeping the network alive. FDDI transmits at
100Mbps and can include up to 500 nodes over a distance of 100km (60miles).
FDDI networks are wired as a physical ring, it has no hubs, and machines are
generally directly connected to each other by means of fibre-optic cabling.
Unlike token ring, machines in
this solution are not required to wait for the token to make a full circle
before transmitting another token. When a computer possessing the token has
more than one data frame to send, it can send additional tokens before the
initial frame completes its journey. This allows data to be transmitted
around the network more quickly. Also, once the computer has finished
sending its token, it can immediately pass the token along. Again, it need
not wait for the token to complete its circuit around the network. FDDI
network use a ring topology, and occasionally a star-ring.
Unlike token ring networks,
FDDI permits administrators to assign priorities to certain nodes, for
example a server running time-sensitive data or video. |
