| The following is the argument which I put together for using the FDDI protocol
in a backbone as opposed to Fast Ethernet. I was greatly helped by several
people within DIGITAL who responded by mail to my base note (you know who you
are - thanks!), but any omissions or mistakes in the following are mine.
The technical merit of FDDI over Fast Ethernet as a backbone protocol is "a
no-brainer", but it would seem that technical considerations take second place
to the market's belief that FDDI is a technology on its way out and that Fast &
Gigabit Ethernet are the way to go.
I'm posting this in the hope that someone will find it useful.
-John
Fast Ethernet vs. FDDI
======================
FDDI is used as the backbone in most Fortune 500 companies, and is a proven,
mature, 100Mbps networking technology. An exclusive feature of FDDI is the
dual-ring architecture which provides built-in redundancy and superior
fault-tolerance.
FDDI's signalling rate is 125Mbaud, but the extremely reliable encoding
techniques deliver a 100Mbps data rate. FDDI also runs over UTP copper, and
adapters and uplinks are available from all major manufacturers. FDDI does not
have the distance limitations of Fast Ethernet, and is used extensively in
corporate backbones. Built-in redundancy, topological flexibility, exceptional
reliability, and extensive distances make FDDI an ideal choice for a network
backbone.
Fast Ethernet is effectively Ethernet at a much higher clock-speed. The
tradeoff is that increased speed is achieved at the expense of smaller
distances. Fast Ethernet is built on the same MAC layer found in Ethernet, and
uses exactly the same packet format which is limited to 1500 bytes in size.
Collision detection is still used, and Fast Ethernet uses the same (rather
long) delay before attempting to retransmit.
A station transmitting data requires that the data be sent as rapidly and
possible. Ethernet cannot guarantee that another station is not going to
transmit at the same time, thus causing a retry. Furthermore, another station
can interrupt before the first has finished sending everything - for relatively
obvious reasons of equity of bandwidth. Fast Ethernet requires an inter-frame
gap of almost one microsecond, and there are also a significant number of
overhead bytes in its design. In fact, Fast Ethernet's protocol efficiency
(i.e. the percentage of the bus bandwidth that carries user data) ranges from
65% to 87%, depending on packet size.
FDDI on the other hand, is a timed token-passing mechanism with a protocol
efficiency close to 99%. FDDI is very efficient for a variety of topologies,
independent of packet size. For normal packet sizes, roughly 256 bytes or
greater, it provides 16% to 30% greater throughput than other protocols of the
same media bit rate.
Packets can be 4500 bytes rather than Ethernet's 1500. While a node holds the
token (a fairly long time), it can transmit as much data to as many targets as
it wants. FDDI also has integral management capability for fault identification
and isolation (SMT).
Some direct comparisons between FDDI (over copper) and Fast Ethernet:
1. FDDI uses an intelligent hub that monitors each link and ensures complete
network integrity. If an interface card malfunctions, it will be automatically
removed from the network. If a NIC in an Ethernet network malfunctions, the
whole network can be brought to a standstill.
2. Collisions (and hence "runts") do not occur in an FDDI network.
3. Ethernets must be designed so as not to exceed a certain length, both for
timing and electrical reasons. Although copper cabling and termination problems
also affect FDDI, its architecture and protocols ensure a lower error rate.
4. FDDI guarantees that a node can transmit within a given time (negociated
when the node joins the ring), whereas Ethernet does not. Also, in networks
with a small number of nodes (e.g. the trunk between a switch and a host), the
phenomenon known as the Ethernet Capture Effect can occur. This is a situation
in which a station can be "locked-out" by another that continuously has
something to transmit.
5. With dual-homing, FDDI can failover to a redundant link almost
instantaneously, whereas Ethernet relies exclusively on Spanning Tree for
redundancy. Spanning tree reconfigurations take at least 30 seconds, depending
on the complexity of the network.
6. Fast Ethernet is a relatively new technology for most vendors and there has
not been a lot of inter-operability testing done yet. FDDI, on the other hand,
is a mature technology with support from - and guaranteed interoperability
between - every major network manufacturer.
The advantage of FDDI is maturity and performance. The advantage of Fast
Ethernet is lower cost, which is not necessarily a desirable backbone trait.
Ethernet switches have no inherent advantage; a switch is a switch, the data
link isn't the issue.
|