| >> The reason why several corporations are considering
>> Fast Ethernet again 100-Base-VG is that there is no need for
>> frame translations since the frame format is the
>> same as traditional Ethernet, apart from being a one-vendor
>> proposal (Hp, who by the way has began to invest in
>> Fast Ethernet >;-> ).
As far as I understand, both Fast Ethernet and VG use the standard
802.3 framing. That's why there was little to no transport or OS
impact in adopting these technologies.
/l
|
| >> Fast Ethernet wins the price segment against FDDI. But,
This is true only if you let the comparison be based on the cost of a NIC
and the cost of a concentrator port to connect it to.
If you look at over all _system_ cost which includes the overall network
cost plus the sum of all computer costs including the things needed to
achieve equal performance, then the difference in cost is not significant
or runs the other way.
Of course, this requires going against the conventional wisdom, often backed
by data from vendors who only make routers. "Truely well designed networks
connect segments by routers to ensure that the network is stable and so
that overloads in one place don't adversely affect the rest of the network."
If you then design the network around routers, then the cost of any significant
amount of FDDI is huge because the typical router isn't able to handle many
FDDI connections.
And FDDI's ability to offer better performance than Ethernet in the computer
requires protocols that are designed to take into account the properties
of the network. TCP/IP is not designed for this, but its still relatively
easy to build translating bridges that will make up for TCP/IP and offer
increased performance over Ethernet, as long as you buy good bridges.
Note that there is less overhead required in "translating" a packet from
Ethernet to FDDI than there is required to route a packet from Ethernet to
Ethernet. And when "translating" from FDDI to Ethernet, ie., fragmenting,
the cost is still less than routing. All the bit munging, etc. is done
by chips, so there is nothing more complicated about FDDI versus Ethernet.
The two big reasons for differences in cost between FDDI and Ethernet components
are volume and fiber optics. If you chose fiber as your infrastructure based
on the eventual availability of gigabit and higher networks, and then you
build a fiber based Ethernet network, you would find Ethernet to be more
expensive due to the cost of components and the increased pain from having
to convert from fiber to copper because fiber Ethernet is less common than FDDI.
The advantages of FDDI are
- higher performance in the "expensive" component: the CPU
because of fewer packets (FDDI allows bigger packets)
and because a single adapter is faster and less costly then
multiple slower adapters
- use of infrastructure that will never be obsolete
short Cat 5 UTP runs with core network built on fiber
- integral management capability for fault identification and isolation
The last point is important. How many times have you received notice that the
Ethernet service will be disrupted while the network people segment Ethernet
segments to isolate a problem cause network performance problems. That doesn't
happen with FDDI, nor does it happen with managed Ethernet hubs. Now, compare
the cost of managed Ethernet hubs to the cost of FDDI concentrators and the
price difference narrows significantly.
Basically, what it comes down to is what part of the system are you most
concerned about. If the only issue is how many computers can you buy for
a given number of dollars and you don't give a damn about how well the network
works, then, sure, Ethernet is the winner. On the other hand, if you are the
guy whose job depends on keeping the network up and you get to reject system
orders if they don't meet your standards, then FDDI is the most likely choice.
|