carrs@TROUT.NOSC.MIL (Stephen M. Carr) (03/30/88)
-------
Folks,
1. We are in the process of admiring the pros and cons of
implementing a fiber optic ring backbone versus a straight
vanilla 50 Ohm baseband coaxial cable backbone to support IEEE
802.3 CSMA/CD Ethernet and TCP/IP. This backbone would support
other subnets, as well as support a gateway for all subnets into
the DDN MILNET.
2. We had somebody propose installation of a fiber optic ring as
the backbone. They were adamant that this is an IEEE 802.3
CSMA/CD Ethernet solution that supports TCP/IP on top of it. I
was taken aback, for I never heard of IEEE 802.3 implemented in a
"ring" topology. Seems senseless to implement IEEE 802.3 in a
"ring" topology to begin with, since IEEE 802.5 Token Ring is
specifically matched to a ring topology.
3. Nevertheless, I am assured that just such a TCP/IP IEEE 802.3
CSMA/CD Ethernet solution in a fiber optic ring environment is
available from Fibercom Corporation.
4. Aside from the fact that at first blush it appears silly to
me to even try to implement IEEE 802.3 in a ring topology, I have
other misgivings.
a. The cost per device connection would be significantly
higher.
b. A fiber optic ring backbone, even if feasible in an IEEE
802.3 environment, is not as generic as a 50 Ohm baseband coaxial
environment. In a 50 Ohm baseband coaxial environment, I can
procure interface devices from several vendors at least. I fear
that a Fibercom Corporation implementation of IEEE 802.3 in a
fiber ring environment would lock us into procurement of
gateway/router and interface devices from only one firm,
"Fibercom". I wouldn't like to see us being maneuvered into
procurement of a "vendor proprietary" solution with little
flexibility regarding choice of gateway/routers and vendor
interface devices.
c. Implementation of IEEE 802.3 in a ring topology seems to
me would require something akin to the opposite of IEEE 802.4
Token Bus. In other words, implement me a bus protocol in a ring
topology. Not that IEEE 802.4 doesn't make sense, but it appears
that essentially the MAP folks have implemented a ring protocol
in a bus environment. I am sure they have their reasons, the MAP
community isn't stupid. But what about implementing IEEE 802.3
in a ring topology? Is this for real? I confess, I am ignorant.
d. I would think that in a ring topology, be it fiber or
coaxial cable, by definition requires an order of magnitude
greater configuration management overhead, in the sense that you
can't just pick a 2 meter mark on an Ethernet cable and screw in
a tap. Adding or deleting nodes in a ring topology seems to me
requires very involved physical and software configuration
management processes. Whereas management of a baseband Ethernet
appears relatively simple, requiring no extraordinary services or
talent. Hmmm, maybe somebody is setting themselves up for a LAN
configuration management follow on contract.
e. Fault isolation. I had been warned a long time ago about
the problems encountered in fault isolating and repairing a ring
topology. A baseband coaxial cable plant seems so much more
straightforward and maintainable.
f. We are talking about building a fiber optic backbone for
one large building approximately 2 football fields long, and one
football field wide. It is a one story structure. Since we do
not exceed the limits of 500 meter segment and 2.5 kilometer span
restrictions of 50 Ohm baseband coaxial IEEE 802.3, I don't
understand what advantage a fiber optic ring is going to buy us
in terms of distance. Besides, most if not all of the subnets
within this building are going to be other 50 Ohm baseband IEEE
802.3 gatewayed/routed/bridged into this backbone.
g. To the best of my knowledge, the FDDI (Fiber Optic
Distributed Device Interface) standard has yet to be promulgated,
and is still a draft standard. If we wanted to maintain
interoperbility and conform to standards, I would think we should
be picking FDDI as a target for a fiber optic backbone. Right
now, I am not even sure of the correct physical specifications
for an FDDI cable. Is this cast in concrete? Personally, I
would avoid entertaining fiber optic backbones until such time
that the FDDI standard is promulgated.
5. We have precious little networking experience, but this fiber
optic ring solution seems to have gained a lot of popularity. I
fear that people, in their zeal to become state of the art, are
throwing caution, reliability, interoperability, and the KISS
principle to the wind.
6. Can anybody comment and shed light on this situation? Does
anybody have first hand experience with such a TCP/IP Fibercom
Corporation IEEE 802.3 Ethernet fiber optic ring? My apologies
for any statements above which may appear patently stupid from a
technical perspective. Your candid response citing pros/cons and
errors in my logic would be greatly appreciated.
Very Respectfully,
Steve Carr
LCDR, SC, USN
Navy Management Systems Support Office (Code 42A)
Naval Air Station
Norfolk, Virginia 23511-6694
Commercial: (804) 445-2171, 445-1595
AUTOVON: 565-2171, 565-1595
MILNET: carrs@nosc.mil
navmasso42a@nardacva.arpa
-------PETTY@MITVMA.MIT.EDU (Mary Bastian) (03/30/88)
I have never heard of a fiber optic ring for TCP/IP. I agree with you that it sounds a bit bizarre. I do, however, know of another company that might be able to help you with your fiber optic network as well as your FDDI questions. The company is Fibronics, and you can contact a regional office in your area at 703-648-1533. You'd want to talk with Sharon Bittorie. Good luck! Mary
neerma@COD.NOSC.MIL (Merle A. Neer) (03/31/88)
Hi, I must confess I am totally unaware of 802.3 "ring" technology. I have never heard of such a beast, however, I will comment on some of your other concerns to the effect that I agree totally. I am currently involved in evaluation of local net architecture solutions for NOSC Hawaii. In this effort I have been looking at 802.3 vs. 802.5, FDDI, SAFENET, etc. I concur with your comment that you dont want to lock yourself into a vendor proprietary solution; open systems is the name of the game. In the absence of a valid reason to do otherwise(need for process control, real-time applications, etc.) others have already made the observation that Ethernet is the safest, most conservative, and ubiquitous technology here today. The product choices both for hardware and software favor this solution. Tomorrow FDDI and for the military, SAFENET(a variant on FDDI), will probably be the technology of choice. For maximum interoperability, consider that gateways(routers), bridges, etc. are available now for Ethernet. As far as performance goes, we have an Ethernet here(at NOSC) used by some 50 hosts for the usual traffic mix of interactive login, file transfer, transaction processing, tape backups, etc. I have never seen the peak utilization go above 2%. This is typical I believe. It could be that the kind of applications you will run will require higher performance than 802.3 can offer; if this is the case, of course, the correct solution is not Ethernet. Anyway, it will be interesting to hear more about this 802.3 "ring"! Regards, Merle Neer NOSC neerma at nosc.mil 619-553-4135 -------
nelson@sun.soe.clarkson.edu (Russ Nelson) (04/01/88)
In article <8803292224.AA13808@trout.nosc.mil> carrs@TROUT.NOSC.MIL (Stephen M. Carr) writes: > c. Implementation of IEEE 802.3 in a ring topology seems to >me would require something akin to the opposite of IEEE 802.4 >Token Bus. In other words, implement me a bus protocol in a ring >topology. Not that IEEE 802.4 doesn't make sense, but it appears >that essentially the MAP folks have implemented a ring protocol >in a bus environment. I am sure they have their reasons, the MAP >community isn't stupid. But what about implementing IEEE 802.3 >in a ring topology? Is this for real? I confess, I am ignorant. Well, I did a Master's thesis on a network that was similar to what you're talking about. Physically and electrically the network was a token passing ring. Logically, it was a bus, in the sense that there was only one transmitter active at a time, and all information was received by all nodes at the same time. So, I do think that IEEE 802.3 could be implemented in a ring technology, albeit without token passing. As an aside, the header that I used was remarkably similar to the Ethernet header, but with one octet addresses. I guess round wheels get reinvented all the time. -russ -- -russ AT&T: (315)268-6591 BITNET: NELSON@CLUTX Internet: nelson@clutx.clarkson.edu GEnie: BH01 Compu$erve: 70441,205
smith@cos.com (Steve Smith) (04/05/88)
In article <8803292224.AA13808@trout.nosc.mil> carrs@TROUT.NOSC.MIL (Stephen M. Carr) writes: >[Asks for info on an "802.3 fiber optic ring network" from Fibercom] I am not familiar with Fibercom or their products. However, there is currently an effort within the IEEE 802.3 committee to come up with a standard fiber optic star network (FOSTAR). This network would be compatible with all other 10 MBit/sec 802.3 networks at the AUI level - all you'd need to change to go to fiber optics would be the MAU (or tap tranceiver, for you non 802.3 types :-) Technically, the problem with using fiber optics in a bus topology is that (unless you're No Such Agency :-) there is no such thing as a high impedance tap. As a result, you either have to have a true ring topology (as in FDDI) or have some kind of gizmo (called a hub) in the middle of your network to handle signal distribution. The hub can be either passive or active. Both ways have their adherents. Passive systems are cheaper and don't need power for the hub. Active systems can cover longer distances, handle more nodes, and are easier to balance. Unfortunately, the two types are not compatible, except through repeaters. Also, some of the discussions within the FOSTAR committee seem to be approaching the holy war state, and it looks like we may end up with two separate, incompatible standards (active hub and passive hub). Since you want to use this stuff as a backbone, you might investigate a new standard, IEEE 802.3, Section 9.9, December 1987, "Vendor Independant Fiber Optic Inter Repeater Link". It might be closer to what you need. (Anything that looks like an Official Opinion probably isn't.) -- -- Steve (smith@cos.com) ({uunet sundc decuac hqda-ai hadron}!cos!smith) "Truth is stranger than fiction because fiction has to make sense."