bruce@ssc-vax.UUCP (02/26/87)
I manage a broadband LAN here at Boeing and deal with ethernet and some other
LANs as well. Most of the original flack about why baseband is a better
system than broadband (or vice versa) has died down in recent years due to
the overwhelming acceptance of baseband ethernet as the LAN of choice. Lets
face it, if you want an ethernet interface for any misbegotten unloved piece
of digital gizmo, you can probably get it. Try that with broadband!
Having said that, I still like broadband. But I agree with the poster who said
that you are better off with a broadband guy on your staff if you run a
broadband net. But this certainly doesnt have to be a "trained, experienced
professional". My background included RF communications some 15 years back,
so I was no stranger to Spectrum Analyzers and sweep generators. I find
broadband a snap. It can be learned by anyone who takes the time to put the
same sort of study into broadband that he would into learning a new programming
language or a new logic family.
But back to the subject. I feel that baseband and broadband have definite
niches and should not compete, but rather coexist. Each does certain things
superbly. Rarely, however, is either sufficient as the sole networking
solution.
As a practical example, we have both RF broadband and baseband ethernet trunks
serving a multi-building campus. There has never been a single failure of the
broadband trunk in 6 years of operation. In contrast the ethernet trunk is
renowned for its ups and downs. Why is the ethernet less reliable? Because
of bad vampire taps and loose transceiver cables. The same wonderfullness that
allows a tap to be simply installed allows a simpleton to bring down a whole
network (or large part thereof). Ethernet distribution technology has a long
way to go to match the maturity and reliability of RF broadband. But on the
other hand, I am severely limited in the computing device interfaces available
for broadband.
So why has there been all the bad press on broadband? For one thing, because
it is not understood by the majority of digital folks and they have not
bothered to learn about it (" +35dbmv ? What the &^%$# does that mean?").
Secondly it is true that broadband can be critical to maintain, BUT (IMPORTANT
PLEASE NOTE--> ) this is primarily in fully loaded systems, covering large
distances, in environmental conditions which would frost your manhood. Ie,
situations that ethernet would be at a loss to cope with. For simple, lightly
loaded systems, broadband is relatively simple and very forgiving (a prime
example being the guy who didnt bother to calculate tap values -- he just
swapped until he got one that worked! ).
If I were starting all over again from scratch I would use broadband for
highly reliable trunking between buildings and (where right of way could be
had) between sites. Within buildings the LAN of choice would be baseband
(probably ethernet). This would be bridged onto the broadband trunk at each
building. The ethernet topology would be designed to minimise the ability of
the unwashed to either short or open the cable. It would further be easily
segmentable either manually or electronically to facilitate maintainability.
The only broadband distribution within buildings would be for video, which
would go only to selected sites such as media rooms, conference rooms,
learning centers, etc.
Please note that the "real world" topology and specific user requirements
will be the final arbiters of what is the best technology to use. LANS are
too complex for simple homilies such as "baseband is better than broadband"
or "token ring is better than CSMA/CD" to be more than half-truths.
Bruce E. Stock
Boeing Aerospace Co.
uw-beaver!ssc-vax!brucekwe@bu-cs.UUCP (03/03/87)
Just some notes on the very fine discussion of baseband vs broadband. In response to a comment by Phil Ngai on the unreliability of broadband components, it has been my experience that broadband modems, amplifiers, headends, etc are no different than any other electronic equipment. Mature products tend to be very reliable. Many LAN cable plant equipment vendors (Sci Atlanta, C-Cor, etc) sell pilot tone generators for amp AGC circuits, redundant amp circuits, and status monitoring equipment. This kind of thing lets net managers sleep better but I don't think that a well-designed and maintained system benefits much from these extras. In response to Anders Andersson's thoughts on using broadband for terminals, it sounds like he is thinking of using point-to-point modems. Please don't! My experience with point-to-point RF modems has been very poor compared to using a broadband LAN (Ungermann-Bass terminal servers, to be precise). We thought that we would save bandwidth on our 5Mbps broadband LAN by putting heavy use point-to-point channels on separate sub-channels. A bad idea, in hindsight. These things need to be tuned when installed and they use fringe channels that are difficult to keep in spec. There is no LAN software to handle the inevitable idiosyncracies of point-to-point equipment that was never designed to have anything other than a hard-wire between. A software-based LAN (broadband or baseband) is far superior to point-to-point subchannels. In response to Jim Bigelow at Tektronics who had problems expanding his broadband and didn't want to hire specialized RF techs, I would say that good design is more important than super-techs, although a really good tech like mine is worth his weight in gold. Good design requires building expansion into the design at the start. This means laying out all trunk runs between buildings at the start, even if some will not be installed initially, and designing in the amps where needed to support future legs. The problem is that most vendors still design broadband networks to minimize components and not to maximize expansion flexibility. I imagine this to be a carry-over from cable TV. Most designers, like Allied Data, do their design with computer systems that use the minimum component algorithms. Using these programs you don't have a prayer of getting a system that is expandable. How do you get a good design? I have two people on staff, and engineer and a tech, that both understand good design and good maintenance. This is the main reason that BU's broadband system has gained a reputation on campus as being very reliable. If you are hiring design and maintenance, give your designer two plans. The first shows where your network could be in a few years and the second shows what you want up on day one. See if he can handle it. Bruce Stock is right on the money when he says that each type of network has its place. BU uses lots of Ethernet in separate buildings and the broadband runs our terminal servers. Broadband gives us the campus coverage and Ethernet suits our DECnet and Unix machine environments. Broadband can link your Ethernets, provide access with broadband terminal servers to buildings with terminals but no Ethernet, and carry video, security and environmental control signals. Broadband LANs are proprietary now. This is probably a major reason broadband will be supplanted by fiber-based FDDI (the 100Mbps non-proprietary token ring standard in the works) when the time comes. You have to commit resources to make broadband work and you have to exploit its strengths (campus coverage) and minimize its weaknesses (exotic technology) by using it as a backbone system. I hope this hasn't been too long-winded or boring. Thanks for reading my musings.-- ---------------------------------------------- Kent W. England Manager, Network & Systems Engineering Boston University Information Technology 111 Cummington Street Boston, MA 02215 kwe@buit1.bu.edu ARPAnet, CSnet kwe@buit BITNET ----------------------------------------------