Frank:
Re: Metcalfe's position on the new ethernet.
I suspect that since his original patent (1975, now expired- see below) didn't cover port
switching, he would be saluting other people's ideas. The original ethernet was a shared
passive network topology. Also recall that Metcalfe famously predicted that the internet
would "go spectacularly supernova and in 1996 catastrophically collapse." I'm still getting into
SI and getting emails, so I'm not so sure he understands today's networks all that well!
However, ignoring the poetry of the word ethernet, I still am wondering about how the last
mile network will work. Playing around the web I found some historical notes on the
ethernet. This, actually, was not only fun, but may be relevant to how the last mile will work.
Since we are anxiously awaiting a market solution for last mile residential broadband, it may
be helpful to go back to first principles.
A classic 1976 paper by Metcalf and Boggs ( acm.org ) actually
still provides relevant concerns about topology.
We cannot afford the redundant connections and dynamic routing of store-and-forward
packet switching to assure reliable communication, so we choose to achieve reliability through
simplicity. We choose to make the shared communication facility passive so that the failure of
an active element will tend to affect the communications of only a single station. The layout and
changing needs of office and laboratory buildings leads us to pick a network topology with the
potential for convenient incremental extension and reconfiguration with minimal service
disruption.
OK, ethernet must be shared for simplicity and salability. The problem is that the net is still a
single point of failure, and a service upgrade potentially can impact the entire
local/neighborhood network.
To me, PON and HFC both suffer from the fact that they are shared; only a 'home run' point
to point network would achieve the reliability and availability of the copper phone line. But
I'm not sure this is what the FTTH or 10GbE or PON or whatever are supporting. It sure
makes for a complex business case.
My guess is that some hybrid system, with shared neighborhood access, must be accepted for
all solutions to make a reasonable business case. This drastically reduces the apparent
advantage of 10GbE bandwidth, though not fatally. The big 10GbE advantage, as I've noted,
is freedom, more than bandwidth.
The topology of the Ethernet is that of an unrooted tree. It is a tree so that the Ether can
branch at the entrance to a building's corridor, yet avoid multipath interference. There must be
only one path through the Ether between any source and destination; if more than one path
were to exist, a transmission would interfere with itself, repeatedly arriving at its intended
destination having traveled by paths of different length. The Ether is unrooted because it can
be extended from any of its points in any direction: Any station wishing to join an Ethernet taps
into the Ether at the nearest convenient point.
This salability is the big advantage of shared systems. Unless you design a network for
ubiquitous access, you will probably rollout residences on a almost random basis- you don't
know where you must tap the network until you get the work order. HFC and ethernet do
this well.
Looking at the relationship of interconnection and control, we see that Ethernet is the dual
of a star network. Rather than distributed interconnection through many separate links and
central control in a switching node, as in a star network, the Ethernet has central
interconnection through the-Ether and distributed control among its stations.
You can see why ethernet was originally 'not a star' network- to me making Ethernet use
'many separate links and central control in a switching node' means, well, it isn't Ethernet.
However, the IEEE folks have successfully redefined the issue, so ethernet doesn't mean what
it meant in 1976.
Ironically, the issue of distributed control vs. central control moves up to the application level
now. We would all like, in the coming decade, more local, decentralized control over our
applications. Even if the 10GbE has more links and central switching at the network layers, IP
and various modifying protocols will still support a distributed control at the application level.
Unlike an Aloha Network, which is a star network with an outgoing broadcast channel
and an incoming multi-access channel, an Ethernet supports many-to-many communication
with a single broadcast multiaccess channel. "
This is an important point. If you do have home run links, the switching element must support
a sophisticated multicast, for, say, broadcast TV or radio.
Finally, I can't agree with your point:
Which, when you look at what a switched port is, is nothing radically different in principle,
although the newer ports are backed up with a lot more
intelligence and speed, with some of them looking at the upper layers for instructions, to
boot..."
It is clear from the above quotes from the 1976 papers, that the authors believed there as a
radical difference between shared access and point to point links with switching. I believe this
patented difference made Meltcale famous and a multi-millionaire from the royalties.
The Original ethernet patent for those with far too much time on their hands, may be found at:
164.195.100.11 |
