re: passive optical networks [PONs], and references to optical b/w management
Doc, good find.. thanks.
It's uncanny that you should post that article referencing Quantum Bridge today. I just sent out a reply last night to a private mailing list in which I used QB as an example of one firm's approach to doing what they characterize as dynamic allocation of optical bandwidth in metro and residential settings. I'll post a couple of my contributions to that discussion at the end of this message, in a moment.
While much of what has been written here and on the ATHM and LM threads about 10GbTTH, Rabit, etc. has been more akin to prospecting and some lightly spirited banter (at least in part), I now find myself at some interesting crossroads where it's time to 'walk the walk' on at least two different fronts.
In each case that we now face, the overriding theory is that optical can do it all for less. But let me tell you that the obstacles and considerations which must be attended to in order to cross the chasm to optical from legacy for a large enterprise are not trivial ones.
I liken the exercise of transforming a large enterprise network to optical to human organ transplants thirty years ago. In optical networking it may not take so long, however, because we now have something in our favor, something that Dr. Michael DeBakey would never have dreamt about, unless he was devoutly following McLuhan and Toffler at that time. And even Toffler, at least, has been stunned today by these effects. And that 'something' is comprised, even accelerated, by all of the drivers which encompass what we now call "Internet Time."
Internet Time dynamics, even in this space, will have the uncanny effect of shortening, rather than lengthening, schedules of newer and more sophisticated platform deployments than the older ones which they are intended to supplant.
Despite this phenomenon, legacy thinking and FUD among decision makers, along with proper levels of care and diligence which demand that they move very gingerly into the unknown, remain the biggest obstacles to crossing new chasms, and these seem to defy any kind of 'time,' by any metric.
But we shall persevere and move forward, because some bleading edge lunatics like (note: I said 'like') yours truly. Hey, I'm not necessarily volunteering here, yet, at least not until I'm assured that my family is taken care of, first. Those lunatics will prove the concepts in first (some will be shot, to be sure), and then others will follow, moo-ing their ways to "look what I just invented here."
-----
The following are two replies to the mailing list, which I alluded to above. They are in response to a cited assertion which was made by the CEO of JDSU, whose opinion was alleged to be that metro level bandwidth allocation schemes which would be used to dynamically allocate resources do not exist yet:
=========
[All,]
There are numerous optical bandwidth approaches being developed in
wide area network [elements]. But your point(s) seem to be directed at the
metro venue, which I'd also take the liberty of meaning residential, eventually.
So, I'll focus on the latter (metro and future residential) for the purpopses of this
reply.
While there are photonic technologies under development which do
address dynamic bandwidth (linewidth) allocations over various base
element designs, some using photonic chips employing familiar planar-like
principles, I'd be inclined to agree with the CEO of JDSU. At the present
time none is (widely) available in commercial form. [But my first inclination
which I'll expand on in a moment, below, is: So What? Well, "eventually," so
what, in any event.]
There are companies whose products can dynamically "slice" wavelengths,
or so they claim. But these are not "pure" optical solutions as much as they
are those which are still tied into other more established frameworks, like
SONET, ATM and passive optical networks to residential neighborhoods
and commercial buildings. See Quantum Bridge, for one:
quantumbridge.com
Here's a passage from QB's product page which explains what I just
stated:
==========begin snip: (go to site for diagrams)
""QUANTUM BRIDGE ARCHITECTURE AND DYNAMIC WAVELENGTH
SLICINGTM
The Quantum Bridge Optical Access System is comprised of two network
elements, the QB5000 Optical Access Switch (OAS), located in the central
office, service node or head-end, and the QB100 Intelligent Optical Terminal
(IOT), located at the customer premises. These two components are networked
with fiber and low cost, passive optical components that reside at each fiber
junction. In a typical deployment, the system supports hundreds of IOTs, each
of which can be provisioned to provide on-demand service bandwidths ranging
from 1Mbps to multiple dedicated wavelengths.
Fundamental to the Quantum Bridge Optical Access System is its Dynamic
Wavelength Slicing (DWS) protocol. Extending the benefits of Dense Wave
Division Multiplexing (DWDM) - which made it possible for a single fiber to
transmit multiple wavelengths of traffic - DWS is the key innovation that enables
the Quantum Bridge architecture to "slice" or fan out each wavelength, or
lambda, to deliver broadband services to multiple business customers. This
gives service providers a cost-effective way of leveraging existing fiber
infrastructures to rapidly deploy service to many business customers thereby
maximizing their market share and profitability.
<delete>
==============end snip
Actually, there are two still prevailing reasons AT LEAST why they must
still do conservation that I can discern. One is that line widths are still too
broad (they are not granular enough, and are limited to " n " lambdas per
strand at this time, so they must be used prudently. Secondly, the
upper-Layer 1 and Layer 2 two protocols which they support (Sonet, ATM,
Frame Relay, ISDN, etc.) are themselves limited in supply (real or
imagined, but they are expensive either way) once the optical elements
meet the legacy SONET and other networks which are, by definition, more
costly and cumbersome to manage than pure optical, in an increasing
number of metrics. Here we have the tail still wagging the dog, so to speak,
in several different ways.
---
But a more pertinent idea occurs to me when I contemplate such a
capability as minimizing and then maximizing the amount of bandwidth over
lambdas in a metro venue. And that is, it almost seems antithetical to be
promoting such a capability when the whole idea behind metro fiber is to
avail itself to virtually unlimited b/w (of far more than can easily be used by
current applications) to begin with! (and I seldom use exclamation points).
The direction which dark fiber should be taking us to should be more in line
with the fiberspheric environment that GG spoke of in his 1992 ASAP
article, taking advantage of the tens of THz of bandwidth that each strand
has to offer. When we start talking about "conserving" it in the metro area
through legacy concepts which were used to optimize TDM and metallic
alternatives, such as the dynamic allocation (of wavelengths), we simply
revert back to those minimalist attitudes which were formed over the years,
when bandwidth was more legitimately scarce (or retarded by the
incumbents, in order to control supply).
---
I recently wrote a brief article in SI positing the creation of an all IP
10/100Gb/s residential architecture. Here, I too proposed such a set of
capabilities. I called it dynamic wavelength tuning, or tunable wavelengths.
In one part I proposed a 10GbTTH (to the home) design in order to
maximize the number of homes passed by such a system while using
dynamic resource allocation. This of course I was proposing as an
alternative to the existing analog HFC fiber-coax framework, such as the
MSOs have been upgrading to.
In this case, however, I was proposing to not only replace the 1 GHz per
home ANALOG FR which is already achievable over RF and coax, but to
allow multi-gigabit rates to data end points in each residence, as well. And
of course, such a model would support video (all forms: program grade
NTSC, HDTV and Videoconferencing) as well. One setback that this
immediately presented, however, as any optical purist would be quick to
point out, is that as soon as you introduce Ethernet you are violating the
principles of transparency for the contamination that electronics represents,
creating an opaque environment, or only semi-transparent one, at best.
Which is an issue which has in my mind been relegated to the status of
permanent conundrum, ever since reading the fiber sphere article eight
years ago.
Dynamic "tuning" I felt was essential, in order to keep as high a population
of users as could be optimally done in such a community on the same set
of physical elements for cost containment purposes.
But after some additional review of the matter I saw that even I was lured back
into a minimalist view which I alluded to, above.
After writing the above mentioned piece I later came to learn that a very
similar architecture was being proposed as a pilot in Canada, under the
Canarie umbrella. I'll search for the link and post it here if I can track it
down. Ah, here it is (late edit: this is the point where I hit the wrong key earlier ;-)
It's taken from David Isenberg's column in America's Network Magazine.
americasnetwork.com
<delete>
Message 12507995
---
All comments and corrections welcome.
Best Regards, Frank
ps - in case you get bored over this coming weekend and would like
some light reading to keep you going, try this. I was just carpet bombed
with it while typing the above memo. It's actually a very telling list of names
and companies who are neck deep at this time in 10Gb Ethernet and
optical adaptations of same, in case you are interested. Enjoy.
grouper.ieee.org
[late edit: the preceding came to me via a prominent member of this thread whose level of desire concerning annonymity I'm currently not sure of]
======
Group,
One of the main points that I attempted to convey in my earlier
message was the distinction between (i) managing the supply of
optical bandwidth/spectrum (which is purely the province of photonics),
and that of (ii) managing the protocols which ride atop the photonic
layer, which take place at the higher layers in the OSI-RM stack...
.. i.e., the ISO's Open Systems Interconnection Reference Model.
JDSU's CEO was referring to the latter, in the sense that a "spigot"
effect is still not widely available in the metro, yet, at the photonic layer..
Frank
|
