Hello Dave, Thread,
Okay, we've had enough time to rest and cogitate over this thing. What do you say we pick it up where we left off and design a delivery platform? I know, Dave, that you suggested that too in the upstream message. Okay, I'm ready. But before we begin say hello to some new acquaintances of ours over in Raging Bull CLWK thread. You may not know them yet, but they know you:
ragingbull.com
ragingbull.com
As a refresher, i.e., in order to get back into the groove, I've taken some time to post a rather long message in the SI Gilder Thread, ostensibly to defend a cutesie that I'd made earlier concerning one of George's observations about the number of lambdas and the bit rates that could be achieved through wdm. But it was equally a warm up for me to get back into the FTTH mode.
I'll copy it below in a moment. In it I make some points, however, which I'd like to highlight here, because I think these would be good to keep in mind as we proceed:
1 - Incumbent Fiber (actually, the lambdas which reside within the incumbent providers' fibers) will be up for grabs in much the same way as telco twisted pairs and DLCs are today, a la copper spectrum piggybacking by the DslLECs.
2 - Related to 1 above, it will be possible to have multiple service providers, each supporting one or more lambdas to the residential user, via wdm. Fiber sharing could take place in the distribution mesh (backbone), or on the residential fiber "drop," or on both segments. When spacing and slicing (that's slicing, with an "l", no 'p') techniques mature, there will be plenty of lambdas to go around. Getting them might be a series of long fights, but they'll be there, nonetheless.
(As an aside, we auction radio spectrum, but we don't think in terms of doing the same with fiber spectrum when it is carried over monopoly held rights of way. Something to think about, I suppose. What do you think?)
3 - There is no true FTTH reference model that has been defined by any standards body today. FTTH models which have been proposed (and even the ones which purportedly have been installed) in the past have borrowed from the tree and branch topology norm, or they've employed protocols which are deeply entrenched in the fabrics of the telcos. That's not newly enlightened telcos, its the other part which is still clinging on to their vertical integration model.Such hierarchical models need not survive due to the altered nature of the medium when pure fiber is introduced end to end, save for a minimal number of wdm and optical switching nodes. Here we'd need to discuss whether transparency is required throughout the mesh, and/or how opaqueness plays into matters end to end.
The FTTH model going forward will take advantage of fiber's transparency and distance neutralizing characteristics, enabling polymorphic topologies, logically, if not also physically. We should NOT be thinking of a blending of PONs [which derive from the Bellcore-ITU Full Service Area Network Reference Model (FSAN-RM) using SONET and ATM]; HFCs; and deeper forms of fiber builds such as FTTN and FTTC.
None of the foregoing at the present time are capable of lambda transport into the home because they lack an optical orientation from beginning to end, except to port analog stuff in its native form along a mis-span section where it is usually r-f modulated. But more to the point, none of these earlier modes will support lambdas because they each use coax and/or twisted pair to get into the home.
4 - I momentarily broached the issue of hybridized fiber/wireless, which is a component which I think should be included in the base design from the outset. In this same vein, perhaps we should also examine satellite (community internet access) solutions icw line of sight relays to user clusters, for extreme cases of boon dock.
5 - and this I didn't post earlier, but we should decide if we are to be strictly facilities-based providers of transport services (including ISP, transit, peering and end user delivery of multiple media forms), or whether it makes sense to get into programming, POTS, ISP functions (web, email, news, dns, etc.).
In other words, remain horizontally distinct from those who would use the services we provide in order to provide content themselves, or do "we" provide the video content ourselves, along with POTS, LD, cellular, 'Net access, etc. in a vertically integrated manner? (like the rest of those fools are doing, who consequently can't react to change, if it meant saving their lives?).
I posted the following passage on the Gilder thread in response to a query by Dan B.:
Message 13764193
-------
To: Dan B. who wrote (4438)
From: Frank A. Coluccio Wednesday, May 24, 2000 1:04 AM ET
Reply # of 4439
Hello Dan,
Of course, I was being (although, only mildly) facetious in making that remark about the 864,000 lambdas. I should learn to be more liberal in the way I use emoticons when I'm writing on the road ;-)
The intent behind by statement, aside from some collegial chiding, was this: Although you will some day be able to fit as many as 864,000 or more lambdas into a six-gross (864-fiber) cable whose outside diameter is less than 1.5 inches, assuming 1,000 lambdas per strand, managing this many fibers at a single network element's interfaces would be really ugly, to put it into IETF parlance.
Furthermore, implying that a stream valued on the order of 6 or more Petabits might in some way, in the foreseeable future, be useful is also stretching the elastic beyond its snap. This is a common form of profile characterization that the optical box guys like to promote. They make claims of 2.4 Tb/s this, and 4.8 Tb/s that. But in reality, the largest flows that can be harnessed within those aggregates are still on the order of 10 Gb/s. Today's routers and switches strain when they are asked to do 10 Gigabits of read and forwarding. Hopefully they will reach a terabit some day. Peta? Not until that guy comes along with the magic fix that allows us to read, filter and forward optical headers on the fly, while respecting policy, security and so on. For raw fire power distributed over several fibers, though, yes, multiple.. perhaps up to the tens of, terabits. Give me a million strands, and I can then multiply this by a million... say whut?
Yes, the aggregate amount of bandwidth derived from the mass of 864 strands is impressive, nonetheless. And I should feel this way, I'd better think that it makes sense, because I've written the same type of account, using the same rationales and derivations over on the Last Mile, MRVC and elsewhere here in SI and over on Compuserve, several years ago, if I'm not mistaken.
But I acknowledge that we're not here to talk so much about the limitations of the present as we are about the possibilities of the future, so let's move on.
------
At some point you will begin to hear talk (it's already an issue) about proprietary lambda spacing schemes (grids) on units of different manufacture. If done without proper thought this will (already has begun to) adversely impact interoperability, and ultimately stifle communications between virtual pipes (lambdas) from different vendors' wares at the physical and interfaces and lambda transmission and filtering levels. Stated another way, it will be increasingly difficult to link these elements of different manufacture at their native optical interfaces, unless they all use identical [spacing] grids. That's one way.
The other way is to employ dynamic sensing and accommodation: In the absence of such static standards as I've alluded to above, there would need to be some means of dynamically sensing and then accommodating foreign spacing conventions, on the fly, or under an out-of-channel common control function (like, but not necessarily based on, a form of SS7.
[Don't cough, those of you out there with a more discerning taste for religious warfare... there are talks underway as we type among some pretty smart internet architects, some from the Internet Architecture Board, or IAB, albeit, who are met with less than glowing appeal, that suggests that there may be a need for such a control function for surveillance, signaling and control overlaying for the Internet at some time, perhaps soon. This, in order to orchestrate the monster flows we're now beginning to see, and the even larger ones to come at the link layer, which would call for more than a mere modicum of standardization.. a set of standards which doesn't exist today.
-------
Having stated that, let's get to your point about PONS, and other as-yet-unidentified and un-named cousins relatives of PONs in the last mile that would constitute fiber to the node/curb/garae/home.
Yes, with time I certainly see a point when hundreds of millions, billions of lambdas prevail amongst multiple local regional and national/international/intergalactic systems (hey, we may as well use TeraBeam to get us "out there" while we're at it, no?), in topologies which, likewise, have yet to be specified or even conceived.
How many homes in the USA alone are now passed by MSO- and Telco- fiber-enhanced builds? It's certainly conceivable to me that each residential structure, on average, receives its own lambda sometime over the course of the next ten years, maybe two or twelve of them, when the fibersphere finally arrives. No?
As we approach a purer form of optical, in successive degrees, we move away from a dependency on the familiar hierarchical, tree-and-branch topologies, while moving closer to a fully meshed optical and wireless network framework. Like the power grid. This ability to break away from branch and tree (which really isn't as efficient anymore as it once was), of course, borrows from one of fiber's greatest advantages, aside from its high capacity, and that is its ability to neutralize distances while permitting the backhauling and re-routing on demand of payloads in ways which today appear to be entirely non-intuitive. These capabilities will soon be technologically and economically doable.
Let's examine a couple of future last mile fiber, or PON, possibilities. But first I think it's good if we differentiate between today's PONs and "future" PONs, which I'd equate to being the last sections of builds in, again, still-undefined, FTTH-like architectures.
Today's PONs are almost entirely SONET, GR303 and ATM dependent. SONET on the backbone, GR303 in the central office and field node feature standards, and ATM transport to the residence. Also, today's PONS still adhere to black coaxial and/or twisted pair to get across the last thousand feet from the pole, or wherever, to your shack.
Future PONs will be extended to termination points via fiber directly into the home, or via high-cap wireless, as a function of terrain, pop densities, climate, business model, service mix, and so on, dictate.
If we're talking about PONs in the context of their delivery to the home in a future FTTH design, then they may be configured in a variety of ways, which, without agreeing to a specific scheme would be useless for us to discuss. But assuming full FTTH here, we can wing it, and say that a "sole provider" may support several thousand lambdas to a neighborhood or small town. Those lambdas have been derived from just a few backbone strands through wdm at field nodes, and satisfy the needs of everyone in the serving region.
In this case, a single lambda from a single provider might support a single home, or four or five homes through some wavelength sharing technique, or through some form of broadcast and select protocol... who knows, somehow it will work.
Alternately, in a shared fiber arrangement (where more than one provider rides over the same strand, like today's DSL CLECs who are piggybacking on the ILECs' POTS lines) one or two lambdas of narrower widths might serve an individual home for different types of services, from different providers. Or, for different classes and grades of service: entertainment (TV), casual 'Net access, telecommuting, POTS, etc., from the same solitary (or two different) provider(s).
I think it's fair to say that we will be seeing this form of territorial interloping, or facility piggybacking, soon, which will be characterized much the same way as today's CLECs have demonstrated, riding over the basic POTS line in the upper frequency region for DSL services. ISPs and others will be demanding the right to lease individual (or multiple) lambdas on the incumbent cablecos' and telcos' backbone (and future residential!) fibers.
In sum, if one metro-grade "six-grosser" cable which is less than 1.5 inch in diameter is capable of 864,000 lambdas, then how many lambdas will exist nationwide, ultimatlely? A billion is not an exaggeration. That's a lot of light sources, wouldn't you say? VCSELs will make it a breeze.
I already have reservations for sixty-four. Do I hear 128?
FACt, not FACetious (except for the TeraBeam part ;-) |
