Hiram, DT, MM, All,
Taking the past couple of days of posts into consideration, I'd like to make some observations, if I may.
>Unless there is a new supported IP protocol over SONET which this Gilder chap seems to be implying. Then it would be okay. But his statement of "physical layer infrastructure becoming more abundant" is gobbeldygook nonsense since the same supported framing formats, but just more SONET out there, makes no difference. Its the part about being more capable that I am not sure of.<
George Gilder certainly doesn't need me to stand in for him in his defense, but I'll say this: Some of the quotes I've read from him here in the past couple of days are highly generalized ones, and he probably just didn't see the need to get very granular in those passages when they were written, given the general audiences he reaches sometimes.
In any event, on the surface, supporting IP over SONET may not appear to be that difficult to achieve. Although, there was never any convergence or mapping function designed between the 1st and 3rd layers, specifically, of the OSI stack for this feature at the time that SONET was being specked, during the Eighties.
The problem with vendors plowing forward on their own with Packet over Sonet (POS) solutions is that there are no agreed-on standards at this time to facilitate it on a widespread basis. By this I mean that there are no standard network management and configuration tools to satisfy inter-provider (inter-carrier) handoffs, no interoperability guidelines, no alarming and surveillance conventions, no traffic grooming techniques specified, etc. etc.
But we will begin to see changes in this, gradually, as will be the case also with Packet-over-Fiber and Packet-over-lambda protocols. Cienna, Cisco, Lucent, Nortel, Pirelli and others of less stature [at this time] are joining ranks with the IETF and ITU folks and other industry standards setting bodies in seeking solutions to these issues.
Some of the advantages of POS are obvious. For example, if one chooses to place IP over SONET directly, then the IP payload size is increased considerably, with minimal wasted bandwidth going to useless overhead, because there are no Layer 2 protocol overhead bytes taking up valuable space in the frame. Bandwidth efficiencies jump from the 60%-70% range when ATM is used [under IP], to over 90% when IP is used alone.
It's not exactly that simple a comparison, since other dynamics get in the way at times, but that is a good characterization of the difference between the two modes.
What we've just examined is packet over SONET. When it's packet over "Fiber," directly, now eliminating the SONET overhead as well, you cut the overhead in half, or by more, again.
But this begs the question of POS's true utility, in a truly "networked" sense. It can do very nicely in a Point-to-Point situation, such as a high capacity route between Chicago and St Louis, at this time, where the end points are very large routers that feed off to conventional tail circuits. And it will even do well in large meshed networks where there is consistency of pipe sizing throughout.
But how does it fair when placed in a situation that demands "adding and dropping" of smaller, sometimes very small, streams along the way? Not too well, due to the still incompatible nature of the facilities along the way. At least not at this stage, since these special routers [which will be very scarce during the initial stages of POS] are needed at every ingress and egress point on the network of interest.
To the best of my knowledge this will not be a classical case of upgrading the myriad routers now in existence, like we do every time a new RFC is ratified. This will be different. POS operating at super OC-48 speeds will be a new breed of animal, the likes of which will have to be added in every location where its benefits are to be realized. We're talking about a multi-billion dollar make over of existing networks. The Internet[works], to be more precise. No upgrades of older 2500s and 4500's, in other words.
When you begin to digest the implications of that last sentence, then you can begin to appreciate the principle that states that embedded investments must be preserved on the Internet, as well as on the PSTN. Up until now, upgrades on the internet were more often than not, software patches. Not any more... at least not where higher throughputs through advanced optics are concerned.
--Re: ATM and SONET
ATM is not a standard belonging to SONET. It is an application that rides atop SONET, and even though it fits like a glove (that's the way it was designed after excruciatingly long battles between the Europeans and the Americans), it has its own headers, its own end to end protocol, and routing mechanisms.
It is the combination of ATM rules and protocols along with SONET and AIN that comprise the three-pronged strategy that makes up what is known as that gobbeldygook [to borrow a term] Broadband ISDN, or B-ISDN. B-ISDN is not an individual protocol, rather it is an architectural framework.
===
On the matter of OC containers, Hiram, your post was very funny.
> I could be wrong on this, but isn't OC based on TDM? I heard from someone,can't remember whom,that OC was eventually going to be dropped,as its time is just about over,well time is the operant word. Is there any truth to this,TDM is being replaced by WDM,so what replaces OC,OHNO?<
TDM-derived Optical Carriers (OC's) will eventually be dropped at the super-OC48 levels for the most part, at first. There may still continue to be higher transmission rates called OC-xyz, but they will not be TDM derived from a multiplexer.
There will continue to be contention between DWDM and OC192 rates for the next couple of years, and perhaps OC192 will settle in on its own. I don't know how stubborn the traditionals will be on this one. Some are already swearing by it, and others wont go near it with a ten foot pole.
The "OC" designated containers, nonetheless, will still be around at the "OC48 and under" levels until my now thirteen-year-old daughter graduates from College.
Most of what we've been discussing in these past ten or so posts concerning fiber and lambda usurping the roles of SONET will be at the super- ultra- high- capacity levels, first, and eventually (and very gradually) at the lower levels such as OC12.
These are not rules, rather my opinions. And what holds true for the deep core may not hold true on a time-synchronized basis for the distribution plant to the home.
Eventually, when PONs begin to proliferate in the network's edge, and when commodity dynamics kick in, we'll see this to the residence as well. I hope to see that day when it comes.
Regards, Frank C. |
