Paul, prior to this post, my views were pretty much pointed to general directions and the provisioning of enterprise network elements and protocols in an integrated, larger setting. I still hold to those views. However, after reading your posts more carefully, I would allow that there are a set of applications such as you have alluded to which remain as niche areas at this time that need special attention. I attempted to post this last night, but the SI server was down. Let's try it again:
-----------
Paul, now I see the context of your argument, and from a practical perspective I agree with you, with some qualifications.
"I do not believe that the Internet is capable of delivering these services in its present state."
For those applications which demand consistent, uninterrupted delivery of isochronous services (such as real time video conferencing and high quality voice communications), IP has clearly not optimized itself yet, at least not on the public Internet (and there is a great deal of work still to be done on private IP backbones as well), and these spaces would be best served by more deterministic methods of delivery.
There are communities of users, however, who will swear by desktop IP Video. Granted, many of these are too cash-strapped or underfinanced for anything else. Then again, there are commercial grade services being supplied by conferencing providers who are charged with delivering high-Q video content, who could never meet their service level agreements (SLAs) if they were forced to depend on anything but dedicated lines, or ISDN, or ATM.
As with most assessments and final decisions, the choice of protocol (and the platform to be used) will depend on the venues involved, geography, and the tolerance for price-performance tradeoffs.
Right now, despite the oceans of optical capacity being laid, the carriers and SP's have still not harnessed that capacity in ways which would allow for sufficient and consistent "head room" in the pipe that would make IP acceptable for real time services, and at the same time allow for ubiquitous connectivity.
Therefore, using ISDN, or ATM in a quasi-private network [on-demand clustering] manner actually does work out better for situations like corporate video conferencing, distance learning, and live video broadcasts, as you have suggested.
The spread between these two scenarios (i.e., between the packet loss and jitter of IP on the one hand, and the clarity and steadiness of PLs, ISDN, and to a sometimes lesser extent, certain grades of ATM, on the other), however, will narrow and eventually come to blur as greater amounts of bandwidth are put to use more inexpensively, and as IETF ATM-like protocols (such as label switching) are refined, ratified and deployed over IP network platforms.
As we speak, and this is where your argument makes the most sense, if I had to put in a fifty-school distance learning network, say, which depended on video conferencing, I would be hard pressed, still, to use IP for all but the least demanding sets of expectations. I probably would use either ISDN PRIs where they made sense, or BRI ISDN, and the ATM based system technology which you've been referring to would also make sense.
I could also see putting in an IP based video solution at this stage for certain levels of distance learning programs, especially those which are related to internetworking and computer sciences, and where viewing of fine visual detail is not crucial. These IP solutions, complete with whiteboarding at the desk level are becoming more prevalent each day... but they do not satisfy many of the more demanding corporate solution requirements, especially for large enterprise [including many large universities'] video conferencing needs, where most of the dollars are being spent on big ticket systems... yet.
What would I use if the distance learning program were directed to the hearing impaired, i.e., those who depend on extreme crispness of image and superior motion tracking?
I would use ISDN (probably at least six bonded BRIs worth) or some ATM derivative if one were available (such as those which have been installed on a number of the larger municipal I-net backbones), due to the needs that those students possess for reading subtle changes in hand and facial dynamics, especially in reading the syntax of American Sign Language.
Most IP Video platforms would not be able to consistently catch those types of movements today over the public Internet, under normal conditions. Someday, however, and IMO, they will.
I've intentionally oversimplified this to a great extent in order to keep it brief, since all of the parameters which must be considered and the actual level of tradeoff analyses that must be performed are quite extensive. Netting it out, there are still some very large opportunities for circuit switched technologies (and even straight dedicated private lines, if the tradeoffs were in favor of them) to support isochronous payload deliveries, especially when Service Level Agreements stipulate stringent performance criteria.
Having said that, however, I should also note again that these networks could easily become stranded for all intents and purposes other than for those purposes for which they have been specifically installed. And again, sure... you can always integrate one-off subnetworks with larger enterprise networks such as extra- and intra- nets, but very often these are not done with optimal costing, and very often involve architectural sacrifices as well. As is the case any time you add additional vendors' wares to an already over mixed-and-matched environment.
Ubiquitous reach by other applications and end points within those same enterprises who use them will increasingly depend on the IP model for both local distribution and wide area transport. All, of course, IMO.
Regards, Frank Coluccio |
