Hi Eric, thanks for the thought that you put into that reply.
First let me correct some of my sloppy terminology from my prior post -at least to the extent that it is commonly applied- since I was, of course, referring to what is commonly termed "private peering" as you so duly noted late in your own post, and not route peering per se, as I had stated. Also, I omitted stating that ABOV also caters equally to Internet content providers as much as they do to ordinary end user-oriented types of ISPs.
------
Private peering has appeared from time to time as a means to avoid the NAPs and other Internet components, just as you note later in your post, as well, and it has been called various things such as P-NAP, NAP Bypass, and a few other labels which slip my mind right now.
Some earlier attempts at this by regular ISPs (some of whom used ATM exclusively at first) led to dilution of mission because they were also attempting to administer to ordinary access accounts at the same time, and some have failed altogether because they never considered the full brunt of supporting what has ultimately ensued in this space from a volume perspective (while also catering to their own subscribers) for such a mission, from the outset.
Some of the earlier NAP bypass players who operated under dual roles (ISPs and Exchange operators) soon learned that managing major Internet Backbone Routes and exchange points was a full time job in itself, and that ISP access and account administration, likewise, was a full time job. They learned that these two areas, while related in many ways, did not necessarily lend to the same types of synergies as they thought they would. Note, as an example, how AOL itself does an arguably excellent job of adminstering to its subscriber base, but at the same time outsources a good deal of their backbone chores to others.
ABOV would have you believe that they fall into a category all their own. There is some merit to their claim, though, and I say this for several reasons.
According to David Rand, CTO of AboveNet (prior to AboveNet he worked for Cisco) ABOV's original intent was to be a colocation site about four or five years ago during their early days of operation, and not a networking company, per se. They never intended to become a network company, despite the fact that that is what it has turned out to be in many ways.
You ask:
"Can you explain in more detail how you define (1) route peering [[fac: private peering]] for ISPs and (2) what are the requirements to build a "route peering" [[fac:private peering]] system (ie. a. software and/or b. hardware dedicated fibre backbone)."
I don't feel that I am fully qualified in answering many of these questions. Although I have covered them sufficiently in order to state with some degree of confidence that ABOV goes about their business in some ways that only remotely resembles any of the others. They have some inherent philosophies which are closely tied to the lore of the 'net, regarding how their customers should be treated, for example. They have unique ways for billing for bandwidth utilization through some unusual algorithm-driven processes. And they have some unique ideas about how bandwidth should be transported between their own nodes and those of their subscribers.
Also somewhat unique is their constant display of network utilization statistics (customer network management windows) that they provide to each of their users, which is a feature that many of the big five other "BackBone Club Members" refrain from availing themselves of, due to the sometimes "double-sided sword" nature of doing so. But they provide full utilization monitoring of their routes, nonetheless.
"RMII has already installed broadband switches at their larger ISP customers sites within Colorado..."
Here lies one of the major difference in modality between what RMII and many other Tier Ones are proposing, or saying they are doing, and how ABOV would respond to the same situation. ABOV's stated policy (as I suggested previously) is to refrain from going directly to their target ISPs and content providers with their own bandwidth facilities. They do not want to become more of a transmission company than they absolutely have to (they rationalize that others can do that equally as well, and they are in the routing business, not the telephone or local fiber optic construction business). Makes sense to me.
In other words, they will not backhaul their customers to their own location in the fashion you have described concerning RMII. Rather, ABOV asks that their ISP customers and their Content Providers to come to them, to bring their traffic to ABOV's major node points, instead.
Another difference might be in the nature of ABOV's versus RMII's proposed protocol mixes. You say:
"They utilize a software centric multiservice WAN platform, based on ATM cell switching for interworking among Frame Relay, SMDS, ISDN and ATM. Isn't this really very specific "Route Peering" technology? What prevents RMII expanding this concept to their large ISP customers throughout the U.S.?"
ABOV, thus far, and wherever possible, has refrained from using ATM, ISDN and Frame Relay, preferring instead to allow each ISP and Content Provider customer a minimum of T3 (DS3) access to the Internet at Layer 3 using Internet Protocols... not ATM or Frame Relay.
When you get into the alphabet soup boxes all kinds of complexity begin introducing themselves, and one of the principles behind ABOV's operating philosophy is to keep it simple. Robust, fully capable, but simple.
"I can only conclude that AboveNet has in place some routing agreements with other ISP's that utilizes a "private peering" service model that guarantee's a minimum bandwidth (on demand) to each of their subscribers."
Absolutely, and this is the semantic difference I alluded to above whereby I corrected my previous use of the term route peering, when I meant to state private peering. I should expand just a bit on this notion of bandwidth on demand, as it relates to the evolution of how ABOV got into the private peering game to begin with. They used to rent from the Big Five backbone players [UUNET specifically] at, say, the T3 level. When they realized that they were good at a couple of meg but started losing packets as soon as they turned it up a notch, it became clear to them that no one was getting the full T3's worth (or whatever capacity was being leased at the router port level) that they were paying for.
The larger players were clearly oversubscribing beyond the normal bounds of statistical latitudes. From what I have read about this, I have inferred that they weren't even able to "burst" to T3 when needed. Again, however, that is only an inference of mine. In any event, they saw this as an opportunity, and the rest should be obvious: they sought to fill the need for a better means of providing Internet trunks to ISPs, and they proceeded building ABOV as we now know it. It's not as simplistic as the way I am explaining it, but that is the general idea.
In order to ensure against this same level of short-changing from taking place to their own customers, they use a rather unusual megabit accounting scheme based on sampling and averages, governed by an algorithm which assesses monthly charges in ways which are not considered intuitive, but work out very fairly for their customers.
It's kinda complicated, and I don't want to go on about it here in this post. See if you can borrow a copy of the May 1999 Cook Report on Internet from someone. Therein you will find a 12-page interview printed in microscopic font size taking place between ABOV's David Rand and Gordon Cook, which took place in February of this year.
If you recall, in my previous post I did indicate that most other ISPs also engage in peering. Indeed, peering between networks and other Internet components has always been one of the fundamental reasons for the growth and survival of the Internet in the first place, from its very outset. Peering, private or otherwise in itself, is not what sets ABOV apart alone from RMII, or from EXDS, or from ISLD. One of the main ways differentiators which sets them apart, again, is their unwillingness to compete with their own customers. It is their contention that this develops a higher level of trust and cooperation, hence loyalty, between them and the entities they serve. This is not my personal observation, rather I am reporting to you what their claims have been, although I find them to be highly believable and rational.
Most of the remainder of the points you raised, while valid in themselves concerning private peering, do nothing to alter the fact that RMII falls into the category of "all others" in many ways, since, they, too, like all the others in their class, do in fact compete with the ISPs who they are peering with and for. And ABOV, ostensibly, does not.
I'd like to thank you for putting together that excellent reply. I can tell that you put some work into it and I appreciate it. Let me know how you make out securing that May '99 Cook Report.
Best Regards, Frank Coluccio |
