[October 18, 2000]

Choice in DSL Transport Not So Clear

Reader expresses a different viewpoint about ATM vs IP transport for DSL deployment, particularly in terms of scalability.

[Response to the Executive Perspective feature article Strategic Choices in DSL Deployment from Oct. 17th.]

Dear Editors:

I wonder why this article was not accompanied by a "paid advertisement" label.

Mr. Gilbert puts such a slant on this article such that a mountain goat would not be able to maintain its balance. I was intrigued to discover what choices Copper Mountain Networks' products offered. I see from their corporate web site that they actually provide ATM as a transport.

This mildly surprised me, given Mr. Gilbert's clear choice. I do notice, however, that ATM switched virtual circuits (SVCs) are not supported on their equipment. Thus the solution for ATM's scalability problem cannot be provided by Copper Mountain Networks. I would venture that if asked their response would be that it will be supported in the "next" revision or RSN.

From my past experience I would definitely support the premise that PVCs are not a viable solution for scaling to support provisioning of any large number of customers.

The real choice is whether ATM SVCs or IP Class of Service (COS) will provide the required functionality and scalability to support the estimated increase in DSL subscribers.

The article seems to imply that the ATM solution would occur on a totally ATM network and ignores what would appear to be a more logical architecture. If fact this architecture can be found in Copper Mountain Networks' own literature, namely that the ATM transport is only used to connect ILEC and/or remote central offices to a central aggregation point where the data traffic is handed off to an IP network and the voice traffic is handed off to the existing POTS infrastructure.

So the scalability issue then is less of an issue than the article appears to infer. The issue over whether ATM is a new and little used protocol, while IP COS is not, is misleading.

Scratch beneath the surface of many, if not most, large IP carriers and you will find an ATM transport infrastructure.

On the other hand, while IP is the fundamental building block of the Internet, Class of Service is a yet unproven and not supported by any of the major ISPs.

So from my perspective, both ATM SVCs and IP COS are untested in whole scale scalability and the choice is NOT so clear

Brian Fellows,
NewsEdge Corporation Network & Security Manager