Discussions of the migration from an ATM network to an all-IP network usually focus on the backbone, where carriers can realize huge increases in bandwidth that will allow them to respond to the growing demand for services such as the triple play of voice, data, and video. To unicast 50 IP video streams, each representing 2-3 Mbps, to 50 different subscribers requires a sustained performance of 100-150 MB. That much traffic, combined with the overhead required to encapsulate IP into ATM, would overwhelm a 155 Mbps ATM backbone. For a 1-10 Gigabit IP backbone, on the other hand, 50 IP video streams is a drop in the bucket—and there is no need for ATM.

However, not much attention has been paid to how to migrate from ATM to IP in the last mile between the Central Office (CO) and the subscriber’s home or office. Yet the efficiency penalty of ATM on the local loop is even greater, as the local loop still represents the bottleneck in the end-to-end video stream. A 10 to 15 percent efficiency penalty on a 24 Mbps loop translates to enough bandwidth to service a full stream of video. The full benefits of IP will not be available to everyone until carriers and providers migrate to a true end-to-end, IP from subscriber through to core network. 

Is ATM’s Virtual Pipe Really Needed?

Historically, the ATM class of service mechanism was developed to allow the deployment of voice and data services alongside one another. Digital Subscriber Line Access Multiplexers (DSLAMs) employed the already existing ATM protocol because it was the closest thing to carriers’ familiar circuit switching. ATM creates a virtual pipe or tunnel from the Customer Premises Equipment (CPE) to the DSLAM all the way across the server to a centralized Broadband Remote Access Server (B-RAS). The DSLAM became an ATM switch, transporting voice traffic to the voice gateway and data traffic to the B-RAS.

Figure 1 illustrates the ATM DSLAM deployment model employing RFC-2684 Multiprotocol over ATM encapsulation. The figure shows how much processing is required to convert IP data at the CPE into ATM and back into IP again in B-RAS. This model not only is profligate of computing power but also turns the B-RAS into an IP keyhole entry and exit to and from subscribers that acts as a bottleneck to quick delivery of broadband services. 

Figure 1—Current ATM DSLAM Deployment Model

For years, users were willing to make the trade off of ATM’s overhead in return for a virtual pipe that isolated their traffic from everyone else’s. In addition, ATM pipes look so much like circuit switching that carriers could just forget about them. Once a carrier implements an IP link to the backbone, however, there is no need for a virtual pipe, and thus no need for circuit switching. In fact, ATM encapsulation, pipes, and switching become obstacles to the delivery of broadband services. 

Next page >>

page 1 of 3

Newsletter FREE Trial  |  Subscription Info  |  Marketing & Advertising  |  About Us  |  Contact Us
Copyright © 2005 Converge! Media Ventures, Inc.  All rights reserved.
Designed for use with the latest version of Microsoft Internet Explorer