Why IP beat out ATM for use in Next-Generation Voice Networks

For a good part of the 90’s, conventional wisdom in the telecommunications industry held that asynchronous transfer mode (ATM) and Internet Protocol (IP) were competing technologies. IP, the prevailing notion held, was a “best effort” service because IP-based networks indiscriminately discarded packets if there was congestion. There was no standardized protocol to identify and prioritize video and voice. The industry at that time maintained that best effort protocols would not be recognized by carriers as acceptable for voice traffic. Because of this, ATM’s ability to create virtual connections and to prioritize voice and video so that packets would never be dropped and quality of service standards were met gave ATM a vital advantage.

ATM also had speed advantages, capable of speeds of 155 and 622 megabits per second. Ethernet LANS at this time were limited to 10 megabits per second, and IP used between networks was also slower than ATM. For these above reasons, when carriers wanted to improve their networks, they decided on ATM equipment. I personally was involved in Fleet Bank’s multi-million dollar loan to LDDS Worldcom (now MCI) in the late nineties for ATM gear for their UUNET data network subsidiary.

However, despite all of its inherent advantages, ATM gear was costly and complex to install. There was a slight push around this time for ATM to be used in LANs, especially in campus backbone networks and NSF research nets, but ATM was far too expensive to deploy on the desktop. So ATM was relegated to use in large corporate backbone networks and carrier traffic-bearing data networking.

So, as you can imagine, mainly due to the speed and quality-of-service advantages, established telecom vendors and most new softswitch vendors initally at least based their next-generation voice switch architecture on ATM rather than IP. Meanwhile, improvements in routers and faster speeds on IP networks were making IP networks much more suitable for voice. Also at this time, Cisco’s TAG protocol, the forerunner of today’s MPLS, was being developed and was maturing. The MPLS protocol marked packets so that voice and video could be prioritized. This capability let IP packet flows be handled similarly to ATM virtual connections, which treat various types of traffic differently. Concurently, IP speeds improved from 10 megabits per second to 100 megabits per second speeds and, eventually, gigabit speeds.

With these notable improvements in speed and service qualities, along with the fact that corporate endpoints were already equipped to deal with IP traffic, the founders of Sonus Networks (Westford, MA), in 1997, choose to base their next-generation, softswitch-based voive infrastructure on IP. In this manner, Sonus was granted a head start over competitors who initially developed platforms based on ATM, losing time and previously invested development money when they switched over to IP – too late.

In related news, Sonus Networks of Westford, MA recently (11 March 09) announced it is “restructuring” again, cutting another 60 employees to complete its third round of cuts in three months. The company said this cut will equal out to about 6% of their workforce. The total job cuts within the three months has added up to 160 jobs lost at the networking equipment vendor. Sonus has a baseline resource level of approximately 1,000 people.

BT (British Telecom) has also recently (3 May 09) announced that it is cutting back on deployments of equipment and resources for its 21CN Next Generation Network (NGN) project.

Jefferies & Company analyst George Notter points out in a recent research note that Sonus was slated to in late 2007 to provide an Access Gateway Controller Function (AGCF) to enable communications between core IP and PSTN access networks. However, Sonus may have to take a revenue hit now, as BT discovered that its NGN network architecture is too costly. They have halted the NGN Network cutover project.

Update: Sonus Networks announces 2009 Q1 results