What happens when you flip on the softswitch? After the
marketing bubble-wrap and engineering Styrofoam have
been pulled away, what is this piece of equipment going
to do for you? A few years ago, you might have been
satisfied to have it manage the setup of a carrier
quality voice call over a packet infrastructure, using
any reasonably cosmopolitan media gateway control
protocol. But a lack of precision in the initial
definitions of a softswitch combined with the
intoxicating rush and subsequent crash of funding sped a
roller coaster ride of expectations. This is a good time
to step back and assess where the action is today in
softswitch deployment and where tomorrow's development
is likely to take it.
BACK TO BASICS
According to legend, one of the first comments received
by IBM researchers upon the unveiling of their new
invention of the microchip in 1968 was "But what is it
good for?" The birth of the softswitch was greeted with
no such lack of imagination. While the academics debated
the requisite degrees of "softness" (hardware
independence), "openness" (reliance on standard
protocols), and other criteria in competing softswitch
definitions, pragmatists raced to market with
demonstrations of compelling value.
The first intrinsic value provided by the softswitch
is in network "flattening." Because the bandwidth
required for the packets that contain the bearer path or
voice encoding of a call is anywhere from 50 to 5,000
times that of the signaling required to make and take
down the connection, the softswitch's role in separating
these two communication channels profoundly shifts the
capital and operational expenses of legacy voice-centric
networks. No longer is a dedicated mesh of "fat"
voice-only pipes required to interconnect each switch in
the network. By consolidating voice and data traffic at
the customer site on one link, virtually eliminating
trunk provisioning, and simplifying operations, service
providers have seen access cost savings of up to 35
percent and operational cost savings of up to 50 percent
as well as up to 85 percent faster deployment.
A second benefit that flows directly from the
softswitch's delegation of bearer control to the media
gateway is the greater geographic span of a given
instance of network intelligence. With a five- to
ten-fold increase in the number of air miles between a
softswitch and a media gateway, we have seen a 25- to
50-fold increase in the serving area of a softswitch.
This greatly reduces both the costs of geographic
expansion as well as the time it takes to deploy new
services.
A third de facto benefit from carrying bearer path
over packet is that it allows traffic to skirt taxation
and other regulatory barriers in certain areas of the
world. Some of the initial European deployments of
softswitches were able to make compelling business cases
through the promise of the avoidance of burdensome
tariffs alone.
BEYOND NETWORK SIMPLIFICATION
So three of the key benefits of softswitches derive
simply from their role in enabling carrier class voice
over packet networks. Nevertheless, as these early
promises of softswitches became well known, a cycle of
increasingly exuberant expectations was established.
First, came the purists' drive for 100 percent software
switches based on commercially available hardware. We
have seen early adopters of this approach already reap
the benefits of Moore's Law by migrating onto newer
generations of hardware offering better
price/performance between Beta trial and General
Availability. Second came the open architecture
advocates' promises to meet and exceed Class 5 service
sets on the basis of an ever-increasing litany of new
protocols and hordes of eager third-party software
developers. Third, and perhaps most important, was the
ability of the softswitch to lower the barriers to entry
for "greenfield" opportunists in the world of
communications. As the expectations multiplied, so did
the number of vendors: By mid-2000, dozens of vendors
had unveiled new entries, and the International
Softswitch Consortium had 185 members.
Over a year later we see a comparatively modest set
of real world deployment scenarios for softswitches.
Least ambitious, but perhaps most widely touted are
Internet Offload deployments, where media gateways
controlled by softswitches actually route TDM dial-up
modem traffic onto stripped down PRI trunks for cost
savings. Independent analysts are now saying this market
has already peaked, due to the decline in new ISP
subscriptions and overcapacity in DSL and Cable access,
which are alternatives to dial-up. Second, but perhaps
less publicized, are the 'toll bypass' solutions that
have seen greater success outside of North America,
where comparatively high rates for long-distance and
especially international long-distance offer more
favorable arbitrage opportunities. Slower to develop are
the carrier class end office replacement softswitches.
Several roadblocks have emerged to slow the progress
of softswitch deployment. First, while a basic VoIP call
can be emulated in the softswitch fairly readily, there
is a mountain of non-trivial specializations that lies
between a basic call and full Class 5 transparency. The
incredible complexity of design assumptions accrued over
decades of connection-oriented networks do not
seamlessly graft onto a connectionless network. Among
the boulders at the foot of this mountain lie such
showstoppers as emergency (E911) and CALEA or Lawful
Intercept (wiretapping). While marketing adrenaline was
diverted toward the revolutionary possibilities of the
softswitch, these relatively unglamorous table-stakes
requirements did not necessarily get the right priority.
Beyond design issues, archaic inter-LATA transport
restrictions obviate some of the network flattening
capabilities of softswitches for the regulated service
provider space. Also, the TDM entrenchment within
inter-company billing resolution mechanisms continues to
hamper the full flowering of next generation
possibilities.
While vendors reassure the industry that solutions
for all these hurdles are eminent, it is clear that some
of the initial luster and simplicity of the softswitch
vision will be difficult to reacquire. For example,
CALEA requirements are more appropriately met not by a
softswitch but through dedicated media servers, just as
SS7 network interworking usually scales better on
dedicated gateways. While the softswitch continues to be
a focal point within the network, it is clear that it
has helped create a world in which the 'switch' has more
limited importance as it devolves into a family of
specialized servers. As the dreamlike vision of
softswitches becomes somewhat scaled back, many of the
more exuberant early expectations are finding a new home
in one of its spin-offs -- the SIP-based application
server.
REMEMBER THE TITANS?
We have seen a tendency for new technologies to devour
or co-opt the spin-offs they enable. The story is so old
and universal it was captured in the ancient Greek myth
of Kronos, the leader of the Titans. When Kronos was
foretold that his son would overthrow him, he began a
habit of swallowing his newborn children. Eventually his
wife tricked him by presenting a stone wrapped in
swaddling clothes, thus saving her newborn son. This son
grew to be Zeus, who upon reaching maturity, overpowered
Kronos and himself became the king of gods and men.
Similarly, as new possibilities enabled by
softswitches have emerged, vendors have been eager to
swallow these into their existing softswitch product
descriptions. But just as in the myth, there comes an
idea that is too revolutionary to be fully absorbed by
the earlier model. For softswitches, this "problem child"
may be SIP, the Session Initiation Protocol.
SWALLOWING SIP
SIP is an application-layer protocol that allows direct
peer-to-peer communications between intelligent clients
without the need for centralized call control. The SIP
architecture is articulated entirely through the
vocabulary of Web-based application development.
Therefore, SIP is poised to change how people
communicate just as HTTP and Web browsers changed how
people use the Internet. Because SIP treats voice,
video, and data the same, it can deliver as an
effortless byproduct the perennial white elephant of
telecom: Spontaneous video calls. While SIP's ascendancy
over H.323 in the enterprise is increasingly accepted,
its potential to disrupt carrier-based communication
services is perhaps even more dramatic.
Note that increasing client intelligence is not a
death knell for core networks. As we have seen in the
evolution of Web applications, applets (client-based
intelligence) eventually fuel the creation of servlets
(server-based intelligence). This trend has gone so far
as to create a renaissance for mainframe computing as
industrial strength Web servers over the past few years.
So network-based server intelligence as provided through
SIP-enabled application servers will be an exciting
growth area over the next few years.
As we have seen, softswitches provide compelling
value for streamlined communication networks. SIP
provides an evolution of communications into
peer-to-peer multimedia. While dedicated SIP application
servers are rolling out under their own banner, we also
see softswitches being positioned as SIP servers.
The de-evolution of the switch is spawning increasingly
specialized servers, and SIP is likely to accelerate
that trend. In addition, SIP's ability to span different
network types will help further the differentiation of
SIP-enabled softswitches dedicated to different markets.
Together and apart, softswitches and SIP will continue
to accelerate the transformation to the next-generation
network.
Jim
Thomas is senior manager, Carrier Voice over IP
Marketing, Nortel
Networks. Nortel Networks is a global Internet and
communications leader with capabilities spanning Optical
Internet, Wireless Internet, Local Internet, eBusiness,
and Personal Internet.
[ Return
To The September 2001 Table Of Contents ]
|