Any creation worthy of the name does more than reflect the vision or
embody the toil of a creator or creators; it also assumes a life of its
own. And, no matter what measure of perfection it may achieve, it also
remains a work in progress. Such is art. Such is engineering. Such is the
next-generation network. And, to art historians and analysts alike, no
important work exists in isolation. For example, important works of art
are seen as harbingers or culminations of sweeping movements, rich
traditions, and epoch-defining styles. And this same perspective, this
encompassing view, may reveal technological trends. A fine thing, too, for
the destiny of any technological artifact is quick obsolesence! But, if we
had but the eyes to see, we would recognize how the ingenuity of the
moment may demonstrate that which is lasting and profound.
One example of this point of view, applied to technology, is a book
entitled The Soul Of A New Machine. In it, the author, Tracy Kidder,
describes how Data General Corporation not only created a new 32-bit
minicomputer, but also met a seemingly impossible deadline. The computer,
the MV-8000 Eagle, has long since become outdated, but the work practices
engendered by the Eagle project live on. You could even say the Eagle
project anticipated the hacker culture. Rarified and rigorous, the hacker
culture is one in which intense single-mindedness and incredibly long
hours are now taken for granted.
And just what is the soul of the new machine? The book admits any
number of interpretations. It could be the dedication of bleary-eyed
hardware and software engineers. Or it could be a sense of urgency more
typical of war than of everyday commerce. And what is the new machine? Not
just the MV-8000, surely. Rather, it could be a new kind of corporation,
one that instills a sense of urgency and demands a new work ethic.
And now we come to the next-generation network, that is, the new public
network, which will accommodate both voice and data, and which will first
isolate and then obviate traditional, circuit-switched facilities, making
way for a thoroughly packet-based infrastructure. At least, that is the
superficial view. But what, I ask, is the soul of the new network?
The new network has a soul inasmuch as it is an expression of humanity
-- not of individual genius, mind you, but of the organized efforts of
many gifted people. Organization is a crucial point, the defining point,
because designing and building the new public network is beyond any one
corporation. The new network signals a new industry, a new economy. So
demanding is the new network, in terms of reliability, capacity, and
flexibility, that corporations will work together as never before, weaving
an ever finer fabric of core competencies and outsourcing schemes,
acquisitions and spin-offs, stalwarts and startups.
The soul of the new network is mutability, in the components that
constitute the emerging infrastructure; in the corporations that provide
and assemble and add value to the components; and, ultimately, in the
economy driven by the new network. And this mutability will be taken to
new extremes, in answer to unprecedented time-to-market pressures.
But enough generalities. Let's examine the emerging public network,
some of its various components, as well as its overall shape. We will see
that regardless of the variety of network elements, the new network
represents a remarkably consistent vision.
A ROUGH CHRONOLOGY
The early PC- or server-based systems, that is, the early IP gateways,
were scarcely carrier class, but they aroused interest anyway. Their
potential could hardly be ignored, not following deregulation in the
telecommunications space. And while these gateways were lacking in terms
of density and availability and reliability, they did have the virtue of
audacity. Characteristic deployments involved bypassing the public
switched telephone network (PSTN), skirting long-distance charges and
exploiting arbitrage possibilities.
Aware of the opportunities in IP telephony, and being loathe to cede
them to impudent upstarts, established vendors began retrofitting existing
data networking platforms so that they could carry voice traffic. Soon,
the IP telephony market was joined by router-based gateways as well as
equipment with their origins in remote access servers (RAS-based
gateways). Such equipment was introduced by Cisco, 3Com, Bay Networks
(soon to be acquired by Nortel), and Ascend (soon to be acquired by
Lucent).
In addition, established switch vendors began outlining schemes whereby
circuit-switched solutions could migrate to packet telephony. The Nortel
Succession platform is a case in point, as is the Lucent 7R/E (the
"R" stands for revolution, the "E," for evolution).
Whether the switch vendors were reacting to IP telephony as a threat, or
whether they were leading the parade, depends on who you ask.
And, finally, we come to a couple of hardware approaches that are much
in the news of late. First, we have the next-generation switch. Typically,
this platform is custom-built, boasting of carrier-class attributes,
including reliability and (more recently) high port density. Vendors
supplying next-generation switches include Converent Networks, Salix
(acquired by Tellabs), Sonus, TransMedia (acquired by Cisco), and
Unisphere Solutions (formerly Castle Networks, acquired by Siemens).
Second, we have the open development option, which relies on commercial
"off the shelf" components. This option, thanks to continued
improvements in CompactPCI, promises to introduce PC economies and
development speeds to the telecommunications space. Board-level gateways
compatible with the CompactPCI universe include offerings from AudioCodes,
Brooktrout, Dialogic, Natural MicroSystems, and Radisys.
What are the advantages and disadvantages of each of these hardware
approaches? It can be hard to say, especially since innovation proceeds
apace, often blurring distinctions between categories. (Mutability,
again.) Also, most approaches agree on an overall vision for the
next-generation public network.
At any rate, we'll consider the input provided us by vendors
representing each of the approaches we've just described. Specifically, we
received input from Lucent's Internetworking Division (discussing
RAS-based platforms), from Unisphere Solutions (discussing the
next-generation switch), and Dialogic and Natural MicroSystems (discussing
open development). Finally, we'll review what Cisco and 3Com have said
about the next-generation network as a flexible service creation and
delivery infrastructure. Their comments are as illuminating as any on this
subject, and conform to widely held, prevailing views about the future of
the next-generation network.
WHERE THERE'S A WILL, THERE'S A WAY
Lucent may take service providers over various paths to the
next-generation network, an observation confirmed by our communication
with Sherman Scholten, technical marketing manager, Edge Access Networks
-- InterNetworking Systems, Lucent Technologies.
Communications Solutions: What "next-gen" migration
approaches might Lucent accommodate?
Lucent: Depending on the carrier's voice and data convergence
strategy, Lucent is either equipping the customer's migration to
packet-based switching from legacy, circuit-switched equipment with
custom-built hardware (such as the 7R/E) that leverage past investment
(such as the 5ESS), or we are facilitating the build out of a
voice-enabled data network using existing data/RAS switching equipment
(such as the MAX TNT and APX 8000) -- often with a simple software upgrade
(such as MultiVoice).
CS: And what kinds of service providers could you work with?
Lucent: We are delivering a variety of products to market that
address the specific needs of every type of carrier, including incumbent
local and long-distance telecommunications carriers, competitive local
exchange carriers, Internet telephony service providers, and emerging
application service providers targeting niche markets like Internet call
waiting, unified messaging and click-to-call Web applets.
All carriers, regardless of where they are in starting down the
convergence highway, demand open, programmable, multi-purpose,
multi-service networking designed for high-performance and carrier-class
reliability, not to mention high port density and low power consumption.
The universal port (voice, fax, remote access) capabilities of the MAX TNT
and APX 8000 access concentrators are a proven, successful example of a
convergence-minded capability.
CS: As for remote access concentrators... How are they
relevant? How might they be advantageous?
Lucent: The access market is undergoing a time of remarkable
transition combined with dramatic growth. Driven by the Internet's
unprecedented rise, the evolution of this element of the public network is
viewed by analysts as the next significant step in the information
technology revolution. As the access market develops to offer a converged
network infrastructure optimized for transporting multiple traffic types,
and to support exciting next-generation applications such as unified
messaging, service providers are at the forefront. Without question,
service providers of all types, shapes, and sizes will undergo fundamental
changes, as they will also be required to evolve in an intensely
competitive market.
Access platforms allow service providers to position themselves in the
market for integrated voice and data services. Their ability to function
as both a remote access concentrator and VoIP gateway enables the service
provider to decide when, where, and how to engage each service, all the
time investing in a single platform with common hardware and software
components, thus significantly reducing infrastructure and operating
costs.
CS: Going forward, how will access platforms contribute to the
evolution of the new network?
Lucent: The real challenge for a service provider is to build converged
networks that are as reliable, robust, and scalable as today's voice
networks. Our approach in addressing the next-generation public network is
to keep intact the feature richness of today's PSTN and couple the power
of IP networking and more specifically the Internet with products like the
Lucent Softswitch using the MAX TNT and APX 8000 as media gateways.
Next-generation public networks are being built today in preparation for
the implementation of highly advanced, functional, and integrated
applications. Advances in transport technologies, switching technologies
and application integration have made it possible to deliver enhanced
voice and data services in volume and with certainty.
CONVERGENCE FROM SCRATCH
To help us understand the customized approach, that is, the advantages of
the next-generation switch, we talked to representatives of Unisphere
Solutions: Susan Blanchard, senior product line manager, and Kate Zebrose,
senior engineer.
CS: Why a custom implementation?
Unisphere: We have found that a custom implementation using standard
hardware components allows for the most cost-effective solution for our
carrier customers that are looking for density, reliability, and resource
management. When the carrier networks actually rolls out services, the
volumes are quite large -- more than sufficient to justify the development
cost of a true carrier-class implementation.
CS: What about alternative approaches? CompactPCI? Remote access
solutions?
Unisphere: CompactPCI solutions simply cannot meet the density and
scalabilty requirements for full-scale carrier deployment. A good example
of this is on the AudioCodes Web site. The highest density compact PCI
solution is the TP-400. This supports only 192 ports per module, while
their PMC module the TPM-800 supports 240 ports.
Remote access solutions that are field proven also do not have the density
requirements. Modem function is a significantly greater DSP requirement
than VoIP, which means the RAS solution has a significantly lower density,
particularly compared to uncompressed and ADPCM compressed voice. These
solutions also lack the resource management, signaling, and redundancy
support for deployment in the voice network. The current solution is to
provide these elements with adjunct equipment, which also needs to be
redundant. While adequate for initial roll out, this increases the
required footprint.
CS: But when you say custom, do people hear proprietary?
Unisphere: We have found that a proprietary design has given us greater
flexibility in addressing this market. We learned from the RAS experience
and have insured adequate air flow through the chassis for initial
deployment and extensive expansion. Equipment vendors will need
proprietary solutions in order to meet the cost and footprint requirements
that will make this network architecture cost-effective for carriers.
ADDING VALUE, AND QUICKLY
In our communications with providers of commercially available components,
several themes were pronounced. First, these providers emphasized the
distinction between platforms that accomplish routing or switching, and
platforms that provide for enhanced services. (At present, enhanced
services platforms are where commercial components are most commonly
deployed.) Second, the importance of outsourcing as a way to achieve
faster time to market. (That is, relying on commerical components instead
of developing all components in-house.) Third, the potential for yet
broader application of the open development approach as CompactPCI
improves in terms of density and availability. For these insights, and the
commentary which follows, we thank Jack Chase, Natural MicroSystems, and
Chandresh Ruparel, product line manager for Internet Telephony, Dialogic.
CS: Could you give us an idea where open development fits into the
creation of the next-generation network?
Natural MicroSystems: A little background might be helpful. First, new
carriers, in the wake of deregulation, deployed PC-based systems. And,
eventually, router-based systems became available. These have expanded the
market considerably, and have introduced lower cost/port alternatives. In
addition, vendors such as Cisco have increased the time-to-market
pressures to which traditional telco equipment providers must respond. And
so, the traditional providers have put increased emphasis on outsourcing.
I should note that outsourcing has always been a theme; it's just more
pronounced now. Also, we've seen increasing emphasis on enhancing existing
platforms, which creates opportunities for those who may provide design
and customization services.
But the key is time to market. This is what accounts for the popularity of
open systems with network applications. (I need hardly add that open
systems are already popular in the PC/computer telephony space.) It is a
relatively new phenomenon, that is, the ability of open systems to fulfill
network application requirements. This ability directly attributable to
advances with CompactPCI. Its "off the shelf" capabilities
simply weren't an option before.
CS: What about other hardware alternatives
Natural MicroSystems: RAS-based systems... these are easily available, but
generally harder to work with when assembling larger systems. And they are
not necessarily an attractive approach for developers looking to add
value. Of course, the RAS vendors themselves may enhance the value of their
own platforms.
More recently, large systems, such as those from Sonus, are entering
trials. These systems, which present a more or less "from
scratch" approach, offer high port density, typically between 50,000
to 100,000 ports, typically within a large chassis, say, 9U.
Such density or scale is not yet available with CompactPCI. (At present,
boards from Natural MicroSystems provide for up to 240 ports/board.) It
remains to be said, however, that open systems providers will keep
"raising the bar," that is, they will keep scaling up CompactPCI.
In any case, the current developmental emphasis with CompactPCI in on
value-added applications, that is, on deploying enhanced services, having
them run on the same platform. Most developers are adding value at the
softswitch, and not so much at the DSP level.
CS: We've heard about the importance of flexibility. Could you
elaborate?
Dialogic: Currently, we see considerable uncertainty with respect to
design alternatives, that is, with how multiple types of functionality --
including gateway and concentration functionality -- may be encapsulated
within fewer boxes. Given this uncertainty, the flexibility promised by
server-based solutions is very attractive, particularly since server-based
solutions, based on CT components, continue to acquire carrier-class
attributes. Moreover, these carrier-class attributes emerge from open
technologies as opposed to proprietary technologies. For example,
CompactPCI solutions are improving both in terms of density and
sophistication, offering high availability and remote diagnostics and
maintenance.
CS: So, in addition to time to market, the need to manage uncertainty
is an important factor.
Dialogic: Yes, in addition to uncertainties with respect to
infrastructure, there are uncertainties with respect to standards. Which
will prevail? MGCP? SIP? Where should developers and service providers
invest their resources? PC-based solutions are popular here, since they
offer a lot of flexibility. And the trend will become only more pronounced
as the viability of standards-based approaches becomes clearer. In other
words, once this viability is demonstrated, there will be less incentive
to develop proprietary solutions. In addition, the focus of large players,
such as Intel, on providing carrier-grade solutions to the service
provider market make commercial components very attractive for deployment
in the public network.
CS: And what of the need for services differentiation?
Dialogic: Again, this comes back to flexibility. You need flexibility if
you are to add services differentiation based on features. If you lack
flexibility in creating new services, you will have a difficult time
competing. But this is where the strength of a broad development community
can be helpful. And, indeed, the current developmental emphasis is on the
enhanced services side. (Enhanced services, I should point out, cover such
things as softswitch and call agent functionality, and billing and
signaling.) Also, currently, deployments are mostly on the access side,
and not so much in the core of the network.
WHAT'S NEW? WHAT'S DIFFERENT?
So far, we've reviewed a variety of design approaches to the
next-generation network. But what does it all mean? Let's return to my
opening trope, that is, my reference to The Soul Of A New Machine. While
this book draws an admirable picture of ingenuity and dedication, we may
feel that Data General's work was nonetheless limited in scope. We are,
after all, talking about a machine, a box, something which suggests
confinement.
Today, having experienced the Internet, we're more accustomed to expansive
and even ethereal visions. So, we may consider The Soul Of A New
Network.
Within this network, functionality may be distributed so finely that few
would care to know where any particular function actually resided. For
most, would it matter if we could isolate a particular box or set of
boxes?
Granted, alternative approaches to design are nothing new. And neither are
outsourcing options. And neither are calls for standardization. But, with
packet-based networks, design challenges are being broken into ever
smaller pieces. Ever more discrete functions are being exposed to
dedicated engineering talent. The point is, a sufficiently large
difference in degree may constitute a difference in kind.
An analogy to the physical sciences may be instructive. It is a fact that
chemical reactions proceed more quickly when the surface areas of the
reactants are maximized. And so it is that while a lump of coal may glow,
a cloud of coal dust is actually explosive. We could liken the glowing
lump to the traditional public network, and the explosive cloud to the
packet-based next-generation network.
In the next-generation network, the distribution of functionality is
sometimes referred to as decomposition. For example, it is fairly common
to hear of the decomposed switch, which is basically a Class 5 switch
broken down into its functional components, an operation that relaxes
co-location requirements while implementing packet-based communications,
as opposed to circuit-switched communications. Typical elements within a
decomposed infrastructure would include a media gateway (for transport), a
signaling gateway (for SS7 functionality), a softswitch (for gatekeeper or
call agent functionality), and the media server (or application server or
feature server).
THE GRAND DESIGN
The proliferation of design options and network elements needn't cause
confusion. There is actually a method to the apparent madness of the
decomposed network. And that method typically involves separating network
functions into planes (in Cisco's words) or tiers (in 3Com's words).
According to Cisco, the traditional switch architecture may be broken into
these layers: the connection plane, the call control plane, and the
service plane. Between these planes, interactions may be managed via open
protocols.
Cisco calls the approach Open Packet Telephony (OPT). According to a Cisco
whitepaper, OPT "opens up the telephony infrastructure to give
service providers unprecented control over their networks. Freed from the
constraints of proprietary solutions and overlaid single-technology
networks, service providers can move into new markets, services, and
revenues."
The same basic vision is described in slightly different terms by 3Com, as
we learned in our communication with Houman Modarres, 3Com's director,
product management, IP Telephony, Carrier Networks Business. Modarres is
responsible for the evolution of media gateway, softswitch, and
back-office components of 3Com's CommWorks IP telephony framework for
service providers.
CS: Please describe the challenges facing service providers.
3Com: In the last several years, a multitude of leading service providers
have accelerated their efforts to build next-generation networks that
bridge the gap between legacy circuit-switched infrastructure and more
efficient and cost-effective packet-based IP networks. Their goal is
threefold: to meet the exploding demand for integrated voice and high-speed data access, provide a common
"look-and-feel" to their array of bundled applications, and
offer compelling new services that enhance customer loyalty.
However, in doing so they face significant challenges, including the need
to interconnect a multiplicity of networks, technologies, and systems.
Most are controlled by divergent protocols and incompatible network
management and operating systems. And the list of applications service
providers require to offer differentiated services grows longer every day.
CS: And how might the service provider bring all these elements
together into a converged data and voice network?
3Com: The key is to build an infrastructure that is both open and
standards based, one that combines standard hardware and software
solutions designed to interwork with legacy PSTN equipment and services as
well as next-generation IP networks. Such a system offers service
providers the ability to mix and match best-of-breed components to create
a custom configuration for individual applications, plus the ability to
add new elements and upgrade their networks on demand.
Building this distributed network requires three basic components: 1)
media gateways, which provide universal connectivity; 2) softswitches,
which provide open call control and signaling intelligence; and 3)
back-end servers, which enable enhanced services and applications. This
three-tier approach offers a flexible and scalable way to meet the needs
of today's service providers and carriers.
CS: Could you elaborate on each of these components?
3Com: Media gateways integrate multiple types of traffic and accommodate
disparate access networks. They provide open interfaces to all back-end
servers, making them available to all network users, independent of access
medium and speed.
Softswitches offer a truly open system design lacking reliance on any one
operating system or hardware platform. As a result, they can support
generic interfaces that enable rapid customization and faster introduction
of new services. In addition, they easily accommodate technological
advances, such as enhanced call control protocols and innovative
mechanisms to integrate with existing IN-based service platforms to
deliver higher call handling capacities.
Back-end servers, which are based on the service providers' standard
computing platforms of choice provide all the required network-centric
services, such as provisioning and directory mapping; application-centric
services, such as unified messaging, IP fax, and IP Centrex; as well as
user-centric servers that provide a baseline for personalized services,
subscriber profiling, managed push, and self-provisioning.
CS: Is it all that easy? It sounds almost plug-and-play.
3Com: Actually, work remains to be done before distributed networks are
truly plug-and-play. A positive sign is the emergence of industry-wide,
accepted standards such as H.323, SIP, and H.248 Megaco. These imply that
the monolithic PSTN switching platforms with complex and often proprietary
interconnections will soon be a thing of the past. In the face of this, a
strong case can be made for a three-tier infrastructure based on open
interfaces, built on standard computing platforms with multiple vendors
providing services and equipment for each tier. Such a solution provides a
scalable and cost-effective means for merging legacy and next-generation
networks, while providing a framework that meets the requirements of
tomorrow's emerging services and applications.
[ Return
To The October 2000 Table Of Contents ]
|