With Voice over IP (VoIP)
technology reaching maturity, the triple play
of Voice, Video and Data is a promising application
of convergence, offering the potential of three
revenue streams from a single high speed packet
network.
To live up to this expectation,
Freescale has developed an implementation independent
system architecture built on Freescale's PowerQUICC™ communications
processors containing PowerPC® cores and
DSPs based on StarCore® technology. In
order to broaden the scope of end applications,
an open and modular approach has been taken
to allow flexibility and scalability in both
hardware and software. This method gives development
teams the freedom to concentrate on their own
high-value IP components, enabling both system
application and network convergence.
Traditionally, enterprises
have maintained two separate communications
networks – one
for voice and one for data. By delivering voice
in internet protocol (IP) packets to the desktop
over a managed LAN, VoIP now enables the convergence
of these two networks. Corporate networks are
increasingly embracing the use of VoIP as a
means of enabling new applications, and in
time it will become a dominant form of communication
in the enterprise. With only a single IP network
to maintain, the cost savings can be substantial,
but the greatest opportunity lies in the ability
to deploy and integrate new productivity applications
and enhanced voice services. It's a similar
story for wireless broadband networks. While
providing fast and competitive internet access
technology the killer application will be video
and voice – specifically Voice over IP.
While convergence onto a single IP/Ethernet
packet network is perhaps inevitable within
the enterprise LAN, it is not a certainty for
wireless, metro or long haul wireline transports.
Wireline carriers have conventionally used
Synchronous Optical Network (SONET) or Synchronous
Digital Hierarchy (SDH) and Asynchronous Transfer
Mode (ATM) to deploy Wide Area Network (WAN)
services. Both technologies have immense feature
sets, and of particular importance are inherent
quality of service, high speed, availability,
service differentiation, and Operations, Administration
and Management (OAM).
With increased bandwidth and numerous IEEE ® 802.1
standards, Ethernet is now ready to address
these requirements. The 802.1q Virtual LAN
(VLAN) and 802.1p Class of Service (CoS) standards
especially, have made Ethernet more robust
and resilient. With these two IEEE standards,
carriers can deliver high-priority, low-latency
services, such as voice, over an Ethernet-based
network. It is supported within fiber, copper
and even wireless and has bandwidth support
ranging from half and full duplex 10Mbps, 100Mbps
to 1Gb and now up to 10Gb. This wide range
of media and bandwidth options increases flexibility
in deploying all-IP Ethernet-based network
solutions.
In general terms, network
services that need highly reliable connections
use TCP/IP. Services such as voice and video
requiring real-time delivery use UDP/IP, although
the threat of denial of service and call snooping
have clouded that picture somewhat. VoIP security
enhancements are now fundamental, and the solution
lies in the standards. Securing the call signalling
channels allows a call to be authenticated
and set up securely between two parties. Standards
define the use of the Secure Sockets Layer
(SSL) and/or Transport Layer Security
protocols to secure Session Initiation Protocol
(SIP) and H.323 with H.325 VoIP sessions. While
these call signalling protocols use SSL/TLS
over TCP/IP, the actual voice data is sent
over an unreliable packet-based network, UDP/IP.
While SSL/TLS can be used to implement Secure
RTP or Datagram TLS, it carries a significant
performance cost in hardware and/or software
terms.
The Enterprise or Media gateway will be instrumental
in enabling IP convergence. Gateway functions
(with or without routing, security and switching
functions) can be implemented in standalone
devices, or as modules in multi-service switches,
edge routers, remote access concentrators,
or PSTN switches. In all cases they share a
common set of hardware and software design
challenges.
Divide and conquer
One of the biggest issues facing manufacturers
today is minimizing time to market while maintaining
and building product differentiation. Careful
selection of components is required to produce
a competitive solution that meets the demands
of service providers. In addition to hardware,
the manufacturer must choose a product ecosystem
that enables them to take their system to market
quickly and with ease. The ideal ecosystem
includes the supporting software, the tool
chain and available reference designs.
A communications or network processor with
a fully programmable, multi-protocol enabled
communications engine can provide the conduit
to and from the packet network performing uplink
aggregation, downlink routing and protocol
interworking. Combine this with a powerful
communications processor on which the protocol
stacks and control functions can be hosted,
and the result is a single platform architecture
that is sufficiently flexible and scalable
to meet the demands of a converging network.
Media processing, including the conversion
between packet and TDM, involves voice encoding
or decoding using standard ITU codecs, such
as G.711, to provide toll-quality voice. Other
voice compression codecs including G.723, G.726,
and G.729A as well as the wireless codecs like
AMR, EVRC, SMV etc, can provide near toll-quality
voice by compressing the data, thus using network
bandwidth more efficiently. To improve the
user experience, echo cancellation is employed
to remove near end echo, introduced by mismatches
between hybrid impedances and low-cost end-user
equipment. Packet networks can introduce problems,
such as lost packets, packet errors, or even
delayed and out-of-order delivery. These can
all affect QoS, and an effective jitter buffer
needs to be implemented to deal with these
issues. Voice coding, echo cancellation and
jitter buffering is most suited to a programmable
DSP.
A single scalable system architecture
An integrated communications processor with
flexible protocol termination, hardware interfaces
and interworking features is a key component
of media gateways. Communications processors
that offer these capabilities in a cost-effective,
single-chip solution are becoming mandatory
in media gateway designs as component count,
board real estate and power consumption requirements
continue to be squeezed. Freescale's PowerQUICC
communications processors are key players but
a flexible front-end is only part of the solution.
A media processing engine
comprising an array or “farm” of programmable DSPs and a comprehensive
and open software portfolio complete the system
architecture. Using a single DSP core architecture
with scalable performance not only sustains
the system architecture but also the software
investment. An open software architecture that
accommodates integration of proprietary developer
or third-party software preserves the manufacturer's
differentiation and intellectual property assets;
with a single DSP architecture that scales
from low to high end processing, all this can
be achieved using a single code base across
a wide variety of markets and applications.
At the high-end, industry leading performance
is provided by Freescale's MSC8122 DSP which
l everages the computing power of four high-performance
cores based on StarCore® technology. Yielding
a performance equal to 2.0GHz and delivering
up to 8000 DSP MMACS, the quad-core MSC8122
offers highly optimized multi-channel processing
of voice, modem, fax and video compression
algorithms. Combining raw processing power
with system interconnect options including
10/100 Ethernet (MII, RMII and SMII), a 32-
or 64-bit host port for connection to host
processors and four multi-channel TDM ports,
enables developers to create next-generation
networking products that offer tremendous channel
densities while maintaining system flexibility,
scalability and upgradeability.
As network requirements continue to evolve
toward IP convergence, it makes sense to design
next-generation media gateways around scalable,
cost-effective and software compatible processing
platforms. Leveraging a single, flexible system
architecture with one software base, from low
to high density, will extend the equipment
manufacturer's investment and time-to-market
for today and tomorrow's network convergence
needs.