Freescale Semiconductor, Inc.
Network Processors:
A Definition and Comparison
A growing class of communications silicon, the Network Processor, promises to
revolutionize how networking vendors architect, develop, and support their products.
Network Processors deliver dramatic improvements in time-to-market, product lifetime,
and system capabilities. This paper examines the benefits of Network Processors in
comparison to other networking silicon offerings.
A Brief History of Network Product Design
The design of networking products has undergone continuous evolution as the speed and
functionality of local and wide-area networks have grown. In the early days of
packet-based networking, networking devices (such as bridges and routers) were built
with a combination of general purpose CPUs, discrete logic, and ASSPs (Application
Specific Standard Products), including interface controllers and transceivers. The
software-based nature of these devices was key to adapting to new protocol standards
and the additional functionality required by networks, such as the early Internet. Although
these designs were large, complex, and comparatively slow, they met the needs of these
early networks (generally comprised of a few Ethernet or Token Ring connections and
slow (56kbps) wide-area links).
Over time, as network interface speeds and densities increased, the performance of
general-purpose processors fell short of what was needed. This led network vendors to
develop simpler, fixed-function devices (such as Layer 2 Ethernet switches) that could be
built with ASICs (Application Specific Integrated Circuits). These devices traded-off the
programmability of software-based designs for hardware-based speed. As ASIC
technology progressed (and vendors invested heavily in hardware-oriented design
teams), more and more functionality was incorporated into the hardware. This was
enabled in part by protocol consolidation around IP and Ethernet as the dominant
enterprise network technology, which reduced the need for product flexibility.
By: David Husak
C-Port Founder and
Chief Technical Officer
The relative simplification of network products has allowed merchant silicon vendors to
“commoditize” some networking segments through specific chipsets, such as Layer 2
Ethernet “switch-on-a-chip” products. Some of these solutions offer significant
functionality within a narrow range of applications, such as ATM switching or basic
Ethernet/IP switching. However, network vendors seeking clear product differentiation
still required long and risky internal ASIC development programs.
Today’s Network System Development Challenge
“It’s the software, stupid!”
Vint Cerf, Senior VP for Internet Architecture and Technology MCI WorldCom, and “Father of the Internet”
ComSec Seminar, January 1999
Today, the convergence of public voice and data networks is speeding up the pace of
change in the communications industry. This is leading to increased time-to-market
pressure and shorter product lifecycles — just when product development cycles are
growing due to complex ASIC designs and associated software re-designs.
Although IP is emerging as the dominant protocol, newly defined IP capabilities, such as
Quality of Service (QoS) and Multiprotocol Label Switching (MPLS), require vendors to
continually support new applications. In addition, the number of different interface types,
White
Paper
For More Information On This Product,
Go to: www.freescale.com