Access Infrastructure Market 

Access networks have traditionally required a separate infrastructure for each physical media and each type of service provided. Although the Internet has significantly transformed data networking, carriers still derive much of their revenues from voice.

The prior method for balancing demand for data and voice was to overlay broadband data networking equipment on existing voice networks. Carrier voice and data infrastructures were essentially separate and non-integrated, even when the media itself, such as DSL and SONET, carries multiplexed voice and data. This not only increases expenditure, but also complicates customer service, management, and interoperability.

 The next generation of access networks will converge voice, video, data, and wireless infrastructures. It will combine soft-switch architectures, media gateways, broadband multi-service access platforms, wireless infrastructure base stations and radio network controllers, and intelligent access devices that support services like VoIP and VPNs. One immediate benefit of the converged network will be lower costs due to reduced equipment expenditures. Converged networks will also be easier to operate, manage, and provision. The cost-efficiency of the Internet is forcing the converged network to migrate toward IP packets once quality-of-service issues are solved in IP.

 Although the benefits of convergence are plain, access networks today are still far from this goal because they are highly segmented. The diverse infrastructure built around competing media, such as copper, coax, wireless, and fiber, competing formats, such as TDM, cells, and packets, and competing regional and regulatory issues clearly shows the segmentation within access networks.

 The competing protocol formats can be classified into the following major types of "protocol handling": 

  • IP: Routing, Forwarding, Tunneling, Classification, and Quality-of-Service (QoS)
  • L2 WAN Based: PPP, ML-PPP, HDLC, Frame Relay, and MPLS
  • ATM: Switching, Segmentation and Reassembly, Inverse Multiplexing
  • Ethernet: Termination, Tunneling, VLAN tagging and Switching
  • TDM: Digital Voice, Time Slot Assignment and Mapping, Cross Connects, and Circuit Emulation over Packet Networks
  • Wire-speed interworking support of every combination!
    •  

The market size of all semiconductors used for protocol handling in the existing access infrastructure is expected to be be approximately $3 billion in 2008 (Gartner Dataquest), with growth in emerging segments such as broadband access, voice-over-packet equipment, wireless infrastructure, and the multi-service spaces.

The complexity of the protocol handling situation previously caused some vendors to advocate immediate migration to IP packets; however, there are several reasons why it is now clear the process will be gradual in existing infrastructures and require multi-service support, such as ATM and Frame Relay. First, regional and regulatory concerns limit the migration to occur per a number of media-geography combinations. Established carriers own over 90% of the local exchange market and have large capital investments that must be depreciated-thus creating a significant need to interwork with branches of the network relying on legacy technology. Second, although IP has the advantage of IETF standards process, there are still outstanding QoS issues where IP is concerned. For instance:

  • IP requires a multiplexing solution that allows multiple small voice packets to be efficiently carried
  • IP must offer a priority mechanism that can absolutely guarantee voice latency
  • IP networks need a scalable control protocol that allows traffic to be engineered in a way that can guarantee bandwidth is available for voice services for the duration of each call


Wintegra believes these IP QoS problems are beginning to be solved in commercial IP networks, and will be commonly adopted in the future, and that MPLS will play an important role in this process. In the meantime, IP services can be provided over an ATM link layer, given that ATM allows reliable QoS techniques to mix voice and data properly. However, the ability to offer triple and quadruple play IP-based services to customers, even today, requires intelligent nodes in the access network that can rebuild packets from ATM cells, as well as perform full IP-based handling and interworking.

What is needed in access networking is the ability to mix and match TDM, Ethernet and OC-x at layer 1; ATM, Ethernet framing and Frame Relay at layer 2; MPLS at layer 2.5; and IP at layer 3 with various forms of interworking in any combination. Wintegra believes that this functionality can offer access networking engineers significant new degrees of freedom, innovation, and cost reduction in their next-generation access infrastructure equipment.

Beyond the basic protocol choices, access network engineers make many other demands on protocol handling semiconductor components. These demands include needs for:

  •  Multi-protocol handling on a single chip
  • Multiple and varied broadband interfaces on-chip
  • Versatile chips that can be used in many different places in access equipment
  • High levels of appropriate system integration balanced against cost
  • Ability to handle voice/data convergence today, ready for video, and containing uncompromising levels of QoS
  • Best-of-breed feature-sets. Access network engineers cannot afford to move backward in product feature evolution just to use a highly integrated solution
  • Pre-programmed and validated. Interworking functionality must be developed, tested, and delivered to customers as a working solution, not given as "reference code"
  • Customer ability to add value both internally with software and externally with software and hardware
  • Programming simplicity. Data path interworking functions are difficult enough to develop without requiring microcode, assembler code, or multi-processor partitioning paradigms
  • Soft Upgradeable and Flexible. To evolve with emerging requirements and standards and to allow bugs to be fixed without semiconductor fabrication cycles
  • Scalable in performance by cascading additional devices
  • Scalable in architecture such that code investments may be preserved in future generations
  • Debug-ability. Allowing remote visibility into on-chip behavior


Building on the above demands, the following benefits will flow to the access network equipment engineers:

  •  Compelling in Cost. Not just in terms of dollars/Mbps vs. other solutions, but also in absolute cost.
  • Time-to-Market reduction. Customers would like to deploy a small team of engineers to build what previously took two or three teams, and to do it in 5 months rather than 12.
  • Power Reduction. Protocol handling sub functions in access equipment should preferably exist in the 2 Watts to 5 Watts range, given that many access infrastructure cards have total power budgets as low as 7 Watts to 10 Watts.
  • Space Reduction. This is critical to the extent that it allows new types of access equipment to be built such as single board base-stations. It is also very important in certain high-density applications such as carrier class Voice-over-Packet systems.

What kind of component can meet these demands? General microprocessors are certainly standard in the control path, but are not appropriate for the data path. Communication processors - the innovation that followed the microprocessor - can be highly integrated with many layer 2 communications peripherals, but are designed for lower-end markets like customer premise equipment. And while network processors provide higher-end IP performance of data path functions, they do not typically meet the cost, power, space, and protocol requirements of the access infrastructure.

The Wintegra WinPath® product family brings the affordability, flexibility, and performance needed to build next-generation access networks. Using WinPath, customers can engineer a whole new approach to protocol handling, integrating a broad range of network solutions into a unified design. Wintegra introduced the WinPath architecture in November 2001, has been shipping into production networks since 4Q02 and has won a large number of designs. For more information on WinPath, please visit our products page.