The latest market projection reports are in and all indicators suggest that global Internet traffic will sustain a 40-50 percent year-over-year growth for at least the next 5 years.¹ When correlating this statistic with individual service provider’s broadband networks, the more general experience has been a doubling of packet traffic every 12 to 24 months. Over the top (OTT) video is a major driver of this growth, and this application is still in its early days as a mainstream service. Given the industry’s relative inability to predict the introduction of new traffic generating applications and OTT services, it’s difficult to judge if those projections will prove to be aggressive or conservative on an individual service provider’s network scale.

On the mid-point of that spectrum, this rate of growth represents a near 10x increase in traffic every five years. That rate of growth seems hard to imagine; however when looking back over a 15 year period most networks have experienced a more than 1000x increase in packet traffic. To support this growth, access aggregation routes of the networks were upgraded from single or multiples of T1 transport to OC-12, to OC-48 or Gigabit Ethernet (GbE), and eventually to OC-192 or 10 Gigabit Ethernet (10 GbE).

However, for many service providers supporting the next phase of growth beyond 10G is challenging in multiple dimensions, particularly for access aggregation and collector rings. This challenge is often referred to as the 10.1 barrier. The purpose of this article is to explore various considerations and emerging options for cost-effectively scaling networks and breaking through this 10.1 barrier.

Caching Will Help -- But Will Not Break Through the 10.1 Barrier
Sandvine’s Spring 2011² report indicates that real-time entertainment traffic (e.g., OTT) accounts for 49.2 percent of North America’s peak period fixed broadband access traffic, and is projected to reach 55-60 percent of traffic by the end of 2011. As the video component of broadband traffic scales, there are opportunities to minimize transport bandwidth on select wide area network routes by caching the most frequently viewed video content at data centers with servers which are located closer to the subscriber base. This reduces the duplicity of transport traffic between the upstream Internet content source by caching a copy of the most popular videos at a server in or near the broadband service provider’s network.

Unfortunately, the caching of video content comes with costs, complexities, and constraints that limit the primary traffic reduction benefits to more regional and possibly high-density metro networks. This is merely a function of statistics, where the law of large numbers enables caching to provide high statistical gains when content is shared over large user populations. Caching diminishes in value as the user population supported by caching servers decreases. Consequently, caching will not provide meaningful relief for the collector rings and less densely populated metro rings in the vast majority of markets. Therefore, alternative approaches will be required to break through the 10.1 barrier.

Understanding Where the Barrier Is and Is Not
When trying to solve a problem it’s useful to characterize and decompose the components of the challenge. In the case of exponentially scaling network traffic there are multiple components of the network to consider. Starting from the customer edge (The first consideration will be the broadband access network.), followed by the aggregation of transport networks, which ultimately connect broadband customers to content sources.

The good news for most service providers is that the exponential increase in packet traffic does not directly correlate with an exponential increase in access line rates. Therefore, existing xDSL (ADSL/VDSL) and FTTP (GPON/Active Ethernet) broadband access platforms will likely be able to support increasing usage rates for several years. However, important changes in the configuration of these devices may be required to deal with the implications of exponentially scaling OTT services.

Historically, the planning assumption was that IPTV would dominate the bulk of broadband access services and Internet content was a relative minority. IPTV is multi-cast in nature, so it was reasonable to integrate transport into the access platforms because the majority of traffic was used by all broadband access nodes on a ring. In this multi-cast centric scenario, the broadband access systems were only moderately loaded and the burden of passing a relative minority component of unicast Internet access traffic across all intermediate nodes was acceptable.

With the popularity of OTT video, that model has changed. Every OTT viewing of a popular YouTube video or the latest episode of TV programs such as Modern Family drives a unique unicast video stream from the Internet or upstream caching server to each subscriber. As a result, integrating transport into the access platforms can result in a situation where 80 percent or more of the traffic simply passes through an access node on the ring, and only 20 percent or less is used by locally connected broadband subscribers. As certain thresholds are reached, this architectural model will add considerable cost and processing burdens to the access platforms with no operational advantage. Alternatively, these broadband access systems can scale more cost effectively, and operate better by simply offloading the transit traffic from each broadband access system onto an independent transport network.

Transport: Many Options for The Metro, Fewer Choices for The Edge
In recent years there have been numerous enhancements in the area of packet-optical transport systems (P-OTS) to address scaling transport requirements. These systems provide the option to collapse transport grade Ethernet switching systems, SONET multi-service provisioning platforms, and dense wave division multiplexing (DWDM) systems into highly integrated platforms supporting exponential scale with simplified operations and substantially reduced cost. The ability of these systems to support 10 Gbps services on dedicated fibers, or combine multiple 10 Gbps services over common fibers provides critical relief for metro and regional bottlenecks.

In most cases those transport systems were designed for larger central offices and other locations where they can be scaled across metro- and region-wide services. The real challenge is how to cost-effectively support this exponential scale out to the broadband access edge, providing aggregation and transport collector rings to offload deeply deployed xDSL, FTTP, and mobile tower locations.

The complex nature of access networks accounts for 5 specific challenges that contribute to the 10.1 barrier.

Challenge 1
Performance
When considering the performance requirements of edge aggregation and transport systems it’s important to understand that these services are no longer limited to best-effort Internet access applications. As OTT video matures and transforms from a YouTube experience on a teenager’s 13-inch PC to a shared entertainment experience on a 42-inch television in the family room, the tolerance for low resolution, jittery images, and download pauses will quickly pass.

Service providers will need to offer premium services with greater performance than conventional carrier grade systems. High-performing, connection-oriented Ethernet transport systems are now available from multiple vendors with the ability to partition transport network capacity in a more structured and deterministic manner. Powerful management systems must provide the ability to proactively and continuously monitor traffic conditions and traffic patterns to make sure that networks are traffic-engineered to assure high-performance of guaranteed SLA services, while providing the best available best-effort services within remaining network resources.

Challenge 2
Scale
Continued advancements in Ethernet switching technologies enable next-generation platforms to aggregate services into one or multiple 10 GbE transport services. The challenge is adding these transport capacity expansions incrementally and paying only for the capacity you need. Therefore, modular systems which provide incremental capacity with modest incremental investments will be an important consideration.

The greater challenge for most access edge aggregation systems is the lack of available fibers to drive these incremental transport services. In many cases, existing fibers are already carrying other backhaul services. Deploying full blown DWDM systems to scale one fiber to support two or even three 10GbE services was historically too costly. In addition, these systems were generally too large and not industrial temperate (i-Temp) rated for deployment in OSP cabinets.

Fortunately, new purpose-built edge aggregation and transport systems now offer the ability to overlay one or multiple 10 GbE transport service on top of existing SONET or Ethernet transport services without costly dedicated transponders and high-capacity DWDM systems. New edge optimized aggregation systems provide the ability to software reconfigure 10 GbE transports into G.709 OTN (Optical Transport Network) digital wrappers, which provide forward error correction (FEC) and the operations, administrative, and management(OAM) functions to support DWDM without additional transponder modules. In addition, new purpose-built micro DWDM modules provide the ability to overlay one or two 10 GbE/OTN services on top of the existing SONET or Ethernet services operating in the 1310 nm spectrum, with full management visibility and control.

Finally, when considering the prospect of a sustained doubling of traffic capacity every 18 months or so, support for two or three 10 GbE services may not be sufficient over time. Platforms that are deployed today should be designed to support the future addition of 100 GbE service modules, which will provide a roadmap for sustained scalability with minimum cost and upgrade effort.

Challenge 3
OSP Ready Form Factor
One of the most rigorous challenges of deploying edge aggregation systems is the availability of scalable, small form factor transport systems that can fit in space constrained cabinets and operate in non-temperature controlled environments. Once again, new edge aggregation and transport products are now available and are designed to support the above requirements of high-performance Ethernet aggregation, with support for multiple 10 GbE transport services and optional micro DWDM functions in compact 2RU chassis, which fit in space-constrained OSP cabinets and indoor racks. These edge optimized transport systems must use i-temp rated components, supporting temperatures ranging from -40ºC to +65ºC. These critical set of edge aggregation platform attributes enable practical snap-in solutions for scaling the access aggregation edge and breaking the 10.1 barrier.

Challenge 4
Operational Simplicity
While performance, scale, and form factors are all essential considerations, these systems must also be easy to provision and operate. The continued maturing of the integrated P-OTS market has yielded dramatic improvements in operational simplicity. The combination of advanced Ethernet OAM and optical OAM with G.709 OTN gives service providers full visibility and control, with simple point-and-click provisioning. The result is a network with the predictability and performance of a SONET-based transport network, with the flexibility and scale of an Ethernet and DWDM network.

Challenge 5
Low TCO
The continued advancements of key enabling technologies makes it viable to offload scaling OTT and IPTV services from broadband access systems for greater scalability and performance. The innovation of new purpose-built edge aggregation platforms has yielded small form factor platforms that are designed to snap into existing space-constrained OSP cabinets and racks to provide robust and reliable services, with optional DWDM. The ability to snap-in and scale networks within existing network constraints is critical to achieving low capital costs.

Increasingly, these edge optimized aggregation and transport systems are part of a broader P-OTS market, which is exploding in growth and as a result is driving tremendous cost reductions in cards and components across the network. The commonality of hardware and operational simplicity from the access edge to the network core provides even greater capital and operational cost benefits to optimize the TCO.

One Barrier Down, More Opportunities to Follow
After examining the challenges and consideration for scaling broadband access networks beyond 10 Gbps, we can appreciate the complexity of the 10.1 barrier. By exploiting continued advancements in enabling technologies, and considering strategies for offloading transport from broadband access systems we can see a clear path for scaling networks and breaking down this network barrier. More importantly, by considering the broader opportunities and changing market requirements, service providers can deploy solutions that break down these barriers and open the gateway to expanding services and incremental revenues.

Endnotes
1. Cisco Visual Networking Index: Forecast and Methodology, 2010-2015  http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/...
2. Sandvine Global Internet Phenomena Report: Spring 2011  http://www.sandvine.com/news/global_broadband_trends.asp

Frank Wiener is the VP of marketing for Cyan. Cyan is a provider of innovative packet-optical transport systems, software and software as a service (SaaS) systems, and professional services. For more information, visit Cyan at www.cyaninc.com.

What is your experience with this? Tell your fellow readers now!