Programmability to prepare for the edgy access
GUEST OPINION: Intense competition and the need for more capacity push broadband service providers to deploy fiber and new carrier Ethernet technologies further out in the network. The access becomes more intelligent, and therefore more important for vendors and service providers alike.
By Per Lembre, Xelerated -- EDN, June 10, 2010
The broadband industry is under constant pressure. Intense competition and the need for more capacity push broadband service providers to deploy fiber and new Carrier Ethernet technologies further out in the network. The access becomes more intelligent, and therefore more important for vendors and service providers alike. System vendors' product portfolio strategies have to embrace the new capabilities while controlling cost for healthy product margins. As with all technology shifts, there will be winners, and there will be losers. What will be the secret recipe of success?
Just up until five years back, architecting broadband access networks was pretty straightforward. Cable players used DOCSIS to provide a best-effort Internet access service over the hybrid-fiber-coax infrastructure, while local copper loop players used DSL to reach the residential customers. Subscriber management and service provisioning was managed by a BRAS (broadband remote access server). This product was a special-purpose edge router mirroring the concept of dial-up, but instead of having modems on PCs calling up the router, the PCs were connected to a residential CPE, which featured an always-on session based on PPP or DHCP to the router. While every service provider used different back-end solutions, the network architecture was well accepted. In addition, fixed-mobile-convergence was mostly slideware, and nothing to consider when rolling out the broadband infrastructure.
Then two things happened. More intelligence was put into metro Ethernet routers, adding similar capabilities as found in the BRAS edge routers. And IPTV and other video services started to become popular. Every IPTV installation had its challenges, and the vendors came up with different network and system solutions to enhance the user experience. This resulted in different network designs. What used to be a common residential edge was often divided into two: one for video and another for Internet access.
Now, half a decade later, we are at the turning point for another shift in the broadband industry. We have learned our lessons from the first generations of IPTV systems. We have seen peer-to-peer technologies fundamentally change access traffic patterns. We have witnessed players such as Google/YouTube, Facebook, and Microsoft/MSN rise to the status of hyper-giants and change the way traffic is routed through the core. And we have started to build out a next-generation fiber access infrastructure.
The new fiber access infrastructure can be implemented using different first-mile access technologies: EPON/10GEPON, GPON/XGPON, active Ethernet, or FTTN/VDSL2+. What type of service awareness would we require from these platforms? They must manage replication of multicast traffic. They need to be user session aware and have IP functionality to enforce service policies. And they need to feature deep packet buffers for asymmetric shaping of user traffic. In short, as we build out the new fiber access infrastructure, we will demand more, not less. The access will become a key part of the distributed edge. The question is not if the access platforms should perform service edge functions, but which of them and to what extent.
The one thing that would hold this trend back is cost. There is always a premium price for added functionality. This was the argument BRAS vendors used to fight back the horizontal split between an IPTV edge and the Internet BRAS edge. Why add another type of edge when you can do everything in one box? The problem is that "god boxes" tend to be capex intensive and slow to adapt to industry changes. And if there is one thing we want from our new broadband network, it would be the ability to cope with the constant pressure for change.
To manage R&D in an every changing environment, you need the technology to be flexible. This is why choosing a programmable data plane for fiber access products makes a lot of sense. It allows for customization of the feature set and a unified approach to designing a range of platform types, be it Ethernet MDUs, PON OLTs, or mobile backhaul switching nodes. Programmability also enables upgrades of the data plane to support future standards or enhanced vendor interoperability.
Service providers are used to upgrade and customizing usage of the control and management planes to fit their environment. But changes to the data plane are far more costly. When the data plane can't support new features required for new services, they may be forced to be postponed until budget permits or until competition makes the change necessary. With programmable Ethernet switching, data planes can be upgraded and customized in a similar fashion as control and management planes. Programmability and capacity headroom holds a promise of product lifetime extension from three to five years to seven to nine years.
One thing is certain: Today's broadband services and standards will one day become legacy. That day is likely to be closer than we want. We should expect the next five broadband years to be fiercely competitive, pushing the limits to what access products can do.
It is always hard to prepare for unexpected, disruptive changes. Adding programmability of the data plane enables feature adaptation and differentiation through software-a safe choice for service providers and system vendors alike.
