Has the Age of Distributed Ether-Networking Arrived?

By Accedian Networks' Scott Sumner

Service providers are all working through their transition to packet-centric networking, never seen as an easy task from the beginning, as this is one of those problems with many solutions, routes and pitfalls. A mathematician would call this a poorly posed, non-deterministic problem – the kind of thing they leave to philosophers to debate.

But engineers and network operators are a bit more practical; talk to any provider and you’ll see that they are forging ahead. Despite the challenge, it’s pretty clear that operators need to migrate to a cost-efficient, scalable, packet-based infrastructure to cut the operational costs and complexity of delivering high-performance IP communications and real-time applications. And customers are waiting impatiently, not a small consideration. As business interconnectivity and communications continue to require more bandwidth, performance and cost-efficiency over large geographic areas, they are demanding Ethernet connectivity that rivals private line performance at much faster rates, and demanding the provider demonstrate that performance is up to snuff.

No two providers can travel the same path to packet-optimized networks and the delivery of Carrier Ethernet services, as their existing infrastructure and end-customer needs provide a unique spectrum of variables with which to contend. This diversity of constraints, combined with a constantly evolving fabric of interconnected carriers and networks, also makes defining best practices and standards a slow, complex process. This means there are many approaches to delivering Ethernet services, depending on the provider’s networking strategy and starting point.

At a high level, the options available to service providers draw parallels to developing networked computing applications. Designers need to decide how processing and storage should be distributed – from traditional client-server applications, where most of the intelligence, storage and processing resides on a centralized server (in the core of the network, so to speak); to software as a service (SaaS) applications like Google Apps, where processing is split between the client and the server, and storage can be in either location; to the extreme of peer-to-peer or cloud computing where, for all intents and purposes, all processing and storage reside on client machines.

Providers face a similar range of options in deploying Carrier Ethernet services. They need to address the questions: Where should the intelligence of the service reside? Where and how should Ethernet and IP services be defined, managed and deployed? I’m referring to the features and functionality that make Ethernet carrier-grade, flexible and easy to manage – the basic function of creating or mapping an Ethernet service (provisioning an E-Line or E-LAN); conditioning traffic (filtering, rate limiting and shaping) to preserve access, core network performance and per-service QoS; establishing Ethernet OAM; and the monitoring and management of each service to ensure SLA requirements are met.

Just like distributed and networked computing, the balance of networking intelligence can be highly centralized in ever-more sophisticated “god-box” core-routers and multi-functional access platforms, or distributed and applied at the very edge of the network – at the service demarcation point itself (customer site, carrier hand-off point or cell tower). Enabling this latter approach has been the recent evolution of Ethernet demarcation units to fully functional, carrier-grade in-line “micro network elements”, with a range of features and performance that rivals (and in many cases exceeds) the capabilities of the latest generation of routers and Carrier Ethernet switches. Just imagine distributing the critical functions of service mapping, OAM, traffic filtering, bandwidth policing, traffic shaping and monitoring functionality to each customer site, eliminating the need to upgrade access elements, while providing highly granular control over every service from end-to-end.

When I explain this model to network architects and operations staff struggling with their Carrier Ethernet initiatives, light bulbs go off ─ they really like the idea. It just doesn’t make economic sense for operators to forklift out their TDM core and access networks to create a pure layer 2 network, even if it somehow could be easily accomplished. The reality of mergers and acquisitions that have been the staple of telecom news for the past several years is that networks are a kludge of multiple technologies, carriers and network elements from a variety of vendors, spelling out migraine-level interoperability issues when you get to deploying Ethernet.

So distributing the intelligence and Carrier Ethernet provisioning, management and monitoring to little boxes at each service endpoint makes a lot of sense. You get transparent, SLA-grade Ethernet services over any network, have all the control and management promised by the god-boxes, and get to scale units into your network in step with service and revenue growth. It also allows providers to offer a fully managed service over wholesale networks and leased-access links since they gain full visibility and control of their service directly from the customer site.

Has the age of distributed Carrier Ethernet networking arrived? Some forward-thinking operators think so, and they seem pretty happy with their choice – problem solved.

Scott Sumner is vice president of marketing at Accedian Networks, which provides Packet Performance Assurance™ solutions that enable carrier-grade, packet-based applications over wireless and wireline networks.

Share this article: Email, Slashdot, Digg, Del.icio.us, Yahoo!MyWeb, Windows Live Favorites, Furl
Add this article feed to: RSS, My Yahoo, Newsgator, Bloglines