Building blocks: sound architecture is key in extending Ethernet access in today's network - Ethernet Magic

Telecommunications Americas,  March 15, 2003  by Rob Cain

• E-mail
• Print
• Link

Metro Ethernet networks have been in the headlines a lot lately. As various organizations forge ahead on standards, a handful of service providers like AT&T, SBC and BellSouth have begun rolling out Gigabit Ethernet offerings. Ethernet access is obviously where service providers see themselves and their customers migrating.

To really extend Fast Ethernet access and make native LAN services widely available, however, service providers face a number of challenges. Most enterprise data services are still burdened with telco-like access, which limits the flexibility, scalability and performance of wide area networks. To get Ethernet access, critics have noted, customers need a fiber drop, or they're out of luck. Then there is the cost of adding equipment to the customer premises and network plant, which may not justify the relatively small number of customers who will first migrate to native LAN.

But fulfilling the current needs of enterprise customers shouldn't mean service providers have to offer 10 GigE right out of the chute. Service providers' goals should be to get native LAN services to customers who need it now, priming the pump for the future deployment of 10 GigE and 40 GigE services that will support global enterprise networks. The good news is that carriers now have the opportunity to offer Ethernet access, use the fiber they already have in the ground and avoid economy of scale obstacles. Today's metro Ethernet or native LAN services can be achieved with less need for training, fewer installation and maintenance errors, and operability that meets telco standards. And most importantly, these services can be delivered via 10/100 Base-T copper connections rather than traditional channelized T1 lines. Service providers have the ability to give their customers a unified network that offers all the benefits of continuous, end-to-end IP.

These capabilities can be attributed more to the evolution of technique than technology--and a resolution in the ongoing battle of "Bellheads" and "Netheads."

Beyond Direct Connect

The reason to strive for native LAN services is not just the speed and robust connectivity that gets lost in translation when data is carried over traditional T1 or frame relay. Enterprise data services today feature the hard-wired or direct-connected cabling typical of data networks, resulting in a lot of inefficiency around installation, management and maintenance. They can be messy; common problems include down or intermittent circuits, redundant active ports and wasted floor space. It's difficult to get these services--where active network elements like servers and routers function in a one-to-one configuration--to offer the reliability and service level standards vital to telco networks.

After years of cultural conflict, data and telecom networking engineers finally are working together more closely. The essential development is that service providers' data people, or "Netheads," are finally getting beyond direct connect cabling solutions and adopting more traditional telco best practices, paving the way to easier network management, greater operability and high service standards needed to offer native LAN services.

The differences in operating paradigms are by now painfully familiar. Telco networks built to last and tested to the nth degree work in a centralized world engineered for scalability, reliability and, most importantly for the purposes of native LAN, manageability. Telcos operate durable equipment where the return on investment is measured in decades. In the IP world, the cost per megabit is much lower than that for voice, servers and routers are a dime a dozen and cost is measured in months, if not less. For service providers, the trick is to put together a native LAN service that offers traditional data economics with traditional telco operating standards. That means, among other things, eliminating traditional "Nethead" practices that leave wiring closets looking like spaghetti, and providing centralized connectivity for essential network elements. For instance, for metro Ethernet boxes situated with DLC (digital loop carrier) equipment in the outside plant, network engineers will centralize cable managemen t and configuration practices, effectively taking traditional CO procedures outside the CO. While the architecture is distributed, the connectivity is centralized. Network topology still looks the same, but managing and maintaining it becomes a whole lot easier. (The actual need for that box is another question. That's addressed in the next section.)

The "Netheads" are borrowing three major pieces of traditional telco network architecture.

* Centralized configuration and distribution. Configuring and reconfiguring every single circuit at a network element is unthinkable. Batch jobs are the ideal. Telco network engineers, or "Bellheads," would much rather go to a single patch panel or DSX lineup, and they can do the same with active network elements like routers and servers. Service providers are introducing centralized termination points--in the Go, on the customer premises or close to it--that give them access to aggregate boxes and cables. Even changes as simple as clear labeling have a profound effect on manpower efficiency. In addition, these distribution points give technicians the ability to effectively segment the network when trying to isolate failing equipment or faulty wiring.