Moving to an all-optic infrastructure enables ISPs to provide more services for less money.
By Bruce Wagner
Cogent Communications Inc.
Historically, bandwidth has been treated as a scarce resource by traditional, voice-based Internet service providers (ISPs) -- and high bandwidth costs and data traffic jams were the unfortunate consequences for the end user. High costs were a direct result of the ISP acting like a middleman, having to lease capacity from voice-based carriers and mark up the price to make a profit. In addition, ISPs were pressured to "overload" the network with too many users.
Bandwidth as a commodity is difficult to provision and expensive to support in existing legacy networks. A new breed of providers is building dedicated, all-optical, facilities-based data transport networks, bypassing the regional Bell operating companies (RBOCs) and competitive local exchange carriers (CLECs) and enabling the enterprise to exploit the Internet and extend corporate local area networks (LANs) at a cost per bit mile (the cost to transfer one bit, one mile) lower than any other offering.
The old world
Overlaying data traffic on networks originally designed for voice transmission made sense -- initially. There was no data network. Data services had to be carried over the voice network because only the voice network could give providers of data services -- either ISPs or voice carriers -- access to customers.
Of course, it was never a very efficient setup. Voice and data are fundamentally different, and one of the areas where the inefficiency of handling both types of traffic on the same network comes to light is protection.
Legacy circuit-switched networks that carry Internet Protocol (IP) traffic over an architecture originally designed for voice transmission include equipment and connections that data traffic alone would not require. Voice circuits demand dedicated capacity and uninterrupted service. When there's a line cut or equipment failure in the traditional, long-haul carrier network, SONET terminals linked by fiber pairs carry out signal restoration at Layer 2. The SONET terminal simply fails over to a redundant protection path within 50 msec to prevent signal loss. This is a "hard" failure mechanism.
The new model
Data can be treated quite differently than voice because it can arrive piecemeal, via packets. Optical networks, built from the ground up for data services, rely on "soft" failure mechanisms in both the long haul and metro. In the metro, for example, a bi-directional DWDM system assigns two paths per customer -- one traveling clockwise around the service provider's ring of multi-tenant buildings to be served, the other traveling counterclockwise. In the event of a signal loss, the Open Shortest Path First (OSPF) protocol calls for data to be reversed at Layer 3, with lost packets being redirected to available routers or along active fiber. Fail-over is completed within 1 sec, and that's sufficiently quick for data services.
Eliminating the need for SONET terminals, protection fiber paths, and the optoelectronics needed to light those fibers in a SONET infrastructure removes capital expenditures that otherwise would have to be offset with higher prices to customers. Building a network strictly for data can cut a service provider's capital costs by more than 95 percent.
So much more for so much less
In the data-only infrastructure, service performance is enhanced, in a variety of ways. First, without the SONET infrastructure, the number of optical-electrical-optical conversions is reduced, improving throughput.
Second, the new breed of all-optical ISPs does not have to oversubscribe. With specific levels of bandwidth assigned to individual customers in the provider's DWDM metro optical ring, one customer's access to the hub terabit router is never jeopardized by another's. Bandwidth connections are dedicated.
Third, because their network transport costs are so much lower, the all-optical ISPs can execute "cold-potato routing" -- keeping traffic within their data-only networks for as much as the transmission path as possible. Network hops are reduced and, therefore, there is less risk that data can be lost on other networks.
The efficiencies available in data-only networking enable these providers to offer all of the bandwidth a customer might ever need at a price lower than it is accustomed to paying today -- more for less.
The bottom line
With such price/performance ratios, the new all-optical ISPs are enabling enterprises to carry out business in more creative ways than ever before. Here are three examples:
- A multi-city retailer interconnects up to 20 locations at a time in interactive videoconferences for training. Because there is effectively no upper limit on bandwidth, compressing video is unnecessary. This enhances image quality and keeps the retailer's equipment cost low.
- Universities extend unlimited bandwidth to students and faculty. With no bandwidth cap, the true educational potential of networking can be realized, creating a new generation of bandwidth-hungry professionals.
- A national law firm convenes online with clients for negotiation, document reviews, and consultation in virtual case rooms. The firm is on its way to becoming a "paperless" workplace, as all document storage is being transferred to the network.
The new all-optical ISPs represent a compelling option also for providers of advanced IP applications such as streaming video. The all-optical ISPs give these application service providers (ASPs) all of the bandwidth they need to properly serve customers and access to a more comprehensive customer-care arm than they could afford on their own -- all at prices better than the voice carriers can offer.
The emerging, all-optical ISPs providers enable enterprises, ASPs, and traditional ISPs to simultaneously cut costs and remain viable while passing along unprecedented service performance to users or subscribers. With so much bandwidth affordably at their disposal, customers find that network capacity is determined only by their imaginations.
Bruce Wagner is vice president, sales, at Cogent Communications Inc.