From Telecommunications Online, this is perhaps the finest ATM article I've seen yet:
telecoms-mag.com
<<<
Switching ATM in the Service Provider Market The improved business models and opportunities stemming from ATM backbone adoption will change the dynamics of telecom to a degree that could equal the changes wrought by divestiture in the early 1980s.
David Nelsen
March 1998
ATM was developed in the mid-1980s to be an integrating carrier backbone technology, and its ascendance in that role has begun. Today carriers use an ATM infrastructure that will underpin every service they offer, their business goals following ATM to a dense core with multiservice flexibility at the edge.
This transition is driven and cost-justified partly by the demand for broadband service interfaces, but primarily by the immediate need to scale frame relay and Internet backbones. ATM and frame relay have worked well together for several years, but their conjunction is now imperative as frame relay services outgrow frame switching equipment capacity. Furthermore, ATM is becoming the Internet backbone. Sixteen months ago, there were no ATM switches deployed at the core of Internet service provider (ISP) networks. Today more than 67 percent of U.S. Internet traffic is switched by ATM at some point in its travels.
At an estimated 25 percent to 40 percent, Internet MCI accounts for the largest slice of the domestic Internet backbone and uses ATM switches at the core. UUNet is the largest dial-access ISP and its network also is built on an ATM core. Many other ISPs, including America Online and the Microsoft Network, also use ATM in the core of their service infrastructures. ATM vendors are active in the market. Many dozens of ATM switches--each handling gigabits of traffic--are installed today.
ATM is not usurping established services; instead existing services, e.g., frame relay, IP, LAN, and ADSL, are driving the adoption of ATM. Unified network visions of ATM have made handsome slides for some time, but suddenly this technology is satisfying carrier needs with an immediacy that is just now being recognized.
ATM Drivers
Frame relay and Internet connectivity have turned out to be key factors in initial ATM adoption. These service providers are often backlogged with orders for new service. Internet traffic is growing at a rate of 10 percent to 15 percent monthly, and frame relay services carry double the traffic that they did a year ago. But frame relay and IP networks don't scale up efficiently, and their continued growth can't be sustained by deploying more of the same equipment.
Growing packet and frame-switching networks suffer two severe problems: nonlinear increase in network complexity and a significant decrease in network performance. These Internet and frame relay networks have become an increasingly difficult-to-manage "spaghetti" of T1 and T3 trunks where full interconnection is impossible. Congestion, packet loss, and delays are increasing while throughput is decreasing. With an average of 16 hops now required to traverse the Internet, a larger fraction of any given node's processing capacity is devoted to transit traffic. Each router hop entails a processing delay that adds to the frustrating Web crawl that leads to complaints from users. Packet processing is also expensive. Router and frame relay nodes cost 10 times more per megabit of capacity than newer ATM systems. Conventional expansion options are no longer necessarily economical. In 1996, the largest service providers reached a point where it was simply not cost-effective to grow their networks larger. Routed IP networks had been built with point-to-point circuits. Adding more routers meant adding more physical links to other routers in the network. The combined cost of switches, ports, circuits, network management, and service administration was rising out of proportion to revenue.
ATM backbones solve both problems--unmanageable network complexity and unacceptable network performance. By installing high-capacity ATM switches at the core of a router net, Internet providers can eliminate four to eight router hops across their backbone. Performance improves as expensive packet-level processing is reduced. Routers feeding the edge of such backbones can access any other edge by a single hop through the ATM cloud, eliminating costly point-to-point circuit provisioning.
ATM does the same for frame relay networks. With ATM trunk speeds currently operating at up to OC-12c, or 622 Mbps, a single ATM switch port replaces nearly 14 frame relay T3 ports or hundreds of T1 ports. ATM trunking will soon support OC-48c, equivalent to 55 T3s. Fewer frame relay switches are required as they take their place at the network edge, and network provisioning and management become significantly easier.
Critics have long faulted ATM's 10-percent to 20-percent overhead as wasteful, but this overhead is, in fact, easily recouped by the streamlined performance that an ATM backbone achieves. Carriers have found up to 50-percent reductions in their total expenses by scaling frame relay and IP networks with an ATM backbone.
For example, Internet MCI previously used a router backbone meshed with T3 circuits. Installing higher speed interfaces on their core routers was a potential stop-gap measure for meeting their expansion needs, but it would buy less than a year of relief. The network would still be using point-to-point circuits that would ultimately result in a bigger but overloaded, expensive router mesh.
In the summer of 1996, Internet MCI deployed eight ATM switches with OC-12c trunks for interconnection. Existing routers were moved to the edge of this ATM core with their OC-3c uplinks. Today the network can be readily provisioned to support increased service and will likely be scaled up to an OC-48c backbone this year. Overall capacity is up by a factor of 10, the network is easier to manage with fewer discrete links, and performance has dramatically improved. AT&T Worldnet, UUNet, America Online, and the Microsoft Network (MSN) have all begun similar ATM migrations.
Decreased operating expenses are the second major ATM driver. Wide area voice/data networking that today costs $10,000 per month can be reduced to $1000 per month through ATM integration.
Present utilization of all the separate, overlaid facilities that carriers operate (e.g., voice network, frame relay, private leased lines, Internet, and ISDN) is dismally low--around 10 percent to 20 percent--because bandwidth is channelized. This means that the physical bandwidth has been divided into separate pathways, each designated for a specific user, and there is no way for a user to access additional bandwidth that may be available on another user's channel. Individual elements do experience high utilization periods, but the whole is less than the sum of its parts because the resources aren't pooled.
"Non-coincident utilization" is the phrase used to describe this condition. Each separate network has to be sized to handle peak traffic load, but rarely do the different services or customers peak at the same time. This effect is most significant in private line services, where bandwidth is rigidly partitioned. This provides the major economic driver for statistically multiplexed services such as frame relay and ATM. Supporting all facilities with a common, unchannelized infrastructure efficiently pools reserve capacity and radically trims operating costs.
ATM was invented for consolidation of network services, and carriers are finding it works. By folding all of today's voice/data services and private lines onto a single ATM network, carriers expect to achieve total cost savings of a full order of magnitude. ATM's inherent attributes--flexibility, scaleability, and granularity--will define the competitive edge of future carrier services. Stiffening competition among service providers is already spurring them to adopt new technologies that can be folded into the emerging ATM infrastructure. This trend will further develop into expanded business opportunities.
The 1996 Telecom Reform Act presents an immediate scenario. Carriers are suddenly free to compete for business outside of their regions, gaining access to customer premises anywhere through obliging local exchange carriers. The most efficient model for offering new services out of region is to lease local access at remote POPs and then use ATM to back-haul the traffic onto the carrier's own backbone.
Asymmetrical digital subscriber line (ADSL) is a fine example of new technology melding with a new business opportunity. ADSL could potentially be extended to the 80 percent of U.S. homes and businesses that are within 18,000 feet of a central office, carrying 6 Mbps to the subscriber and 1.5 Mbps back out to the network. Telecom providers and their vendors generally agree that the resulting volume of traffic can only be supported by large ATM networks; no other technology vies with ATM for the ADSL backbone role.
ADSL is also an example of the order of magnitude cost savings that carriers foresee. A T1 currently costs almost $1000 per month (a distance-sensitive average). ADSL with an ATM backbone provides the equivalent outbound bandwidth and four times more inbound and will cost just $50 to $150 per month, roughly 10 percent the cost of a T1.
The next major industry trend is carriers' aggressive focus on managed network services (MNS). Companies are eager to shed the administrative and capital costs of running their own telecom operations, and carriers are eager to handle that business for them.
MNS will include native ATM services, but initially they will be comprised primarily of non-ATM interfaces integrated through ATM service extension links. Existing private line and voice traffic, ISDN, Ethernet and token ring LAN connections, frame relay, and ADSL data connections may simply be handed to carriers who will manage the adaptation of this traffic into their ATM clouds. Managed services will extend beyond today's typical premises "demarc" into wiring closets and probably even deeper into enterprise networks.
Carriers' ATM infrastructures make such a business shift plausible because ATM is an integrating multiservice technology that blurs the distinction between LAN and WAN. Customers would not have to replace their installed T1 multiplexers, PBXs, or frame relay routers. MNS could easily support legacy devices, and customers would see their telecom costs drop substantially while they focus on their own core business competencies.
As a result of discussions with industry analysts, customers, and service providers, we believe that, in the future, 40 percent to 50 percent of all enterprise networking equipment will be acquired from carriers through MNS and resale arrangements. Outsourcing and systems integration will allow businesses to focus on their core competencies. Carriers will become the preferred sales channels, and the influence of today's one-stop shopping mergers and acquisitions will wane.
The compelling economics are inducing carriers to work toward the long-held vision of a single ATM infrastructure supporting all of their service offerings. Driving ATM's immediate success is the pressing need to scale up current frame relay and IP services. Instant cost justification and performance improvement put ATM in place and prepare carriers for the next phase: voice data integration. Providers are pressing for products that will let them integrate voice and private-line circuits over their ATM backbones. There is a business case for even the most basic voice over ATM technology, using straight circuit emulation with PVCs. Carriers have found cost reductions of greater than 50 percent for long-distance service when mixing these circuits with other services over a shared ATM backbone.
The improved business models and opportunities stemming from ATM backbone adoption will change the dynamics of telecom to a degree that could equal the changes wrought by divestiture in the early 1980s. Indeed, this was the act that initially conceived ATM. As the new networking cycle unfolds, ATM truly is proving its case.
David Nelsen is the senior director of business planning at FORE Systems. His areas of expertise include ATM applications in medical, financial, entertainment and educational markets. Before joining FORE Systems, he was product manager of InterSpan ATM Service at AT&T. Nelsen holds a B.S. in systems engineering from the University of Arizona and an M.S. in operations research from Stanford University. >>>> |
