Great EXPECTATIONS

Zero tolerance has arrived in the provisioning zone. That slogan was once reserved for network failures with the promise of shortening mean time to repairs. Now it is extending to the edge of networks where customers have little tolerance for the three- to six-month waiting period to provision circuits. The current push is for immediate bandwidth packaged in a customized rate that can be launched and then dismantled on an as-needed basis.

Such bandwidth-on-demand is expected in core networks and in the metro market by year's end, according to wholesale carriers and vendors alike.

Much of the high-speed, flexible service will arrive on the back of the optical network that can scale the heights needed for these intuitive features that make applications intelligent.

Before bandwidth-on-demand is widely available, watch for a slew of billing options that will introduce carriers and their customers to the malleability of packet networks and the intelligence they bring.

Changing of the guard

Customized, dynamic applications are some of the strongest selling points for intelligent packet networks. By placing control on the desktop, provisioning becomes a point-and-click exercise where customers not only can choose how much bandwidth they want and for how long, but they also can select restoration capabilities and other options tied to service level agreements (SLAs).

Although these features are not generally available today, the networks over which they will ride increasingly are.

The preferred method to deliver such flexibility in an IP network is to run it on top of an optical core. And by extending optical networks into the metro, carriers save the expense of optical-electrical-optical conversions and unclog the last-mile bottleneck.

Most people are carrying 10 gigabits on their wavelengths. As you go to these higher speeds, electrical systems become exceedingly uneconomical, says Bruce Nelson, director of optical services architecture for Lucent.

As you go to 40 gigabits in the future, these systems probably won't be able to handle that capacity because of the cost structure. Second, because of their electrical fabrics, they will not be able to scale in terms of bit rate on the channel or in terms of size of number of ports that you will need.

Last July, Lucent shipped its WaveStar LambdaRouter an all-optical switch and then followed up in December with a companion wavelength management software. These smart products allow switches to automatically discover each other and dynamically redistribute wavelengths in response to traffic levels.

While the initial release of the Lambda-Router focuses on wavelength services, Lucent expects by year's end to incorporate IP services that will make bandwidth available in customer-driven increments. One of the most powerful features of an optical network is that it gives carriers the tools to transfer control of bandwidth scaling into the hands of its customers.

In today's world, networks are very static with a lot of headroom, says Tom Osha, Broadwing's chief of staff. For example, if you have an application that requires 160 Mb/s, that's bigger than an OC-3 but it is a lot smaller than an OC-12. Today you would have to buy an OC-12 and basically just keep the overhead because of the way that bandwidth is sold. This becomes very expensive.

Last month, Broadwing launched two of three segments in its all-optical network that will carry IP, ATM and frame relay traffic at a debut rate of 10 Gb/s (Figure 1). In the second phase of its expansion, expected within a year, Broadwing plans to add customer-driven options, including metered billing whereby subscribers can chose incremental services based on 1 Mb/s configurations.

We believe that incremental service is actually going to allow for an explosion of applications that never could have come to market before, Osha says.

Some of those bandwidth-on-demand applications are already taking center stage. Last June, Broadwing turned up an OC-48 for less than two weeks for a customer that broadcast an international fashion show over the Net. When the high-speed service was no longer needed, the circuit was taken down and the customer resumed its normal operation at OC-3.

A customer could not have afforded to be able to do that without this incremental ability to go up and down, to buy what it needed, because this client would never have wanted to pay for a one-year OC-48 circuit to do a 10-day event, Osha says.

Getting what it takes

Capacity is the foundation of intelligent packet networks. Carriers may need capacity to offer high-end services, yet there is little economic value in bandwidth that lies dormant awaiting bursts of traffic. That Catch-22 may be part of the reason the provisioning is typically a multi-month process.

I think it's a dirty little secret in the industry that there's not that much capacity out there, Osha says. A lot of folks believe the industry is ripe with capacity and you can get it anywhere. But there are really only three [interexchange] providers that have available capacity today. Most providers don't have the OC-48 capacity. They'll take the sale, book the order and then go out and build or acquire it. That's why it takes 120 days to provision a circuit.

Conversely, Alcatel cites a Telstra Research Laboratories finding that fiber optic cable is being put underground at three times the speed of sound, or 2100 kilometers per second. As that fiber extends into the metro, it will likely loosen the capacity stranglehold in the access market and further the cause for bandwidth-on-demand from the local scene.

It's a vicious circle, says Ben Crosby, senior technologist for Alcatel's core data group. Networks that are growing because bandwidth is available and because this broadband capacity is available to end users, they're finding ways to use it. All of a sudden, people want to use these services and they're generating traffic. Obviously, on the access side, if you've got that much traffic being generated, you have to increase capacity dynamically to accommodate that.

Some very visible failures to respond to bandwidth demands may help facilitate dynamic provisioning. Last summer when the Concorde crashed outside of Paris, the BBC Web site buckled under the demand for information and virtually went offline. Such bursty traffic is precisely that which would be protected by employing bandwidth-on-demand.

Big and little steps

Although bandwidth-on-demand is packed with mass end-user appeal, it is a higher order of magnitude for most networks and not necessarily at the top of every carrier's wish list. The premium intelligent product that carriers crave today is rapid, remote provisioning.

The first application is not bandwidth-on-demand. It is provisioning. It is time to service, says Greg Reznick, president of the Xros division of Nortel Networks. We are talking about provisioning in a matter of days or minutes or seconds. And that scale depends on the particular strategy of the carrier. But the idea is to be able to deliver capacity very soon and not have to depend on it being used for a long period of time.

So if I want to back up my entire campus into remote storage, I might like to have an OC-192 for 30 minutes every Tuesday and have the carrier arrange that. If you have a photonic network, you can. That is the kind of demand that customers are starting to place on their carriers.

Not only is provisioning moving toward the instantaneous, it is becoming a virtual event with Web-based ordering that is expected to debut this year.

As we roll out this next generation of optical cross-connects and extend that reach into the metro, we are looking for that point-and-click provisioning, says Rodger Williams, senior manager of product development for Williams Communications. If a customer needs to nail up another OC-48, they can basically go out there on a Web interface and nail up that service and also receive the typical SLA type of information that they are going to require.

Williams Communications offers another precursor to bandwidth-on-demand known as burstable billing that allows traffic to surge within a port. Customers subscribe to a given service, such as 50 Mb/s within an OC-3 port, but have the option to use that full port or 150 Mb/s at any point.

Burstable billing traffic is capped at the top end of the port speed. Bandwidth-on-demand is unrestrained in its ability to scale up and down, such as jumping from OC-3 to OC-48 and then back to OC-3.

Williams expects to offer bandwidth- on-demand in the fourth quarter. Before it does, it is rigorously lab testing the equipment.

The equipment needs to be very scalable, Williams says. The OSS needs to be integrated into what we are using. It needs to have functionality that we can push out through a customer management system that we developed in-house. So it needs to be very user friendly in the world of point and click to design on-demand service.

There are some pieces that are instrumental in us moving forward with bandwidth-on-demand services such as going to the edge and opening up the edge with smart equipment. We are really looking for a solution that is very cost-effective to push [customer premises]-type equipment to the edge, says Williams.

Williams Communications does not shy away from start-ups in its search to find such equipment. It has drafted its share of industry rookies as part of its farm system approach, whereby it recruits equipment vendors and then puts the products through its paces.

We will be one of the first, if not the first [carrier] to provide the optical cross-connect, Williams says. We are giving our customers what we like to call optical [quality of service], which is basically application-based restoration services. Our farm team approach works closely with vendors to help them design and deliver products that meet our goals and customer needs.

Cisco Systems takes a similar approach to finding talent via its eco-systems partnerships. In January, Cisco began shipping destination-sensitive services supported by its 7600 optical services router. The feature gives carriers the flexibility to tailor pricing based on the amount of time traffic traverses a network. In the third quarter, it plans to release a companion product developed by an eco-systems partner that allocates bandwidth based on time of day.

An SLA management tool is also in R&D at Cisco that coordinates applications at layer two for ATM and frame relay or layer three for IP. One of the functions of this real-time monitoring system is to verify bandwidth use.

If a customer has subscribed to a certain bandwidth, there is no tool on the market that measures exactly what kind of bandwidth he is really getting over a period of time, says Udaya Shankar, product manager for Cisco.

An SLA could very clearly identify if he is really getting the bandwidth that he is paying for and whether or not additional amounts are needed. You can see what kind of traffic is going through what loads and how the networks are really carrying traffic, he says.

A managed approach

If the number of applications developed for a technology indicate a level of maturity, packet networks are well on their way to a fruitful development. Vendors are announcing a bevy of management solutions expected to launch in 2001 that increase the intelligence of packet networks.

ALMA Vision IP is Alcatel's OSS service assurance platform that communicates with devices as traffic flows across the network. It is expected to serve as a foundation for dynamic provisioning and bandwidth-on-demand services with a traffic engineering solution that optimizes the network.

ALMA Vision will specifically help enable bandwidth-on-demand by identifying congestion due to bandwidth bursts and then building an alternate label switch path with greater bandwidth to reroute the traffic, says Stephen Forney, Alcatel director of business development for OSS solutions.

We use the [generic rate adaptation for traffic engineering] algorithm proprietary to Alcatel to determine traffic patterns, he says. The Alcatel traffic-engineering tool first discovers the network elements, he says. We identify the traffic trunks. We go through a simulation model to see what network optimization steps need to occur, and then we can either automatically or manually apply those configuration changes to the network.

IP may be in the limelight for new applications, but for quality of service and management features it historically has conceded to the more mature cell-based ATM technology.

The Internet Engineering Task Force is working to change that perception. It developed a standard called multi-protocol label switching (MPLS) to bring scalability, flexibility and performance management to layer three. But perhaps its most salient quality is its ability to converge disparate network technologies.

Traditionally, a lot of these big carriers have had separate ATM and IP networks, Crosby says. MPLS is potentially one of the most important things that has come along in the networking world in the last few years to allow us to join separate IP and ATM networks and run end-to-end services and manage them across both networks as one domain. It allows carriers to focus their growth on one particular network technology that encompasses all of the requirements that they have today.

MPLS has come a long way in its short life. In addition to addressing Layers 2 and 3, another version of the standard is being drafted to specifically address the optical core. Generalized MPLS, also known as MPLambdaS, is expected to further the aggregation of traffic between edge devices and the optical core. Some manufacturers, however, view the technology as superfluous.

There are some people who claim that bandwidth-on-demand is not going to happen, says Reznick. The people who say you don't need to worry about MPLS are also the people who say that what you really want to do with these huge bandwidth pipes is nail them up and leave them there and let the routers run it all by packets over a static optical network. That is where the argument lies. One argument is that multiplexing packets is how you want to do it and all the switching ought to be in routers with none of it in the optical layer. That is kind of a waste of computing power that is routing those packets because they are going to be taxed very heavily.

Reznick says that the network will want to provide an agile access to high-capacity bandwidth to the point where wavelengths need to be redirected and wavelength services need to be offered.

Regardless if one standard management approach or multiple vendor-neutral solutions prevail, the message resounding from the edge and across the backbones is that customers have zero tolerance for long service delays. Carriers are responding with destination-sensitive services that cut costs and with point-and-click provisioning, which is poised for launch. Bandwidth-on-demand, however, will not likely see the optical light of day until year's end.

Pat Blake is a freelance writer based in Cedar Rapids, Iowa. Her e-mail address is b814@aol.com.