Usage-Based Bandwidth: Pay As You Go

By JOE MULLICH

For years, network managers were content to pay a flat fee for bandwidth, even though they sometimes paid for more than they used. Now, the ISPs are pushing for a usage-based "pay as you go" model, largely to set up an equitable way to charge for high-bandwidth services such as video conferencing.

Under a usage-based system, so the theory goes, IT shops would benefit because they would no longer pay for a T1 line when all they really need is 256 Kbps. And, on the supplier side, ISPs would finally have a way to charge for high-end premium services.

While many IT shops may be content to pay a flat rate for some time, analysts say a pay as you go model is inevitable and will play a critical role in the natural evolution of the Internet becoming more like a utility.

"People have grown up with the notion that Internet bandwidth is free," says Tim Wilson, vice president of marketing at ISP Digital Island Corp. "Although they have no problem accepting that other types of bandwidth, such as their telephone or utility bill, have variable costs.''

Digital Island's Wilson says with usage-based pricing, ISPs will finally have a way to charge for different types of service. Wilson likes to use the analogy of flying coach versus first class. If everyone pays the same price for a ticket, he says, no one gets champagne and extra leg room. Under the existing pricing models, ISPs have difficulty charging for higher types of service in the form of higher bandwidth apps.

"There are different costs for providing different types of service," says Wilson. "Usage-based billing will let ISPs offer different service types because they will be able to charge those costs back to users."

Mark Fisher, senior vice president of network services at ISP Concentric Network Corp., offers some straight talk: "The ISP industry won't have a rational market unless a usage basis for pricing emerges."

Roy Milano, manager of MIS at Hitachi Metals America Ltd., is a good example of an IT manager who took advantage of this new pricing model. Milano says in the past, he had been charged a fixed rate--no matter how much bandwidth he used. He finally decided to take action after he first approached the carriers and asked about preparing his network for videoconferencing. They all suggested he bring in three times more bandwidth than he really needed and said upgrading the network to support videoconferencing could take up to 45 days.

Essentially, Milano wanted to upgrade his network from 56 Kbps to at least 128 Kbps-and he couldn't wait around because the network was bogging down at 56 Kbps. Milano opted for a solution from Concentric's Enterprise VPN service, which lets WAN connections be adjusted from 128 Kbps to T1 speeds.

"A lot of IT managers are paying for a lot more bandwidth than they're using because they want to be flexible," Milano says. "I was talking to one company last week that was paying for a full T1 at every location, when they were only using 56 Kbps to 128 Kbps of bandwidth, because their carrier wouldn't provide the bandwidth as they needed it."

To be fair, Concentric is a bit ahead the curve. By and large, most ISPs offer that kind of service only to customers savvy enough to ask for it. In many cases, usage-based pricing has not been made into an official product offering. "Business customers are desirable enough to the ISPs that the [users] can call their own shots," says Hilary Mine, executive vice president of Probe Research Inc. "But in order to do that, they need to understand their own network traffic so they can determine where usage-based pricing makes sense."

Initially, some observers believe that usage-based bandwidth pricing will make its biggest splash among Fortune 100 companies that want a better way to track departmental bandwidth usage. The current static methods, such as going by headcount, aren't accurate enough to satisfy IT managers.

"Usage-based bandwidth billing is a new concept, because it's difficult to get departments to pony up for overhead accounts," says Ruth Chatterton, Internet consultant for TeleChoice Inc., a consulting firm. "Managers would much rather pay for what they use, rather than the current 'spread-the-peanut butter' method."

Actually, the degree to which ISPs are adopting these new pricing schemes is hard to pin down. Certainly, usage-based bandwidth pricing is already available in several incarnations. There are lower rates for doing data backups and file distribution during off-peak hours, for instance. Frame relay services are sometimes priced based on usage, but these pricing plans lack granularity and flexibility, says Jay Pultz, research director for the Gartner Group, a consultancy.

But the real trick, analysts say, is usage-based pricing that charges by metering IP packets. Right now, ISPs and companies don't have the technology to measure bandwidth at this level, but vendors are starting to offer tools that promise to make accurate charge-backs possible. Tribeca Software, for instance, just released NetCFO 1.2, one of the first usage-based billing packages for WANs. The product is so new that only two customers currently use it, according to Tribeca spokesman Rich Reitman.

One of those companies, J.P. Morgan & Co., says it cut costs by using the software to bring about changes in user practices and the way applications are developed. Before, business units had no idea how much bandwidth they were actually using. "Departments basically paid a flat rate for bandwidth," says Jerry Trollo, a vice president at J.P. Morgan. "Now, they are charged for the amount of bandwidth they use, so they are actively looking for ways to reduce bandwidth costs."

Some units, for example, are designing applications to be less bandwidth intensive. Others are running apps at off-peak hours to cut costs. "Under the older system, the units had no incentive really to reduce bandwidth costs," says Trollo.

Analysts believe demand for these types of billing applications will increase as networks are upgraded to allow for new and more bandwidth-draining apps.

"IT departments don't want to get caught between the CFO telling them to cut back on their budgets and users asking for better and stronger bandwidth," says Mike Stone, president and chief operating officer of Narus Inc., one of the independent companies developing technology to trace network traffic and provide billing. "The problem is 20 percent of the users are using 75 percent of the bandwidth. The important thing is to make sure these 20 percent are paying their fair share."

The hosting segment of the ISP industry has built in the biggest usage-based pricing component. Typically, providers charge according to actual data transfer, disk space reserved, plus bandwidth used.

"The hosting companies by definition need a little more sophistication in the way they provision support for hosting accounts," says Concentric's Fisher. On the hosting side, of course, providers try to deliver additional value-added services, including reports that measure time-of-day usage.

New services, such as content distribution, replicate content closer to the end user browsing the Web. This can deliver dramatic performance enhancements for latency-dependent applications and commerce applications, depending on how far traffic has to travel, explains Greg Howard, principal analyst for the HTRC Group.

Digital Island's hosting service, for instance, is a mirrored service that charges by the gigabyte, with charges that vary depending on distance. Wilson says usage-based services such as this suit electronic software distributors and credit card companies. "Usage-based pricing doesn't mean more expensive," Wilson says.

Of course, usage-based billing has many technical, practical and psychological roadblocks One big problem is that no accepted method for measuring bandwidth has been devised. "The most fair way is dollars per megabytes of payload traffic, excluding header information the network adds to the traffic flow," says Gartner Group's Pultz.

In reality, pricing schemes for different incarnations of usage-based bandwidth are all over the map. Some charge by the bit, per event, per broadcast or per audio stream using a fixed period of time for a certain rate of transmission. Some charge by the amount of port bandwidth, or the clock speed at the entry port going to the Internet.

The biggest technical hurdle is collecting detailed packet-by-packet information for billing and customer reports. Implementing usage-based billing requires costly back-office overhead. The technology exists to measure volume, but many ISPs have shied away from the costs of setting it up, according to Pultz.

"The problem with networks today is they don't have a granular ability to get bandwidth on demand," Pultz says. Burstable traffic streams can provide, for example, a quick download of a massive file and the customer will only pay for that download. "We will likely see more usage-based pricing for very bursty traffic, and companies can take advantage of that in designing applications," Pultz says.

According to analysts, start-ups are just coming out with tools that will solve that problem. ISPs are carefully preparing their systems to leap into usage-based bandwidth pricing when market demand develops--or when their competitors take the first step.

"Very few ISPs have the network management sophistication to bill bandwidth very well," says analyst Mine. "It's fine and dandy to bill for bandwidth for traffic on my network, but once it leaves my network, it gets tricky. For nine months, the big guys in the backbone game have been working together quietly to solve this problem, and new services should emerge in the coming year."

The Narus system places a probe behind each point of presence in a network, collecting information that is eventually packaged to an Oracle database. Because the system works outside the network, Narus says it can re-create these sessions for billing without slowing down the network. Frontier Global Center, a Web hosting and Internet service, has been experimenting with the system for billing enhanced services.

Generally, those tracking mechanisms are platform- or vendor-specific, Howard says. Solutions have to be flexible enough to allow for new technologies and services. "Right now, the ISPs have to tie it all together themselves," Howard says. "New products are just now coming out that will allow them to do this more easily."

Having ISPs handle billing requires a degree of faith that makes some IT directors uncomfortable right now. Milano at Hitachi Metals America tracks his own network usage with Concord Network Health.

"Most IT managers are afraid of going to a provider and saying, 'I want bandwidth on demand,' because they don't know how to monitor it themselves and they don't trust their ISP to give them the right information," he says. "They don't want to spend $20,000 with Concord Network Health just to monitor their bandwidth usage."

Given the relatively embryonic nature of the Internet, analysts say that many users are leery of adopting new pricing models. Probe Research's Mine says when AOL changed to the more familiar flat-rate pricing scheme, many customers who were not using the network much happily paid the $19.95 monthly rate instead of the $6 they had been paying on average in exchange for predictability of pricing.

IT directors negotiating bandwidth contracts shouldn't end the discussions with an ISP's salesperson, Mine says. "You need to talk with the supervisor to find out what their plans are," she says. "You don't want to get stuck in a three-year contract for a T1 with an ISP that might not be developing the usage-based pricing that could be best for you."

The idea is to start thinking about this now, so you can take advantage of usage-based billing when it becomes more mainstream. Most analysts say usage-based pricing will coexist with the flat-rate-fee model for some time. But while flat rates may offer a certain comfort level, it's much more efficient to pay as you go.

Joe Mullich is a freelance technology journalist based in Glendale, Calif. He can be reached at joemullich@aol.com.