Sprint's and MCI's tactical strategies fall short

In baseball, sometimes you hit a home run. Other times, you hit just a single, and that's what Sprint and MCI seem to have done with their new tactical network announcements involving ATM and frame relay, respectively.

Both vendors deserve credit for recognizing that users are changing the way they consume network services. Fixed data rate, leased-line private networks were fine when it took two years to develop a custom communications application for an IBM mainframe. But in an age of shrink-wrapped network software, a new and flexible approach is needed.

Sprint's solution is ION, an ATM backbone network flexible enough to provide economical services that range from constant bit rate voice to very bursty Internet or virtual private network services. To keep all this flexibility from being lost in a rigid, low-speed access connection, Sprint plans to use fiber optics and other technologies to expand the capacity of the infrastructure that connects users to the network. The new local loop will give users high-bandwidth connectivity to ION, Sprint says. Where Sprint seems to have missed the boat, however, is in the way this flexible network connects to the subscriber. Sprint proposes to place a multiservice access device on the subscriber's premises and connect to it with ATM private virtual circuits (PVC).

ATM isn't even credible as an application interface to a carrier service - there simply aren't enough ATM-specific applications out there. However, there are a lot of IP applications, and Sprint acknowledges this by saying it will support IP over ATM. How? With PVCs. Earth to Sprint: This approach has already been tried and is neither flexible nor tactical.

If Sprint can come up with a really inventive way of providing IP services over ATM backbone networks, ION might really offer something new to the marketplace. Without that, it's just another ATM-to-the-office strategy.

MCI dodges the ATM bullet by basing its tactical data announcement in the proven frame relay space. The problem is, frame relay PVCs aren't any more tactical or flexible than ATM PVCs. MCI proposes to solve this problem by offering frame relay in switched virtual circuit (SVC) form. Users can get bandwidth when they need it by dialing up an SVC and get rid of both the bandwidth and its cost by disconnecting the SVC when they're done.

Like Sprint, where MCI seems to have missed the boat is in the user interface. There aren't any applications that set up frame relay SVCs, which will make consuming those SVCs a bit challenging. What might have been useful is if MCI had partnered with a bunch of equipment vendors whose customer premises equipment could allow IP applications to set up SVCs to provide special paths to infrequently used destinations or provided a specific quality of service (QoS) for applications sensitive to network performance.

The Frame Relay Forum issued an SVC document years ago, and the problem then was the same as it is now: Buyers have IP applications, not frame relay applications. There are standards, such as the Next Hop Resolution Protocol (NHRP), that allow IP devices to set up SVCs. In addition, Hughes Network Systems has announced a product that would use SVCs to support the newest, hottest IP networking enhancement, Multi-protocol Label Switching (MPLS). Did recommendations on NHRP or MPLS over SVCs get into MCI's announcement?

Nope.

MCI and Sprint seem to understand that users need more flexible data services to match the more tactical way they buy and deploy data applications. What they don't understand is that those data applications are IP applications - period. Public data services have to offer users an IP interface.

Comments on and criticism of Sprint's ION were directed almost entirely at the fact that Sprint chose ATM rather than IP as a backbone. That's an unreasonable standard to apply to a service whose stated mission includes insulating users from the technology of the transport network. It is reasonable, though, to criticize Sprint - and MCI - for failing to meet the first requirement of any network: to support application connections.

It appears Sprint and MCI are saying connection-oriented services such as ATM and frame relay have a big role to play in the evolution of data services overall and will be important in developing support for new and tactical data applications. OK, folks, then tell us how that's going to come about. What makes ION a more flexible IP service than routers and PVCs would provide today, Sprint? How does NetMeeting or HTML use those frame relay SVCs, MCI?

We seek simplicity in an industry that's growing more complex every day. However, this desire for simplicity doesn't justify announcements that lack substance. The simple message that MCI, Sprint and all the other carriers need to understand is this: The virtual network services that will form the basis of future applications will have to be based on IP.

Nolle is president of CIMI Corp., a technology assessment firm in Voorhees, N.J. He can be reached at (609) 753-0004 or tnolle@cimicorp.com.