Frame relay NNIs in the cross-fire
Big carriers tend to justify the telecommunications merger frenzy by saying users want to get all types of services bundled together into a single package from one player. But many users actually prefer to get different kinds of services from different carriers.
And then there are the users who even want to -- or have to -- get the same service from two carriers.
That's the situation that led to the development of a specialized form of networking called Frame Relay Network-to-Network Interfaces (NNI). One of the earliest frame relay standards,NNI has long been used by local exchange carriers and regional carriers.
Those carriers tend to sell frame relay service to users with many locations in a concentrated geographic area. But the carriers need to find another carrier to connect sites outside their territories - or, in the case of Bell companies, across any two local access and transport areas (LATA).
The challenge has always been to find a long-distance carrier that will agree to hook up its frame relay net to the local carrier's net. Now that challenge may be getting harder than ever because in the past year AT&T and MCI WorldCom have introduced their own intra-LATAframe relay services. Having their own intra-LATA frame relay services means AT&T and MCI WorldCom can provide multiple local frame relay connections in a single metropolitan area on the same switch platforms as their national frame services.
Should you consider a frame relay NNI? The answer is hotly debated among vendors and analysts, who alternately praise recent improvements in NNI procedures and practices or criticize NNIs as a bottleneck strewn with network-management headaches.
Some carriers want it all
These days, if you go to your long-distance carrier - especially AT&T - and ask if it will set up a frame relay NNI with your local carrier, expect to get pitched on moving all your frame relay business to the long-distance carrier.
AT&T is even reluctant to consider an NNI if you are involved in a merger with another company that uses a different carrier. Carrier officials say they would rather take their chances on winning or losing your entire business. "When corporations merge together, unless they deliberately choose a multiple-vendor situation, one carrier usually ends up dominant," says Keith Falter, AT&T's national marketing manager for high-speed services.
Temporarily, AT&T will assist users to put two separate routers on each site - one for each carrier's frame relay network - each attached to the LAN. But that's not a frame relay NNI because the two WANs never meet.
AT&T officials are candid that they will present users with alternatives if they try to force frame nets together. One alternative is AT&T's Local Frame Relay service, introduced last year, which puts multiple sites in a metro area on AT&T's switches for much less than the cost of AT&T's national frame relay. Another is an AT&T service called IP-Enabled Frame Relay, which sends one permanent virtual circuit from each site into an IP cloud that can terminate the connection at any other site. A third alternative is AT&T's Transparent LAN Service, which provides native LAN-speed connections, again generally in metropolitan areas.
Steve Taylor, president of Distributed Networking Associates in Greensboro, N.C., has aa different idea. He suggests that if users find themselves with frame nets from two carriers that won't agree to an NNI, the user can install a frame relay switch at a central site and set up the NNI there.
Sprint has a different approach. Although Sprint doesn't have regular public NNIs with local carriers that can be shared among multiple users, for a single user it will sometimes provision a private link, such as a T-1, between its frame relay net and another carrier's.
Not all carriers balk
But there are several second-tier, long-distance carriers that will happily link together local frame relay clouds across LATAs for multiple users who prefer to employ the regional Bell operating companies for frame relay. According to a recent survey by Distributed Network Architects, these include Frontier, IXC Communications and even GTE -- a non-Bell local carrier with no long-distance restrictions.
Another carrier that has specialized in working NNIs with RBOCs is Intermedia, a Tampa, Fla.-based company that began as a competitive local exchange carrier and has branched into long-distance data services. "Traditionally Intermedia has been very willing to do NNIs and is very aggressive in that market," Taylor says. Intermedia even offers network-availability guarantees across its NNIs that are only slightly less robust than those entirely on its network.
One of the old knocks on NNIs was that each carrier couldn't see across the connection to view the other carrier's network, leaving users in the dark as to who to call in case of an outage. NNI proponents claim that scenario is outdated. To address this question, the Frame Relay Forum approved a variation on NNIs in 1997 - Implementation Agreement FRF.10 - which allows two carriers to pass switched virtual circuits between each other to gain more flexibility.
To further alleviate finger-pointing problems, Sprint and local carriers sharing an NNI for a particular user will attempt to create a common addressing database. "You want to be sure that they have some common mechanism for referring to a particular circuit," says Tom Mennona, director of Sprint's packet data services center.
Make no mistake, says Taylor: NNIs can save carriers and users money. That's because if a user organization employs frame relay access from a local carrier to reach the long-distance carrier's frame relay switches, it avoids often-expensive leased-line dedicated access at 56K bit/sec or T-1 speed.
In fact, a shared NNI "could actually give you a better performance" than some of the connections on a single carrier's network, Taylor says. That's because if carriers set up the NNI with many users in mind - say, a T-3 link -- the mathematics of statistical multiplexing work in your favor and it's very unlikely that the NNI will be a bottleneck, as opposed to a 56K bit/sec branch-office connection that could max out with bursty traffic.