| When it comes to connecting private nets to carrier ATM services,
there are plenty of options. Providers offer a wide variety of access
configurations, including ATM circuits at T1 (1.544-Mbit/s) and higher
rates, to link companies to suppliers or customers in the same LATA
(local access and transport area), across LATAs, or across long-haul
circuits (see Figure 1
). And a broad choice of ATM services means netmanagers get greater flexibility, better connectivity, and better QOS
capabilities than they would with leased lines or frame relay
services.
Trouble is, ATM also can be more expensive than those other options
(at least at first). So why should corporate networkers even bother
considering it over private lines or frame relay? Because corporate
WANs long ago moved away from being one-trick ponies, where a single
T1 handled SNA traffic and another circuit handled voice. Today's
corporate WANs handle far more apps and far more protocols, and they
present far more demanding requirements for high availability, high
throughput, and low latency.
But leased lines and frame relay impose significant limits as to
how individualized connectivity can become. Frame relay, for example,
offers only basic congestion-control facilities, and it doesn't
differentiate among applications unless net managers specify a
separate link for each app—and carriers sell frame relay by the
circuit. ATM, on the other hand, can carry multiple traffic types over
the same link, each with unique QOS parameters.
The Hard Part
Powerful and flexible—but complex. Don't assume that every ATM
network is the functional equivalent of every other ATM network.
Generally, ATM nets vary quite a bit in design and configuration.
Among the variables are Lane (LAN emulation ) services, SVCs (switched
virtual circuits), PVCs (permanent virtual circuits), call routing,
QOS parameters, and traffic shaping and policing.
If the design of one ATM network doesn't mesh with that of another,
then interconnection problems crop up. Nowhere is this more evident
than when a private ATM network must be linked to a public one:
Incompatibilities, standards anomalies, and ATM address variations
between the private and public ATM networks are the interoperability
issues likely to arise.
Specific interoperability issues include whether the carrier
supports SVCs; which version of the ATM UNI (user-to-network
interface) the carrier backs; whether the carrier offers QOS levels at
all; what upper-layer protocols the carrier handles; and whether the
carrier supports the ATM Forum's PNNI-1 (private network-to-network
interface version 1) and MPOA (multiprotocol over ATM) specs. Each are
of major importance in determining how private ATM networks connect to
carrier services.
Addressing issues are also a major concern. There are at least
three different types of addressing offered today. Two of the three,
the DCC (Data Country Code) and the ICD (International Code
Designator), are based on ISO's NSAP (International Organization for
Standardization's Network Service Access Point). The third format,
Native E.164, was created by the ITU-T (International
Telecommunication Union-Telecommunication Standardization Sector), an
international standards body. While close in structure, these
specifications are not entirely the same; if a private ATM network
uses one form of addressing and the carrier uses another, then there
could be routing issues.
Unfortunately, some public carriers gloss over these issues,
thereby oversimplifying ATM and its associated features. Some carriers
boldly talk about "turnkey" ATM networks while treating ATM as if it
was a superset of frame relay. Oversimplification raises unrealistic
expectations among potential users, particularly those with little ATM
experience. In one case, a carrier promoted ATM as a plug-and-play
technology to a company's top executives. Unfortunately, ATM networks
today are not particularly adaptive nor are they plug-and-play. The
company's networking staff was left to correct the carrier's
oversimplified presentation.
Missing Parts
There's something else net managers should watch out for. Chances
are, the campus ATM technology will be more advanced than the one the
service provider uses. Thus some of the features of the private ATM
net may be missing from the public network.
For example, many campus ATM nets use PNNI-1, which allows dynamic
rerouting of ATM circuits around failed links or switches. But
carriers may only support PNNI-0 (an early interim spec that involved
static routing)—or perhaps no routing at all. The upshot is that net
managers may be required to furnish the routing function across the
carrier network by deploying conventional routers at the edges of the
ATM cloud, or by adding ATM call-routing functions to switches
connected to the carrier's ATM network. Unfortunately, this is a
problem that's usually discovered only after it's too late.
Some carrier ATM networks are PVC-based, offering no signaling or
QOS. The lack of SVC capability makes many-to-many connections more
difficult to accomplish between sites, although not necessarily
impossible. Net managers may be able to tunnel SVCs through a PVC or a
PVP (permanent virtual path) on the carrier's network. However, the
choice of SVC, PVC, or PVP connections from the service provider has a
significant effect on network design, interface requirements, and
operational capabilities.
Don't Speak Softly
What a carrier actually delivers may depend on more than just its
network capability. How loudly a customer makes demands also could
play a role. Some customers have been able to get more features than
initially offered by threatening to penalize the carrier—or not award
it the project at all. Rather than lose the business entirely, the
carrier's sales force may press the engineering department to furnish
better services.
Of course, this can be risky. When a carrier is pressured to
provide more than it's capable of—or earlier than it can—serious
problems can occur. Its technical or operations staff may not be fully
prepared to meet a customer's requirements. The appropriate tariffs
may not be in place, or the carrier's billing options may not
accommodate the service.
When it comes to telecom services, the public service commissions
rarely feel a sense of urgency. By asking for too much too soon, net
managers may place their networks and their service providers at the
bleeding edge. This is one area where it's best to proceed with great
caution.
In some cases, net managers may need to deal with more than one
carrier. For example, connecting sites in Boston and Raleigh, N.C.,
might mean working with three carriers: Bell Atlantic Corp.
(Philadelphia), Bellsouth Corp. (Atlanta), and an interexchange
carrier like AT&T (Basking Ridge, N.J.), MCI Communications Corp.
(Washington, D.C.), or Sprint Corp. (Kansas City, Mo.).
While working with more than one carrier isn't unusual, ATM changes
things. Each carrier may offer different capabilities and features.
For example, the Boston-to-Raleigh link might require OC3c
(155-Mbit/s) service, but one of the carriers may top out at T3 rates.
If this is the case, then shop around for alternatives. Remember that
alternative carriers, such as cable television providers or power
utilities (particularly those which own rights-of-way), may be able to
offer better wide-area connections, although they may not advertise
the fact. It's a good idea to ask them if they offer ATM service.
Smart Steps
So in the end, what steps should net managers take when dealing
with an ATM service provider? Try not to be adversarial. Remember,
it's in carriers' best interests to offer ATM services that deliver
value to customers—and put money in their own coffers. Also, remember
that ATM is an evolving technology: Allowing the private network and
the public carriers to undergo a phased implementation over time
assures that the network growth will be sensible. Don't expect all ATM
features to be in place day one—on either the public or private
networks.
It's also a good idea to work together. Get the company's technical
people working with the carrier's technical people as early as
possible—and keep them working together through the design and initial
operational phases. And use the sales team effectively: Let the
carrier's sales team be the conduit between the company's network
staff and the carrier's, not the voice for both.
Finally, remember to be realistic about the ATM sales pitch.
Realize that some carriers have their sales forces out selling ATM
before they've placed orders for ATM switches or even finalized their
own network designs. And while they're probably reluctant to admit it,
the salespeople may not understand ATM even though they're being asked
to sell it.
Here's wha
t it all boils down to: Carrier-based ATM offers greatpotential for connecting corporate sites. But successful deployment
depends as much on a net manager's savvy as it does on the
service.
Jeffrey N. Fritz is principal network engineer for West Virginia University (Morgantown, W. Va.). His e-mail address is jfritz@wvu.edu.
|
