|
|||
 | |||
|
With the ever-increasing need for building better meshed networks and supporting applications requiring high bandwidth, the traditional network based on leased-line time-division multiplexing (TDM) just isn't cutting it anymore.
TDM-based networks worked well when SNA ruled the land, but that's not the case with most new multiprotocol nets. But eliminating the TDM net isn't what most users are doing; many are including it as a security blanket while they move to more advanced and flexible technologies such as ATM or frame relay.
Users beginning the transition from TDM report that careful planning and execution are required so you can reap the benefits of those technologies without getting whacked by unforeseen problems and expenses.
"ATM lets us migrate our legacy traffic and diversify our network architecture," says Greg Britz, systems integration engineer for Burlington Northern Sante Fe Railroad in Kansas City.
The company is shifting from a TDM network anchored by DS-3 trunks to one powered by General DataComm, Inc. ATM switches. The existing 45M bit/sec DS-3 TDM network stays in place for narrowband applications, while the ATM network handles some of that legacy traffic, including voice, and leaves room for more.
"The ATM network allows us to be ready for the next move, which is broadband networking in a collaborative environment and video," Britz says.
General DataComm ATM switches allow network bandwidth to be used flexibly by supporting different qualities of service. Combined with priority queuing, the switches offer more efficient use of bandwidth than their TDM predecessors. For example, if top-priority voice trunks are not being used, data traffic can be sent over that bandwidth. In a TDM network, that bandwidth would be off-limits to other applications, idle or not.
That flexibility allows Britz to wring extra bandwidth out of his DS-3 ATM trunks. "With ATM, you can overbook a DS-3," Britz says.
By leaving the existing TDM infrastructure in place, the railroad can make a gradual, manageable transition. Some legacy traffic runs on the ATM net and some on the old TDM net. The two talk over OC-3 connections from legacy routers to the ATM switches. If the entire ATM network were to fail, the TDM network could keep the essential traffic moving. And the backup would work in reverse should the TDM net fail.
"Building ATM on top really bulletproofs our network," Britz says.
Maybe frame relay
Many of the existing T-1 networks out there, with their always available, dedicated bandwidth, were set up to support tempermental SNA traffic that just won't tolerate delay. But other technologies, frame relay in particular, have learned to coddle it.
Gregg Foley, the senior network analyst for the Cherry Creek School District #5, of Englewood, Colo., is in the midst of a transition from TDM dedicated lines to frame relay.
While the district still keeps some dedicated trunks just for SNA, Foley also has installed an SNA gateway from Novell, Inc.
The gateway encapsulates the rest of the mainframe traffic in IPX packets for transport across the frame relay WAN. With the gateway spoofing the mainframe to keep sessions alive, users don't know the difference, he says.One SNA link did run into trouble because of a problem that the frame relay provider, US WEST, Inc., blamed on a router in the school district's network, but that glitch was eventually resolved.
The back and forth to correct the problem, though, has led the district to invest in ADC Kentrox DSUs/CSUs that include network monitoring probes known as FrameVision. That helps Foley isolate whether link problems are in his gear or in the carrier's network, he says.
In addition, it gives him a more detailed read on actual traffic patterns, which lets him fine-tune the amount of bandwidth he buys. "It gives me capacity planning," Foley says. With proper bandwidth allocated for each WAN link, he can keep traffic moving.
While other technologies might hold some attraction, it is not necessary to swap out your current TDM network in order to make changes, according to U.S. Air Force Maj. Harry Opel, branch chief of information infrastructure engineering for the Headquarters Air Combat Command in Langley, Va.
The command is still using Network Equipment Technologies, Inc. IDNX chassis he bought in 1991 to support TDM. "The chassis is the same, but we are running frame relay and ISDN in addition to traditional TDM," he says. And he's looking forward to the ATM card expected to ship this month.
The common chassis leaves him with a core of gear that's constant, a big help for an IS department that suffers a 60%-per-year turnover in personnel.
"Feature cards are very easy for me to accommodate as opposed to a brand-new box," he says.
The government is constantly watching his expenses, so he has to make sure he can justify the technology changes he makes - a good lesson for anyone. "The onus is on me as a user to do the homework and point out the most reasonable solution," he says.
Still a dedicated market
Despite incursions being made by packet and cell technology, the demand for dedicated T-1 lines continues to grow (see graphic).
And prices have been dropping, in part prompted by newer, less expensive technology that sup-ports T-1 service with a smaller investment in equipment. The next generation of alternative technology, known as HDSL2, will allow provisioning T-1 service over two wires instead of the current four - another cost savings.
While the number of dedicated T-1 lines in service grows, that does not mean TDM is still enjoying boom times, according to Tom Nolle, president of CIMI Corp., a technology assessment firm in Voorhees, N.J.
"Only about 5% of new T-1s are genuinely muxed," he says. The rest support corporate Internet access, voice trunks terminating at PBXs or point-to-point data lines terminating in DSUs/CSUs or routers.
Abandoning TDM for other technologies can make sense for several reasons, he says. First, the cost of a full DS-3 or even a fractional DS-3 may be too high, given that the user may not need the full 45M bit/sec a DS-3 offers. DS-3s also are not available in some locations.
Throwing ATM gear on leased T-1s can offer more efficient use of T-1 bandwidth. Even with ATM overhead eating up 15% to 20% of each cell, users still can end up with more effective throughput, Nolle says, because idle bandwidth on the line is minimized.
