The case for IMA

Asynchronous transfer mode has been touted as everything from the harbinger of an exciting frontier in networking to networking doomsday. These wildly varying predictions hinge on finding the "killer app" for ATM.

The answer may lie not far from ATM itself: inverse multiplexing for ATM (IMA).

IMA, a specification for provisioning multiple ATM circuits in T-1 increments, was created to bridge a bandwidth gap between T-1/E-1 (1.5/2 Mb/s) and T-3/E-3 (45/34 Mb/s). In other words, IMA enables network managers to buy several lower-cost T-1/E-1 lines, aggregate the bandwidth and distribute ATM traffic across multiple physical circuits. But it can do much more.

Using IMA, ATM can propel itself from the backbone niche market to the mainstream and reinvent itself as the pre-eminent networking technology of choice. The only missing piece of the puzzle needed to reach killer app status is a lack of vendor interoperability. But vendors are working to clear this hurdle.

Several companies have formed the IMA interoperability initiative, called I3, to promote interoperable solutions among vendors so end users can choose from a wide variety of interoperable products.

The IMA specification defines the protocols and steps needed to divide a single high-speed stream of ATM cells across multiple lower-speed links and then recombine the cells into a single stream at the other end. The specification also defines link management and device management procedures, as well as procedures for connecting cell sources and providing cell adaptation services for non-ATM based services.

The basic mechanism of IMA is round-robin distribution of cells across multiple links. That means the sender sends the first cell on the first available link, the second cell on the second link, the third on the third link and so on. The receiver has to take the cells off in the same order to preserve the cell sequence.

IMA also provides a mechanism to monitor and control the flow of cells through the links. IMA control-protocol cells, combined with user data cells, form IMA frames. When there is not enough user data to fill an IMA frame, the end devices continue sending IMA frames with "filler cells" instead of user data. The receiving devices delete the filler cells before passing data to the network.

Although nominal speeds are the same for all the links in an IMA group, there is no guarantee that a constant delay can be maintained. Yet it is important that delay variation be minimized so that delay-sensitive traffic such as voice and video are unaffected. To accomplish this, the sender compensates for variations in delay by inserting "stuff cells" into fast circuits and maintaining a constant delay variation.

IMA also describes procedures to drop T-1 lines based on delay, bit error rate or manual command, and to add lines when the error condition is remedied.

Litmus tests How will we know if IMA becomes the next killer app? The label is frequently cited when everyone other than the customer is bullish on a technology or product. Yet from a historical perspective, it's worth asking, "Was desktop ATM a killer app? Was TCP/IP a killer app? What about fiber distributed data interface? Or multiprotocol over ATM?"

A true killer app is a combination of a new technology and an outstanding application that together cause explosive growth in the technology's deployment that can be measured in market penetration.

Figure 1 shows the relative market penetration in number of ports shipped for three technologies that were considered killer apps before they entered the market.

Many parallels can be found in comparing frame relay-the successful application for the 1980s-with IMA. Indeed, IMA may be on the verge of replacing frame relay as the access solution of the next millennium. All the following signs point to the same conclusion.

The most successful applications benefit both service providers and users. Just as frame relay replaced multiple leased line connections by providing a single link to a frame relay switch, IMA helps end users consolidate their traffic on a single network connection of flexible bandwidth.

At the same time, service providers-particularly competitive local exchange carriers and Internet service providers-can design innovative service packages such as data only, data and voice solutions, or data, voice and Internet access, each with its own price point. In addition, CLECs can use IMA to consolidate customer traffic from remote points of presence (POPs) to their own central offices and data centers (Figure 2).

Furthermore, a killer app must be transparent to the user. IMA, in most of its applications, fits the bill.

Other factors include timing, the service provider's scenario and the bandwidth gap.

Timing. The same trend that is changing the service provider landscape is causing enterprise network managers to hand off their communication needs to service providers so they can realize enormous savings. A decade ago, when traffic moved from asynchronous terminals connected to a 3270 to a LAN-based enterprise, X.25 could not provide the enterprise's high bandwidth requirement in a cost-effective and manageable way. That's when frame relay entered the picture.

In today's carrier situation, frame relay above T-1 rate is not commonly available. An integrated service access solution over T-1 ATM would provide new alternatives to enterprise users when the single service infrastructure is fully depreciated and enterprises are looking for the next generation access solution.

The service provider's scenario. The Telecommunication Act of 1996 mandates that competitive carriers have access to terminated T-1s, but it says nothing about CLECs having access to switched services such as DS-3 (44.7 Mb/s). Hence, considering the wide availability of T-1s, CLECs have an opportunity to turn these lines into innovative data services.

The bandwidth gap. Enterprise users are looking for bandwidth relief when interconnecting between sites and connecting to the Internet. With the variety of high-speed networking technologies grabbing a foothold inside the enterprise, it's surprising to see that this same enterprise is connected to the outside world by a single T-1/E-1. Although network managers want higher-speed connections to their service providers, the next block of bandwidth is significant for anyone to swallow-in price as well as size.

IMA-based services, both for enterprises and service providers, have been sporadic. In May 1997, MCI became the first interexchange carrier to announce IMA-based ATM services. Since then, many service providers have announced plans to deploy ATM-based services, but they are waiting for competition among interoperable solutions to evolve so they can choose the best-of-breed solution.

IMA applications IMA already fulfills certain access and trunking applications needs.

IMA can be used to bridge the bandwidth gap as an access solution from an enterprise to the nearest service provider location (Figure 3).

For example, an enterprise can use IMA to connect T-1/ATM to branch offices and T-1/IMA to head offices. Branch offices have the upward flexibility of adding more IMA links in T-1 bandwidth increments. IMA also can provide flexible bandwidth for high-speed LAN interconnection among an enterprise's branches in different cities over the WAN.

Because of the flexibility in adding bandwidth, end users can choose from a variety of videoconferencing services. For example, they can choose the MPEG 2 video at 3 to 6 Mb/s or full-motion quality video at 6 to 10 Mb/s.

IMA also can be used to provide efficient trunking to backhaul traffic from remote locations to CLECs' COs or data centers. A CLEC with remote POPs can then make the best use of bandwidth and be profitable (Figure 4).

The key to transforming IMA lies in providing a transition path from existing disparate single services to multiservices support over a single platform. The end user benefits from service consolidation over a single connection to a single service provider, both financially and in operations procedures. IMA integrates a variety of equipment and applications without a high replacement cost.

IMA also offers the best of the inverse multiplexing technologies to date, especially when compared with other applications that are either proprietary in nature or use an older bit-based inverse multiplexing technology that is proprietary and that may not offer quality of service guarantees.