A layered approach

Asynchronous transfer mode continues to gain popularity with service providers because it meets emerging requirements for scalable transport of myriad end user services, including voice, video and data. ATM also operates over existing transport technologies worldwide, including DS-1, DS-3 and Sonet.

As the next generation technology most likely to succeed, ATM appears to have it all. Yet substantial deployment issues must be overcome if carriers are going to maximize the benefits of ATM.

One obstacle to effectively deploying ATM is its strong dependence on the physical layer transport mechanism. Unlike voice service, which can tolerate error rates as high as one per 10,000 data bits, ATM is highly susceptible to even moderate error rates. As a result, carriers need a thorough testing strategy to verify the physical layer's ability to transport ATM effectively.

A layered, bottom-up test strategy can help service providers eliminate problems systematically at the physical layer before they affect ATM service (Figure 1). This approach will ensure a high-quality transport mechanism to eliminate the problem of excessive cell retransmission. To understand why this is such an important issue, it is useful to examine the relationship between the physical layer and ATM cell retransmission in more detail.

To ensure high-speed transport, ATM switches do not retransmit cells. Instead, the burden for retransmission is placed on the application riding over ATM. An example is transmission control protocol (TCP).

Variable lengths of TCP data are assembled into a protocol data unit stream and then broken into as many as 100 individual ATM cells by the ATM adaptation layer. If even one cell is dropped because of bit errors or network congestion, the application will request the entire protocol data unit to be retransmitted.

This situation can spiral out of control quickly because retransmissions cause more congestion, dropped cells and retransmissions. Ultimately, this can cause the application to simply quit. Additionally, excessive retransmission on a usage-tariffed service means that users wind up paying more for lower quality service. The solution is a comprehensive test plan beginning with physical layer verification.

Physical layer testing Physical layer problems are often difficult to find once ATM service is turned up. Signals of questionable quality can cause some transmission equipment to generate errors while others perform adequately.

Intermittent errors are often the result of loose connectors or network elements with poor receiver sensitivity. Simple in-service tests can yield a good amount of information. However, because it is difficult to break into a network to verify physical layer performance once service is turned up, it is also critical to verify DS-3 and optical backbone operation during installation.

Out-of-service tests performed during installation on DS-3 ATM switches, DS-3 cross-connects, DS-3 multiplexers and Sonet add/drop multiplexers can isolate a host of problems that can interrupt ATM service. In- service tests allow service providers to isolate faults caused by existing DS-3 equipment.

Important DS-3 tests include: Bit error rate test. This commonly used out-of- service test isolates physical layer problems. Note that BERT may not detect problems such as a borderline pulse shape.

Pulse shape and signal level. This out-of-service or in-service test isolates connector and cabling problems such as loose cable crimps and incorrect coaxial cable lengths that can cause intermittent errors.

Bipolar violations. This in-service test detects physical layer impairments that can create excessive noise in DS-3 signals. Improperly grounded equipment, static charges, unterminated plugs and signal taps can cause bipolar variations.

Jitter. This out-of-service or in-service test verifies end-to-end timing of the circuit.

Many existing fiber optic cables were installed to transport lower rates than they are now required to support. As systems are upgraded to increase bandwidth, intermittent errors begin to surface.

Testing the optical backbone during provisioning is essential to ensure that it can handle the higher speed transport often required to support higher bandwidth demand. Optical return loss and insertion loss tests work together to prequalify the span.

An optical return loss test indicates the network's ability to transport high-speed signals by measuring the total accumulated power from reflections in the fiber. Too much total reflected power can interfere with the transmitting laser, causing degraded service. As useful as it is, however, the optical return loss test alone cannot determine that sufficient power actually arrives at the optical receiver because measurements are taken only at the fiber's transmitting end.

For example, a sharp bend in a fiber can allow light to escape the cladding (Figure 2). But because the bend reduces total reflections sent back to the transmitter, the optical return loss measurement actually improves.

Another common problem the test will detect is dirty connectors at the light distribution frame, which will cause a large reflection very close to the transmitter. This problem will not be caught by an optical time domain reflectometer because the light distribution frame falls in a dead zone.

These types of problems may or may not affect the total amount of power crossing the span to the receiver. To detect problems of this type, an insertion loss test must be conducted.

An insertion loss test measures total power loss across an optical span by comparing the power transmitted from a stable source with the power received at a destination. To prevent intermittent errors or a system failure, power loss should be set within a defined range. An acceptable range of power loss for single-mode fibers is 5 to 15 dB, although high-power lasers can require greater loss to avoid receiver saturation.

Correcting improper optical power levels may require using an attenuator, installing new fiber, polishing connectors or adjusting the transmission equipment's output power levels. Improper optical power levels can cause problems ranging from intermittent errors on live traffic to complete signal loss.

Sonet configuration testing Once the physical layer has been thoroughly checked, the next test layer is the network's transport technology. Because of its bandwidth capacity, Sonet is the technology most commonly used to transport ATM service.

Improper Sonet configuration can cause hard-to-find problems on the payload signal. Left unchecked, these problems can degrade service to the customer and delay turn-up of new services. Although difficult to detect after turn-up, Sonet configuration problems are relatively easy to isolate with end-to-end tests before live traffic is commissioned.

One frequently encountered Sonet problem is improperly configured synchronization. Often caused by network elements that are left in holdover timing mode after installation, timing problems in Sonet are automatically compensated for by "pointer adjustments," which protect payload data while it is within the Sonet network. Pointer adjustments cannot protect ATM data while outside the Sonet network-as when carried on a DS-1 or DS-3 signal to an end user-and are often to blame for those tributary problems.

A few Sonet pointer adjustments will not cause problems on the tributary ATM signal by themselves. However, the presence of pointer adjustments may indicate a serious problem.

If the timing of a Sonet element is allowed to drift, a large number of pointer adjustments can be generated days or weeks after installation. Because of the length of time between installation and service provisioning, a slowly drifting clock often means that tests performed on the ATM tributary signal at installation pass, yet service still fails later.

Some testing philosophies verify only the tributary network. Unfortunately, tests performed solely on the tributary ATM network cannot identify a Sonet timing problem as the root cause. This is why an installation test should include a timed test to verify that no pointer adjustments are occurring on the Sonet circuit.

A common cause for turn-up delays is improperly configured Sonet network elements. Often, each network element in the circuit must be manually configured for either concatenated or non-concatenated signals.

A concatenated signal-such as an OC-3c signal-combines payload space to provide greater bandwidth capacity than a non-concatenated signal. Concatenated signals are primarily used for ATM transport, while non-concatenated signals are used to transport legacy DS-1 and DS-3 signals.

Although the physical characteristics of concatenated and non-concatenated signals are identical, important differences determine how the two signals are handled by Sonet transmission equipment. Network equipment that provides access for non-concatenated signals cannot always support the higher bandwidth concatenated payload without hardware upgrades.

If the network equipment is not properly configured to transport concatenated signals, Sonet alarms-such as a path alarm indication signal-will be generated when a customer's ATM traffic is placed on the network. Similarly, if network elements are configured for concatenated signals, non-concatenated signals may not be passed through the equipment. Simulating the proper service traffic during installation ensures end-to-end path continuity before live ATM traffic hits the network.

ATM configuration testing Error-free delivery of data, voice and video services over ATM depends on error-free transmission facilities, but ATM service will not perform as promised without the ATM network elements' error-free performance. To ensure quality service at turn up, it is important to perform out-of-service tests to verify virtual path/virtual channel provisioning by testing the ATM cells and switching components.

ATM configuration layer testing also should include an out-of-service, end-to-end check to verify that ATM cells reach their proper destination and in-service monitoring to identify cell transport problems, such as delay variation and congestion. These problems can cause users to experience lost data, echo during voice conversations, or jitter during video transmission.

Out-of-service testing can verify specific ATM switch configuration on a per-channel basis (Figure 3). A common problem that arises during the installation of ATM switches is the improper provisioning of the ATM addresses and ports, causing misrouted or dropped cells.

Because data traffic is not always predictable, a recommended test is to verify that the ATM switch properly handles data rates up to and exceeding the contracted bandwidth on each incoming address. One way of doing this is to configure a sustained cell rate and generate data bursts above the contracted rate. Depending on the architecture, the excess cells will either be dropped or set their cell loss priority tags to indicate low priority.

Many ATM switches also are capable of modest error correction of individual ATM cells. If a single error is found in the header of an ATM cell, switches can be provisioned to correct this error to prevent unnecessary retransmission.

On the other hand, multiple errors in the ATM header should cause the cell to be dropped to prevent misrouting the cell. First generation switches often do not perform this function, so a simple test to generate single or multiple errors on ATM headers is recommended to verify proper operation.

Another simple out-of- service test that should be performed is an end-to-end connectivity test. This test simulates customer traffic by generating test cells with traceable identification to verify that those exact cells are received at their proper destination.

If cells are not sent to the expected address, they might have been given an unexpected address by an improperly provisioned switch. To locate the improperly configured element, a search mode can be helpful. In search mode, a test device locates the original cell by ignoring all addresses and examining the contents of each cell.

In-service monitoring of the ATM network enables service providers to detect congestion problems early so they can be isolated and corrected before customer traffic is seriously affected. Installing test access monitoring points using splitters, monitor jacks, or dedicated ATM test access points allows service providers to quickly locate addresses that unexpectedly consume high bandwidth and to isolate congestion problems.

Once the offending address is identified, the user's configuration can be checked and network bandwidth can be reallocated as needed. In-service tests will also be able to identify delay variation, overall congestion and local congestion to resolve problems before they can cause excessive retransmission.

For ATM to remain a viable service and backbone technology, the performance of the physical layer and the network's transport configuration must be reliable. Systematic and consistent use of this layered testing strategy provides the best assurance that ATM deployment will not be impaired by physical layer problems.

Laura McInerney is Product Line Manager at Telecommunications Techniques Corp., Germantown, Md.