End-to-End Physical-Layer Testing for ATM Service

Efficiency and reliability are both reasons why ATM is an attractive WAN service. ATM operates over optical facilities such as OC3 where the transmission quality is virtually error-free. As such, ATM does not contain any error-checking capability of the data on a hop-by-hop basis. This results in more throughput, but the service must rely completely on an end-to-end, error-free transmission path.

Error-free transmission is simple, with clean, high-speed optical backbone links, but it becomes a real concern over the end-point copper access circuits such as DS1 or DS3. Access circuits often traverse junction points before arriving at the backbone switch port, making the circuit more prone to errors.

Copper access circuits between ATM subscribers and service providers vary from near-optical quality to barely suitable for voice traffic. For this reason, service providers typically specify a guaranteed BER for an end-to-end ATM connection.

The function of data-error detection and retransmission no longer is the responsibility of the WAN, such as in X.25, but instead the responsibility of the user end-station’s transport layer, such as TCP. It can take up to 32 ATM cells to make up one TCP/IP packet. If just 1 bit is in error within an ATM cell, it can cause up to two complete TCP/IP packets to be retransmitted or as many as 64 ATM cells. For noisy circuits, this produces a snowball effect where the virtual connection between stations is congested with retransmitted data.

A simple method lets you test the end-to-end transmission quality of an ATM virtual circuit for comparison to carrier-service-provider SLAs. In addition, a troubleshooting tactic isolates errors or poor performance within the network between two customer premises.

Traditional BERT

When a new access connection is provisioned between a customer site and a service provider’s network edge switch, a copper or fiber-optic cable is installed. The physical cable runs from a demarcation point in the customer’s telecommunications room to a patch bay within the closest CO site (Figure 1).

Then, the LEC performs an acceptance test on the physical media to ensure it can reliably carry user traffic. The test is called a BERT. The BERT can be performed by the installer at the customer site or by a technician at the CO.

Before running the test, the transmit and receive conductors are connected together on one end of the cable to create a loop back. Then a tester is connected to transmit and receive pairs at the non-looped end of the cable.

A repeating or pseudorandom pattern of binary ones and zeros is transmitted by the tester and looped back toward the tester’s receiver at the far end of the circuit. The tester will compare the transmitted bit stream to the received bit stream and count a bit error if a binary one was expected but a binary zero was measured and vice versa.

A BER is the ratio of bit errors to the total number of transmitted bits. Typically, a BERT is run over a period of time (from 15 minutes to 24 hours) to obtain a specific BER.

Acceptable BERs vary by location and intended use. The ITU Recommendation G.821 (08/96) specifies that a 1-s time period with a BER >1 x 10^-3 (1 bit error in 1,000 bits) for an international DS1/E1 circuit is considered a severely errored second. A toll- quality voice circuit should have a BER of no more than 10^-6.

Additional errors may produce audible clicks in the sound. For constant bit-rate ATM applications such as real-time interactive video conferencing, a BER of less than 10^-9 or 10^-10 is required. Performance standards for higher-speed links are specified in ITU Standard G.826.

ATM BERT

Traditional BERT is performed only at the physical layer over access circuits up to the first ATM edge switch. In contrast, ATM BERT is done at the ATM layer and can be performed all of the way across the virtual circuit to the far-end customer-premise site. ATM BERT allows for end-to-end verification of the circuit’s reliability.

The BERT pattern is carried within the payload of the ATM cells that the tester is transmitting. The payload content of cells is never examined by ATM switches, allowing the bits in the BERT stream to traverse the public ATM network transparently. Figure 2 shows how ATM BERT cells travel across a public ATM network to the far-end customer premise.

If the operator has entered the correct target virtual path and circuit identifier values, the cells will be switched through the public network like normal user traffic. When the test cells reach the far-end customer premise, the cells are looped back toward the originating tester.

After the tester receives the test cell, the BERT pattern is extracted from the payload and compared to the transmitted pattern in the same manner as in the traditional BERT. When run over a period of time, a useful BER is obtained for the entire path of the virtual circuit end-to-end.

When performing an ATM BERT, the tester transmits the test cells to match the specified service-contract rates. The data rate and the service category specified in the service contract, such as constant bit rate, variable bit rate, or unspecified bit rate, are entered by the operator. The BERT is run under the maximum load conditions guaranteed by the service provider to ensure error-free, end-to-end service before real user traffic is placed on the link.

Often, user traffic may be routed through several CO sites before reaching the first ATM edge switch. Each of these additional routes adds to the potential for errors. Traditional BERTs may only verify the local loop to the first CO and neglect to verify the connections between the intermediate COs.

An Example ATM BERT

Figure 3 is a sample ATM BERT scenario. The customer premise, shown on the left, has had a new access circuit installed, and the traditional BERT has been performed to the local CO. The customer premise shown on the right is the company’s headquarters and already has an existing circuit to the local CO that is in service to a number of remote sites. The tester is connected to the demarcation point where the new access circuit has been installed in the remote office. The BERT parameters then are configured, including BERT pattern, data rate, service category, and target virtual- circuit number.

A loop back must be installed at the far-end ATM access device. Two common methods of inserting loop backs are used. The first simply installs a physical loop-back plug at the customer-premise ATM access device on the far end of the virtual circuit. Another method uses management software to temporarily configure the far-end ATM access device’s switch table to loop back the virtual circuit to itself. In this example, we would use the soft virtual-circuit loop-back method because a hard loop back would take the circuit out of service.

Now the operator transmits the encapsulated BERT pattern and starts measuring the received bits. Connectivity through the network can be ensured by noting whether or not the tester can synchronize to the receiving pattern.

The operator can note if there are immediate gross errors that would preclude continuing the test. If not, the test will be left running for 24 hours.

Table 1 shows sample test results. For example, there was a single bit error in more than 4 billion bits transmitted over a 30-min time frame. If a 1-bit error is divided by the total number of bits transmitted, the BER is 2.4 × 10^-10. If the results of the ATM BERT are within the service provider agreement, then the network engineer can confidently cut over end-user traffic and put the circuit in use.

Isolating Errors

If a virtual circuit has bit errors in excess of the acceptable limit, the service provider can isolate the problem to a single physical link between any two network devices using the ATM BERT. To do this, a tester is connected at the new customer installation site or at the edge switch where the new customer is connected (Figure 4).

The service provider then configures a virtual-circuit loop back at the switch closest to the far-end customer site (the egress switch). Then the ATM BERT is performed again for 15 minutes to remove the far-end tail circuit as the source of the errors. If the test passes, the problem lies in the far-end tail circuit. If the test fails, the problem is somewhere between the tester and the egress ATM switch.

The next step is to remove the near-end tail circuit as the source of the errors. To do this, the ATM ingress switch loops back the virtual circuit to the tester, and the ATM BERT is run again. If the results pass the test, the problem is between the ingress ATM switch and the egress ATM switch, sometimes referred to as the network cloud.

Isolating the problem within the network cloud is a matter of configuring virtual loop backs into each switch between the ingress and egress ATM switches for the virtual-circuit-under-test and repeating the ATM BERT. Eventually the ATM BERT will fail, and the problem link will have been isolated.

There are several possible causes of bit errors. Common sources of poor transmission quality include:

Weak or attenuated signal strength.

Marginal equipment transceivers.

Timing errors or jitter.

Improperly terminated cables.

Excessive bending radius (fiber optic).

Environmental (copper).

Equipment and cable aging.

Conclusion

The number of subscribers to public ATM networks is predicted to double in 1998. Applications such as medical imaging and video conferencing will demand nearly error-free circuits between subscribers. ATM BERT ensures that the entire path between ATM service subscribers is clean, not just the local access circuit. In addition, a simple, straightforward process isolates failures to a single link anywhere in the virtual-circuit path using an ATM BERT.

About the Author

John Giles is the R&D manager for the Fluke Networks Division, responsible for the design and maintenance of the division’s research and development network. Before joining Fluke, Mr. Giles was a data communications technician at GTE. Fluke, P.O. Box 9090, M/S 279F, Everett, WA 98206, (425) 356-6258.

Error Rate

2.41692E-10

Errors

0000000001

Total Bits

4137503000

Errored Seconds

0000000001

Elapsed Time

0 Days 00:30:00

Table 1.

Glossary of Terms