Do You Speak ATM?

Anyone who has tried to learn about asynchronous transfer mode (ATM) will appreciate the wide variety of available training described in “Education a la ATM Forum” by Jill Kaufman of IBM.1 At the end of her presentation of courses, computer-based training diskettes, expert speakers, and CD-ROM material, she posed the question, “Do you speak ATM?” This is the question for technicians, test instruments, and networks to ask each other—if their conversations were in English.

Within the ATM community, the accepted vocabulary is as extensive and unique as the structure of ATM itself. Like other languages, it continues to grow, in this case reflecting some of the new services that ATM supports. The ATM Forum Glossary is the official dictionary and available online at

Nearly 600 definitions of two-, three-, and four-letter acronyms, terms, and specifications give you an idea of how complex ATM really is. Because there are so many highly specialized aspects to the subject, an initial answer to the question “Do you speak ATM?” is easy. General-purpose oscilloscopes, logic analyzers, and many protocol analyzers don’t. Properly trained engineers and technicians, dedicated ATM testers, built-in remote monitoring (RMON) equipment, and some protocol analyzers do.

However, the world is not yet entirely ATM-enabled, nor do older networks disappear immediately as ATM is introduced. Dave Ushler, product marketing manager at Digitech-LeCroy, commented, “Whether or not application-specific test equipment is more advantageous than multipurpose tools depends upon the network environment. Theoretically, application-specific test equipment should offer more in-depth analysis of ATM in a particular application. But, this may not be the case when dealing with a hybrid or legacy network where other topologies are in use and a test tool with a broader application range would be more beneficial.”

Primary Areas of Test

“The most critical application-specific tests are those that target the unique features and mechanisms of ATM,” commented Bruce Barrell, ATM product marketing manager at Hewlett-Packard’s Network Systems Test Division. “These include quality of service (Qos), policing (monitoring traffic to ensure it is within contracted levels), and signaling/local area network (LAN) emulation.

“ATM testing must have provisions for measurement of QoS parameters such as cell loss, cell delay, delay variation, and cell misinsertion,” he continued. See Table 1. “None of these characteristics can be tested with generic, multipurpose testers. Application-specific testing is required for R&D testing as well as for installation and maintenance testing.”

Most of these ATM layer tests appear quite basic; however, they can be awkward to perform under some circumstances. For example, you generally want to know CDV rather than the CTD. But in applications where knowing the cell delay accurately between two distant parts of the network is important, you need to use a global positioning system (GPS) time reference.

Also, test difficulty increases when dealing with switched virtual circuits (SVCs) instead of permanent virtual circuits (PVCs). SVCs are set up and torn down dynamically in response to signaling commands, but PVCs are set up by the network management function and remain indefinitely.

According to Oded Agam, director of technical services for RADCOM in North America, “When testing ATM SVCs, you need to address the same issues as with PVCs, but the connectivity issues become complex. They include the actual connection as well as the performance of the equipment enabling the connections to be made between different ATM end stations.

“When SVC environments are scaled, the need for private network-network interface (PINNI) becomes obvious,” Mr. Agam continued. “Testing PNNI networks is very expensive and time-consuming. A switch vendor or a service provider needs to test the behavior of a switch or a network in a multiswitch, multihierarchical environment with dozens of other switches and ATM end stations.”

It is the very good fit of ATM to certain types of applications that has emphasized the need for these kinds of tests. ATM adaptation layers (AALs) massage data from different sources into standard 53-byte cells.

AAL-1 deals with constant-bit-rate (CBR), time-dependent traffic such as voice and video. AAL-2 still is undefined but may be used for variable-rate video. There is some work being done with AAL-2 for voice transmission. AAL-3/4, previously two separate AALs, handles variable bit rate (VBR), delay-tolerant data traffic requiring some sequencing, or error-detection support.

AAL-5 supports VBR, delay-tolerant connection-oriented data traffic requiring minimal sequencing or error detection such as MPEG-2 digital video.

1 As well as CBR and VBR service classes, there also are unspecified bit rate (UBR) and available bit rate (ABR) classes.

In addition to tests at both the physical and ATM layers, Digitech-LeCroy’s Mr. Ushler noted that testing at the adaptation layer should include the frame forwarding rate (FFR) or frame throughput; latency; frame loss; burst depth; broadcast storm tolerance; numbers of cells, protocol data units (PDUs), bad header error control (HEC) code, and cells tagged for cell loss priority (CLP); and the minimum, maximum, and average cell rates.

When the traffic contract is made between the user and the network, the user must specify what type of traffic will be sent over the virtual channel (VC). The user will require guaranteed limits on CLR and CDV, depending upon the application. On the other hand, if the network can accommodate the required QoS, the user should not exceed the agreed bandwidth and burstiness levels.

This is where policing comes in. It is all very well to operate on trust alone, but it is better to have some verification. At the edge of the network, each connection is monitored for traffic levels exceeding those agreed upon. Should the rate of cells from a user become excessive, some cells will be dropped. This approach sounds drastic but it is the network’s means of protecting all nonoffending users from degradation of their own QoS levels.


The cost of QoS depends upon bandwidth—the portion of the network’s capacity that an application requires. It will be more expensive to allocate bandwidth sufficient for the peak cell rate (PCR), for example, than for the sustainable cell rate (SCR). If the application can stand increased CDV, a traffic shaper can average out the peaks and help obtain lower-cost transmission.

“A tremendous amount of field ATM testing will involve the physical layer,” said Doyle Mills, the senior technical support engineer at Digital Lightwave. “The first test done on a newly installed ATM network element is a simple continuity test. ATM cells occupying the full bandwidth, normally only on one virtual channel connection (VCC), are sent through each port of the device to determine a basic capability to transmit and receive cells. Each new component in the network must be verified physically before any higher level in the protocol stack can be tested.”

Fluke prioritizes testing by starting with the physical layer, progressing to the ATM layer, and then dealing with QoS testing. According to Molly Myers, product manager at Fluke’s Networks Division, “First, the physical layer is monitored to verify that no physical alarms or error conditions are present. Secondly, the ATM layer is tested. Correct ATM cell format and ATM layer framing/cell delineation are verified.

“Next the service class is tested,” she continued. “A test unit transmits a traffic stream defined by the user’s traffic contract, including the contractual peak cell rate, sustained cell rate, and maximum burst size. Finally, a bit error rate test (BERT) is performed to provide an overall bit error rate for the ATM circuits under test at the specified service contract rates.”

Test Systems and Equipment

ATM is being used at 155.52 Mb/s (OC-3) and especially at 622.08 Mb/s (OC-12) to aggregate a variety of LAN and wide area network (WAN) traffic. Consequently, the need arises to verify that the original data has been correctly adapted to run on ATM, that it has been correctly transported, and finally that it has been converted back to its native protocol. This is the reason why many ATM testers include at least two interfaces—one for ATM and another for Ethernet or frame relay, for example—to investigate internetworking and interworking problems.

“Packet-layer testing primarily analyzes the applications (software) running over an ATM network,” said Avi Zamir, president of RADCOM Equipment. “When cells arrive at their destination, the ATM hardware reassembles them into packets according to one of the ATM adaptation layers. The resulting packet may be a high-level data structure such as an internet protocol (IP) or internet packet exchange (IPX) packet or an instruction to the ATM device such as an SVC setup message.”


Because ATM is complex, suites of standard tests have been developed. If you make ATM equipment, using these tests allows you to certify that your product meets a defined set of criteria. Such a set of tests also eliminates much of the labor required when you have to test the performance of an ATM network.

According to information presented in an Adtech brochure, there are major differences between ATM testers and protocol analyzers. The company claims that protocol analyzers focus on network management problems by analyzing upper-layer protocol encoding and decoding and handling error and fault conditions. Much of the testing can be done in software. On the other hand, an ATM performance analyzer is hardware-based because basic transmission performance has to be measured in real-time at full rates.

Sometimes, it is a case of semantics: When is a protocol analyzer with hardware ATM capabilities an ATM tester, and when is it a protocol analyzer? HP and RADCOM claim full seven-layer decoding and real-time operation for their protocol analyzers. They both deal with OC-12 rates and can analyze or generate ATM cells in real-time.

In the portable protocol analyzers that these companies produce, a significant portion of the instrument is dedicated to a hardware ATM interface. For example, HP’s analyzers use plug-in interface modules and network-specific undercradles for distinct network technologies such as ATM.


1. Kaufman, J., “Education a la ATM Forum,” 53 Bytes, The ATM Newsletter, Vol. 5, No. 3, The ATM Forum, Sept. 1997, p. 5.

1. The ATM Forum Glossary, The ATM Forum, May 1997.

2. Asynchronous Transfer Mode Switching at 622 Mb/s: A Multiservices QoS World, Hewlett-Packard Tutorial, 1998.

3. Zamir, A. and Cerequas, F., “ATM Testing Evolves to Measure Real-World Applications”, EE-Evaluation Engineering, July 1997, p. 145.

NOTE: This article can be accessed on EE’s TestSite at Select EE Article Archives and use the key word search.

ATM Test Products

Comprehensive ATM Modules

A range of AX/4000 Version II Modules is used in either a four-slot portable enclosure or a 16-slot, rack-mount mainframe. The modules include a generator with eight simultaneous traffic sources, cell and frame sequencing, and error injection. A combined generator/analyzer provides similar generator functionality and automatic identification of up to 2,040 different substreams with simultaneous, continuous, full-rate analysis for each one. A combined generator/dual analyzer deals with bidirectional communications monitoring and analysis. ATM, frame relay, and other broadband networks can be addressed at rates to 155.52 Mb/s. Contact company for price. Adtech, (800) 348-0080.

PC-Based ATM Analyzer

The ATM900 Protocol Analyzer provides 100% capture of full speed OC-3 ATM traffic. Real-time filtering by virtual channel or path identifier (VCI/VPI) restricts capture to only the required data, simplifying later analysis. Features include hardware counters for statistics calculations in real-time and support for T1/E1, DS3/E3, and both single- and multimode (STM-1)UTP-155 interfaces. Cell bit error rate testing (BERT) and physical-layer BERT are provided as well as detailed packet-type statistics and a configurable statistics display. The ATM900 is available as a PCI bus processor card with software or can be configured in a PC platform. Starts at $15,000. Digitech-LeCroy, (800) 821-2265.

Instrument Undercradle

An HP Internet Advisor combined with the ATM-622 Undercradle is called a 622 Vu Advisor and has the capability to work at OC12c/STM-4c transmission rates. Traffic from two full-rate ATM channels can be acquired and stored in 32 MB of memory per channel for later analysis. Real-time capabilities include filtering by VCC, payload type indicator (PTI), CLP, VCI/VPI, and good or bad HEC fields. Full application layer decoding and expert commentary are provided for post-capture data. Externally generated or previously captured and edited data can be used to simulate traffic for testing purposes. ATM-622 with multimode fiber interface: $29,900; with single-mode interface: $34,900. Hewlett-Packard, (800) 452-4844, ext. 6348.

Analyzer on PCI Board

The Wirespeed


622 ATM Protocol Analyzer performs real-time filtering, capture, and transmission at the full OC-12 rate. Automatic identification of active channels and services, 64 MB of memory, 10-ns time stamping of captured data, and collection of statistics for each VPI/VCI pair are featured. More than 300 protocols are decoded online from simple cells to hypertext transfer protocol HTTP and PNNI packets. Operating modes include simulation, bidirectional monitoring, and QoS. A Java user interface supports remote operation from anywhere on the network. Contact company for price. RADCOM, (800) 723-2664.

OC-48 Portable Analyzer

The Model ASA 312 Network Information Computer with the OC-48 Network Protocol Processor verifies and qualifies T-carrier, ATM, and SONET networks and network elements. SONET operates to 2.488 Gb/s (OC-48), while ATM is limited to STS-12c/OC-12 (622.08 Mb/s). ATM support includes adaptation layers 0, 1, and 5; traffic shaping and QoS measurements; HEC error generation and analysis; and stress testing to full OC-12 rates. Up to 250 channels can be transmitted with independent AAL, service class, and bandwidth. Separate cell count, bandwidth, and AAL5 PDU counts and errors can be accumulated for up to 250 received channels. From $20,000 to $50,000. Digital Lightwave, (800) 548-9283.

Hand-Held Analyzer

The OC3port


PLUS is a hand-held ATM analyzer capable of OC-3 (155.52 Mb/s) data rates. ATM CBR, VBR, and UBR traffic can be generated with programmable sustained cell rate, PCR, maximum burst size (MBS), and cell delay variation tolerance (CDVT). QoS measurements include CLR, CMR, cell transfer delay (CD), CDV, and CBR. In addition, an IP Ping capability and an auto-discovery mode that automatically locates all active VPI/VCI are provided. OC3port PLUS with multimode fiber interface: $8,995; with single-mode interface: $9,995. Fluke, (800) 443-5853.

QoS Parameter






Cell Error Ratio


The number of errored cells divided by the total number of cells sent

Severely Errored Cell Block Ratio


A block of N consecutively transmitted cells with about 3% lost, misinserted, or errored cells

Cell Loss Ratio


The number of lost cells divided by the total number of cells sent

Cell Misinsertion Rate


The number of misinserted cells (ones with payload information belonging to a different virtual channel (VC)) divided by the time duration of the test

Cell Transfer Delay


Round-trip delay of test cells, or delay time between two points in the network

Cell Delay Variation


Difference between min and max CTD

Table 1


Copyright 1999 Nelson Publishing Inc.

June 1999

Sponsored Recommendations


To join the conversation, and become an exclusive member of Electronic Design, create an account today!