Try out our Interactive Report Card ®!
For more information about our Interactive Report Card ®, click here As ATM slowly builds momentum, you may be feeling the pressure to at least start experimenting with this technolo gy. Some have been experimenti ng with it for the past year or so and are starting to implement ATM production networks. Whatever your situation, you may be wondering

what you're going to do to ensure that your network runs smoothly enough to take full advantage of ATM's capabilities. An ATM analyzer may be just what you need.

Until recently, most ATM analyzers have been geared toward developers needing to simulate and test various ATM events at the bit level. But as ATM migrates to mainstream networks, there's a greater need to perform troubleshooting at higher network levels. The products that we evaluated reflect the transition: They all have simulation capabilities except for the Network General Sniffer, which should have this feature by the time this review goes to print.

Due to the immaturity of ATM products and standards, simulation capabilities still are useful for implementers who just want to build off-the-shelf products, especially if they plan on mixing vendors. Simulation capabiliti es can be deployed to simulate thi ngs like the most recent version of UNI or NNI signaling to test a vendor's compliance. Simulation also can be used to evaluate the ability of an an ATM switch's to perform worst-case scenarios of processor-intensive tasks such as call setups.

You also may need to use an ATM analyzer for more basic troubleshooting tasks such as monitoring for errors on the physical and data-link layers. In some respects, this is no different than what you would do with an Ethernet or FDDI analyzer. On the other hand, you will have a whole new set of protocols and associated errors to learn about before you will be able to understand their significance. Even if you don't yet have ATM in a production network, the sooner you get hold of an analyzer and start observing how it works, the more confidence you will have when it's time to go in and troubleshoot a real problem.

Help Where You Need It Of course, just because you implement ATM doesn't necessarily mean that network pr oblems related to slow file servers an d misbehaving client software will go away. What are you going to do if your users suddenly start complaining about time-outs and slow server-response times, while the systems administrator insists that server utilization is barely in double digits? Or, suppose you are having problems with the latest client/server application when it makes calls over the network? You still will want to be able to see what the upper-layer protocols are doing, in some cases just to prove that the suspicious new ATM technology is not the problem.

At the top of the heap is the Tekelec Chameleon, the most well-rounded product that we tested. It has quite sophisticated simulation tools and also does a nice job monitoring statistics, as well as capturing and displa ying both ATM and legacy protocols. If you cannot afford its hefty price, you might be able to get by with the RADCom RC-200-C. It provides many of the same features as the Chameleon, although its performance claims do not come c lose to the Chameleon's full-bandwidth c laims.

The Network General Sniffer did poorly in our ratings because it lacked simulation capabilities. It retains many of the strengths that have made it the leader among legacy protocol analyzers, however, and it may suit your needs if you require only basic monitoring and troubleshooting.

If full-bandwidth analyzing and simulation are important, another alternative is the Net2Net CellBlaster. Although the software still appears to be a bit immature, Net2Net has taken great pains to ensure high standards of performance and no other vendor comes close to Net2Net's claim of 33-nanosecond time-stamp resolution.

Tekelec Chameleon Open-X

Tekelec Chameleon Open-X ships in a portable unit that can be crammed with just about every type of ATM interface that you could imagine, including T1, 25-Mbps and TAXI connections. It also can be loaded with legacy LAN and WAN interfaces. But your pockets will have to be crammed with cash to afford the versi on that we tested--it had only two OC3 interf aces. Unless you really need sophisticated simulation capabilities, this might not be your best choice.

The Chameleon Open-X is packaged in its own portable carrying case with an active matrix display that is one of the sharpest you'll find. It runs atop the Unix OS, which can be accessed from the Trigger applications, and each interface card has its own Sparc Lite Processor.

To monitor a full-duplex ATM circuit simultaneously in both directions, you will have to purchase two ATM boards, which add to the cost. The only other unit with this requirement is the CellBlaster, which will be upgraded by the time this is printed--one board will enable full-duplex monitoring. Currently, with one board you can observe one direction of the circuit at a tim e. The Chameleon also supports Ethernet, FDDI, Token-Ring and WAN interfaces in the same box. The only other box that could do this was the Network General Sniffer.

The Chameleon has excellent decodes of ATM an d legacy protocols and displays them in a sharp, color-coded format second only to the Sniffer. We were disappointed, however, that the decodes were displayed as one long file with frames and cells separated by dashed lines. It was not possible to display a one-line summary and a decoded version at the same time.

The statistics application can display major Sonet connectivity statistics as graphical lights. You will need this when tapping into an ATM circuit since there are no lights on the card to indicate when you have connectivity. This could slow you if you want to quickly tap into a production network because you will have to look at the front of the box to tell if you are connected. Although awkward, this setup is a little better than RADCom's, which shows only a confusing screen of Sonet counters.

The well-designed statistics window also displays up to three easy-to-read bar graphs and two line graphs, along with counters showing average, peak and cumulative numbers. You can easily customize the screen with the statistics and formats that you want. Switching amoung the six different displays is a snap and various views can be saved to disk. We were disappointed that there were no LANE statistics and no call-setup statistics that could be displayed. Many of the statistics are probably better-suited for a WAN or backbone PVC setup.

The ATM UNI simulator application comes with a tool kit consisting of icons representing various states and actions to be performed in an ATM network. This makes it possible to build complex testing scripts without having to master a scripting or macro language. The ATM events that can be scripted will automatically be compliant with the protocol that you select. When the tests are run, the results can be displayed on the screen and logged to disk. Additionally, if you want to manipulate the cells at the bit level to introduce errors, they can be loaded into an editor for a starting template.

Radcom RC-200-C

The RADCom RC-200-C is armed with a whole arsenal of weapons to deal with applications on yo ur ATM network, from simple monitoring to sophisticated simulation. It allows you to emulate an ATM switch or an end node using UNI 3.0 or 3.1, and it's the only product tested that claims to support UNI 4.0. The decodes and simulation were not quite as good as Tekelec's, but the package also is significantly less expensive than Telkelec's.

The RC-200-C, like the Net2Net CellBlaster, relies on an Intel i960-assisted ATM NIC that can be used with any PC that can run the Microsoft Windows-based user interface. Unlike the CellBlaster, however, the interface and processor are mounted in an external pod about the size of a toaster oven, attached via parallel port.

It is possible to run multiple applications from the MS Windows interface; however, we did find it to be a bit cluttered on the 14-inch VGA monitor that we were using once several apps were launched. We also found that the icons for the analyzer, capture and simulation apps would easily dis appear behind the open windows. We recommend running the so ftware under Windows95 in order to take advantage of the Task Bar which greatly improves this situation.

In addition, we discovered that the software was sluggish running on a 486-50 SX notebook with 12 MB of memory. Strange, since the notebook is only displaying data. The pod does all the data gathering and filtering. We also found it a little difficult and confusing to navigate back and forth between various configuration windows necessary for setting up some of the applications.

RADCom has stated that it will accurately capture cells at about 70-percent utilization. And it will soon release a new model, designed to capture and generate at 100-percent utilization and include indicator lights.

The capture application displays all the decoded packets and cells in one window similar to the Tekelec. It also provides some color coding. The Tekelec, however, was much more readable. But all is not perfect--we did discover problems with some of the translations. Although it did decode NCP frames, when an NCP frame appeared in burst mode, it merely flagged it as a Burst Mode frame from the IPX layer without providing any other information. Another annoyance was that you are forced to choose one active protocol decode at a time. All of the other products automatically decoded the different protocols simultaneously.

A new ATM Expert application allows post-c apture analysis of ATM control cells. It will determine VPI/VCI pa irs associated with source/destination NSAP addresses, as well as call setup and teardown times for SVCs and LANE control operations. Various graphs of this information also are provided.

The RAD-200-C comes with a toolkit that allows you to simulate ATM and signaling events and record the results. Although the scripting was not quite as deep as Tekelec's, it does allow you simulate either an end node or an ATM switch.

Network General ATM Sniffer

Network General clearly had the end-user market in mind when it developed the ATM Sniffer product. It offers very good LANE statistics and, of course, retains its unrivaled legacy decodes, which it applies to frame views. It was hurt in the ratings because it currently does not provide any simulation capabilities (stil l in beta while we were testing). The ATM Traffic Generator Option, which should be released by the time you read this, will allow generation from one ATM port to another monitor port at a cost of $5,000.

A small pod about the size of a paperback novel is used to tap into the ATM circuit. The pod attaches to the PCI card mounted in the Dolch portable PC. Tapping into a circuit this way was the most straightforward of all the products. You simply plug each cable pair into one of the ports (they have indicator lights), and you can see immediately when you have a good physical connection. There also are indicator lights that tell you if you are in monitor or emulation mode.

When you launch the Network General software you have the option of using the Discovery application that displays color-coded summaries of SVCs and PVCs as well as closed SVCs. The VPI/VCI, protocol, AAL layer and source/destination ATM addresses also are displayed.

LANE statistics, OAM statistics or a four-layer summary of statistics, including a summary of the use of major protocol suites, can be displayed while capturing. This helped compensate for the fact that this is the only prod uct that does not have a Windows-based GUI capable of displaying multiple applications.

If you are already familiar with the Sniffer interface, you may be accustomed to setting an almost limitless combination of packet filters. Unlike any of the other products, however, once the packets are captured and assembled into frames, you have the usual complete range of predefined post filters as well as the ability to easily define your own complex pattern matches.

If you only want to passively monitor ATM traffic and plan on performing a lot of troubleshooting above the ATM layer, then the Sniffer is a good bet. If you want to be able to do some simulation, then check out the ATM Traffic Generator option. But if you need to perform complex simulations at full bandwidth, then you might be better off with the RC-200-C, CellBlaster or, if you can afford it, the Ch ameleon Open-X.

Net2Net CellBlaster

Net2Net has taken great pains to engineer high-performance hardware that guarantees full-b andwidth capture and generation of ATM cells. It was the only product that clai med to have an interpacket time-stamp resolution as low as 33 nanoseconds. Net2Net also claims to be the only vendor that can retransmit a capture file retaining the original time stamp. If you have a need to do simulation and testing on an ATM network that requires high levels of performance, then you should take a look at this product. Unfortunately, the software functionality leaves a lot to be desired, especially for the average user transitioning from a legacy network. The software is easier to change than hardware and Net2Net states that it will be improving the CellBlaster on an ongoing basis.

Each ATM card that mounts inside a Pentium PC has two ATM ports on the back. An Intel i960 as well as seven FPGAs on each card provide the best possible speed without the need for an ASIC, which would not be able to adapt to the changing nature of ATM standards. The product that we tested required two cards to capture full-duplex transmissions. But Net2Net informed us that by the time this goes to print, it wi ll be able to provide this capability on one card, while at the same time maintaining its high performance standard.

The CellBlaster cards have clear indicators on the back that inform you immediately if you have connectivity. The monitoring application also had a series of Sonet lights that are very readable. Although the activity indicators are readable, the statistics monitor is not. It uses exponential notation that's difficult to read at a glance. With this software, it is also possible to change the interface on the card to shift to a different media. It has good OLE support, so it's possible to write custom Excel calls to obtain better graphing and statistics viewing.

The CellBlaster was the only product, aside from the Sniffer, that showed a summary view and a detail view at the s ame time for captures. The decodes however, rarely went above Layer 3 and the post-filtering capabilities were very limited. It was al so the only product that could be queried via SNMP.

Peter Morrissey is a network systems programmer at Syracuse University. He can be reached at ppmorris@mailbox.syr.edu.

The first thing that we did when testing the ATM analyzers was to set up an ATM LAN Emulation network in our Syracuse University lab. The network used only FORE Systems products including an ASX-1000 ATM switch, ATM network interface cards and an ES-3810 Ethernet switch with an ATM/LANE uplink.

We set up a Windows NT 3.51 server and a Novell NetWare 4.1 server directly on the ATM backbone and accessed the servers from clients located on Ethernet ports on the ES-3810. Although it was a learning experience, we did manage to get everything running smoothly after just a few days of work. Having one vendor as a point of contact definitely made the task easier than it might have been with multiple support lines to call. Additionally, this allowed us to avoid the fi nger-pointing contests that often result with multiple vendors.

Installing the ATM NIC in the NT server was no different than installing an Ethernet NIC. We had difficulties with the Novell server that turned out to be a result of using the wrong driver from the installation diskette. Tech support was able to eventually get us past what turned out to be misleading instructions in the manual. At that point it was simply a matter of loading some NetWare Loadable Modules (NLMs).

We configured the ASX-1000 and the NICs for one "default" emulated LAN (ELAN), which also simplified matters. Having a solid understanding of LANE operations didn't hurt, either.

Interestingly, we did not make much use of the analyzers while we were initially setting up the network. This was, in part, due to the fact that we were not able to get a good understanding of how they worked until we had a functioning network. It was also because we were able to use the ASX-1000 console, as well as diagnostics that came with the N ICs, to observe the status of the LANE connectivity.