

Mariposa, 3Com Raise the Bar For Next-Generation ATM Access
March 22, 1999
Finally, we invited newcomer Mariposa Technology to unveil its first product, the ATX-100. This device gave its more established rival products a swift kick in the pants and earned our Editor's Choice award. It exhibited consistently high performance and a management interface as easily understandable as the best FRADs. Although we'd like to see the unit add a few more features and interfaces, the features that the Mariposa ATX-100 offers held strong and took our top honors.
3Com's SuperStack II PathBuilder S330 made a very good showing, falling a close second behind Mariposa in our evaluation. As a more mature offering, the PathBuilder provides a wide range of interfaces and software. Standout features include very deep buffers, powerful queue management and straightforward configuration. Best of all, the 3Com box is the only unit to offer voice compression without driving the cost over $10,000. Although the compressed voice quality could use some improvement, the 3Com PathBuilder takes Network Computing's Best Value award.
We invited several other companies to participate in these tests. Most notably, Cisco Systems declined, citing product availability constraints. Although we could have easily located a pair of Cisco MC3810s installed at customer sites, we honored Cisco's request not to put the MC3810s in a comparative performance test (draw your own conclusions). Others, including Accelerated Networks, ADC Kentrox, FVC.COM, Mariner Networks and Nortel Networks, did not have shipping products ready in time for our tests. FORE Systems also declined, citing its lack of compressed voice service as the primary reason.
Our performance tests employed a series of voice and data traffic measurements, taken alone and in combination (see "How We Tested," page 58). We discovered marked differences in the default handling of low-priority UBR (unspecified bit rate) streams (see "ATM Access Concentrators: Throughput vs. Latency," above left, and "Combined Voice-Data Test," above). For example, 3Com's philosophy appears to be to "pass cells at all costs"--regardless of how long it takes traffic to get there--while GDC and Mariposa choose to keep latency consistent and discard cells under load. Either approach is valid, but we lean toward the latter method, opting for more predictable service at the expense of lost packets.
|