Next, we analyzed the dynamic bandwidth allocation capabilities of each device. Using the Wandel and Goltermann DA-30 Internetwork Analyzer's Flexmit application, we sent a steady stream of 1,000-byte pings across the network at 1.4 Mbps. We then repeated the AM2D voice calling test, capturing data from the ingress to MCI WorldCom's ATM backbone with an HP 7500 analyzer (see "Dynamic Bandwidth Allocation," Above).

We then captured a data-only baseline, using the DA-30 RTBench Application. RTBench generates IP packets of various sizes, increasing the injection rate in 10 percent increments, up to our target maximum of 1.3 Mbps. RTBench reports throughput at the maximum lossless level--the highest point at which 100 percent of submitted packets are still being received. Latency calculations are also made at this level (see "ATM Access Concentrators Throughput vs. Latency," to the right). To analyze data performance with voice calls present, we reran the RTBench test--this time with the Ameritec AM2D holding all 24 voice channels open, except, as noted, with the Lucent box (see "Combined Voice-Data Test," below).

Finally, we analyzed voice quality with calls placed with an Intecom PBX, while capturing bandwidth utilization in the WAN with the HP 7500. To examine voice activity detection (silence suppression) efficiency, we disconnected handsets from the telephones and watched utilization on the HP 7500.