We configured our lab environment so that half the lab represented a remote connection, such as a remote WAN site or the Internet. The other half of the LAN was modeled after a corporate network. We used Cisco 2948G-L3 switches to provide workgroup connectivity at each end of the network. These connected on the LAN side to a Cisco Catalyst 6500 and on the WAN side to an Extreme Summit 7i.

We used Chariot to generate a typical mix of corporate traffic. Our simulated traffic consisted of several types of traffic -- Web (HTTP GIF and TEXT), e-mail (POP and SMTP), streaming media (RealAudio streams over RTP), file transfers (FTP) and mission-critical transactions (SAP R/3 login and purchase order creation). Our mix of traffic was designed to saturate the pipe and force the products under test to manage that traffic so that no traffic type was starved for bandwidth -- especially the mission-critical transactions.

Performance Charts » Response Times At 2 Mbps (E1) » Response Times At 45 Mbps (T3) » Jitter Rate At 2 Mbps (E1) » Jitter Rate At 45 Mbps (T3)

During our baselines we noted which types of traffic were hogging the most bandwidth -- HTTP and SAP R/3 dominated the traffic, taking almost 80 percent of the bandwidth between them at both T3 and E1 speeds. We also noted the response times for all TCP-based traffic and jitter and data loss rate of the streaming traffic.

Each bandwidth management device was placed in line between the router and the workgroup switch on the corporate LAN. We configured each device such that each traffic type would be allowed to use only a percentage of the total bandwidth -- Web (25 percent), e-mail (5 percent), SAP R/3 (25 percent), streaming media (25 percent) and FTP (20 percent) -- and then ran our tests again. We monitored real-time data through each product's management interface and then correlated that data to the results delivered by Chariot.

One of the most important facets of our performance testing was to determine whether each product properly managed the bandwidth during testing. It was acceptable for a class of traffic to receive more than its configured bandwidth only if all other classes of traffic received at least its minimum allocated amount of bandwidth.

We evaluated each product in terms of configuration and management during the testing as well as the performance capabilities of each. We looked at each product's traffic management technologies, the control offered over that technology, as well as additional features offered by the products.