Application Performance on Your Network
Bill Alderson, Technology Consulting Officer, NetQoS, Inc.

Some applications, such as file-transfer applications, need bandwidth but can tolerate latency. Others, such as database queries or voice over IP, need less bandwidth but cannot tolerate latency. When these applications are slow or when transmission times vary too much, users complain. Multimedia streaming applications need bandwidth and a steady data-transfer rate to prevent looking jerky or sounding garbled.

The mix of protocols used by network applications can impede application performance. Most business applications and Web transactions use TCP while multimedia and voice over IP traffic uses UDP or RTP. TCP applications gracefully bow out when faced with congestion, but UDP and RTP applications do not back down, they just drop frames. Important business traffic might grind to a halt when users tune in to Internet radio at their desktops.

Tools that test and monitor applications running on your network can help you assess network performance from a user's viewpoint. Although a protocol analyzer can find CRC errors and runt frames, it is equally important to gather application-layer metrics, such as throughput, response time, jitter, and lost data. If email has slowed to a crawl and multimedia applications are experiencing jitter, it is not necessarily a lower-layer problem.

Application performance testing and monitoring software can send simulated application traffic anywhere in the network, measure the performance, and report the results with highlighted locations and degrees of severity. By contrast, protocol analyzer traces can be hard to interpret because routers remove the data-link control number that identifies a particular user's computer and substitute their own DLC numbers.

Application Performance monitors keep constant tabs on end-to-end connectivity and application availability and indicate whether a problem is caused by a client, the network, or a server. You do not need to keep plugging a device in to different segments to find out where a problem is occurring. You can monitor traffic with a variety of frame sizes, protocol types, and transactions to see how different combinations of applications perform on the same LAN or WAN segment. When gathered over time, performance statistics can alert you as soon as an application's performance starts deteriorating.