Organizations have compelling reasons for prioritizing network access for certain applications and users. You can never be too rich or too thin, the socialite Charlotte Ford once said. In the Age of the Internet, she might have added: Or have too much WAN bandwidth speed.

WAN capacity is still the largest line item on many IT organization budgets, despite years of prophecies that one technological breakthrough or another would make wide area links too cheap to meter. Of course, these prophecies have a hint of credibility, since LAN bandwidth is actually too cheap to meter. As IP-based networks become the pervasive medium for commerce and communication, the costs of wide area links are bound to come down. But between bandwidth-hungry applications and ever-growing lists of network-enabled tasks and activities, the day of negligible-cost WAN links continues to recede into the future.

In the meantime, the incentive to manage available bandwidth remains strong. Without any form of bandwidth management, the order that packets flow across a WAN link is, at best, random. At worst, large batch-like flows, such as HTTP or FTP traffic, can temporarily block delay-sensitive transaction flows or wreak havoc on jitter-sensitive multimedia traffic.

To some extent, the need to prioritize traffic flows can be overcome by throwing bandwidth at the problem. If the WAN link has as much throughput capacity as the LAN feeding it, the incentive to manage the traffic is minimized. But few organizations can justify the expensive high-speed equipment, not to mention the recurring costs of high-speed WAN connections, when the careful deployment of bandwidth management solutions can provide equally satisfactory levels of service.

The problem of unmanaged bandwidth stems from the process of network integration that has occurred since the rise of the Internet. At one time, specific applications such as inventory control or order entry had their own dedicated links from branch offices to corporate headquarters.

Now, not only do multiple applications from various parts of the enterprise traverse the WAN link, but a lot of flow-priority traffic flows do so as well--and there is no correspondence between the demands a given flow makes on the network and the importance of the flow to an organization's business objectives. As enterprises move to fill up their unused WAN capacity with voice and video traffic, the problem will only get worse. What is needed is some method of controlling the several factors that make up service quality.

The Trouble with Traffic There are four measurements of service quality in IP-based networks: reliability, delay, jitter, and bandwidth.

A network's reliability is a composite of several factors, including transmission errors, which corrupt packets; discarded packets, which result from network congestion; and routing errors, which cause packets to arrive at their destination out of sequence. Delay (or latency) isthe measure of how much time it takes a packet to transit the network from sender to receiver.

Jitter is the variation in transit delay; a jittery connection will have widely varying transit times. Bandwidth is the upper limit of how much traffic can flow from sender to receiver. (The possibility that multiple flows may be contending for bandwidth at any given time must be factored into this measurement.)