To begin with, network performance is a process. It is not something you do once, after which you can move on to something else.  The ever-changing mix of applications, services, protocols and usage levels on networks means performance must be monitored all the time, and attended to when conditions suggest that something is out of the ordinary. Because anything and everything make its way across most networks, something new is sure to appear. Likewise, changes in technology, infrastructure, and equipment make managing performance an ongoing, omnipresent task.

Efficiency is the second component in our definition, and reminds us that best usage involves a balancing act. We don't want to purchase or operate resources we don't need or use, but we don't want to suffer when resources bog down or fail because of overuse. A good definition of efficiency in this context might be "the production of the desired effects or results with the minimum waste of time, effort, or skill." At the end of the last century, many network managers unconcerned with efficiency took a “brute force” approach to problem-solving, throwing resources at their problems until they went away. A better way is to find a sweet spot for usage that describes efficiency very nicely—namely, a situation where resources are used enough to justify their costs, but not so heavily as to overwhelm their capabilities, or threaten their continued and ongoing use. Bringing performance management to the right level of efficiency is indeed a balancing act, and can involve a lot of sophisticated tools and technology.

The next element in the definition has to do with use of network resources. This generally refers to how much of the available capacity or capability is actually being used at any given moment, and how much remains available for additional use. Network administrators already know that when a resource becomes fully utilized, additional demands for its use can no longer be accommodated, and that when 100% utilization occurs, problems or delays become increasing likely the longer complete saturation persists. Because network resources include network links, routers, switches, and so forth, keeping an eye on how heavily they're used plays an important role in keeping networks working smoothly. Making sure there's always some unused capacity (even if not immediately available) also helps to keep things that way, and is thus part of the performance management equation.

Finally, minimizing the impact of resource contention focuses on what happens when demand for services or resources outstrips the capacity to meet it. This means taking steps to avoid delays involved in waiting for service when various points along the network are congested.  Some solutions include investigating traffic and protocols in use, and eliminating or reducing unwanted or unimportant elements (thereby making more resources available for the important stuff). They may also involve categorizing traffic by time-sensitivity or priority, so that some types of traffic or application communications take precedence over others.  That way, the impact and severity of delay is lessened for the important stuff, and allowed to degrade slightly for traffic that is time-insensitive or  unimportant.

The practices of network performance management requires detailed study and understanding of the networks being managed, including the traffic, applications, and services those networks support. The better you understand what is operating on or across your network, how it's being used, and how it ranks among the priorities for your business or organization, the easier it becomes to optimize network performance where it counts most, and to make sure network resources are sufficient to meet your needs.