What Is IP CoS, and Why Do I Need It?
Simply put, CoS provides a mechanism by which certain packets are afforded preferred treatment in an effort to provide the associated application with a level of performance required for proper operation. Although the preceding sentence seems simple enough, it implies support for several capabilities that must work together within each node—and in a consistent manner network-wide—for an IP CoS deployment to be successful.
Why IP Networks Need CoS
IP networks are based on the principal of statistical multiplexing (stat MUX), which is a resource-sharing technique that allocates resources on an as-needed basis. A stat MUX provides efficiency gains by playing the odds that a given application or user will not be active at its peak rate 100% of the time. By allocating bandwidth resources only when needed, a large number of bursty applications can be supported over a network with an aggregate capacity that is significantly less than the potential aggregate rate of its user base.
To make all this work, some degree of buffering is needed to accommodate the occasional synchronized bursts. Because no network has infinite buffers, flow control (typically supported by a virtual circuit [VC] technology) or simple discard in the case of datagram operation (connectionless) is needed during chronic periods of congestion. Throwing more buffers at the problem only changes the symptom from one of discard to one of delay and delay variance, which is known as jitter ...