[This is preliminary documentation and subject to change.]
As computing and applications become more mission-critical, not to mention more content-rich and multimedia centric, the bandwidth necessary to service desktop functionality increases. However, bandwidth availability doesn't keep up with the bandwidth appetite of today's desktop applications, creating an environment where there is often more data to be transmitted than there are resources to transmit that data. The "bursty" nature of network transmissions only contributes to the problem.
Traditional data transfers are also increasing as a result of a continual addition of new network nodes to the existing network. At the crux of this problem is the fact that there is no inherent means of differentiating between "important data," such as that transmitted by mission-critical applications, and "excessive data," such as that which may be transmitted by interesting (but not necessarily critical) multimedia applications.
Such traditional business applications are continuing to increase in size and in network use, but they aren't alone in their hunger for the network. New applications are pushing network utilization to its limits and sometimes beyond; such as multimedia applications, and they're making extensive use of the network. For example, video transmission applications require significant bandwidth to transmit with acceptable levels of quality. Due to the "send as much data as you can" nature of the most prevalent networking protocol, IP, even a few active instances of these data-intensive programs can put bandwidth strain onto networks that were never designed to carry it. With data-intensive multimedia applications putting such hefty data loads onto the network, the network often becomes less available for other applications. If the load is significant enough, overall network performance will wane.
Such network performance degradation is especially threatening to real-time audio and interactive conferencing transmissions; they are time-sensitive and especially susceptible to the introduction of delays or individual frame drops during the course of their delivery. Such delays in the delivery of individual packets, known as latency, can render real-time audio and real-time conferencing applications substandard at best, and a burden to even try using at worst.
In the midst of larger traditional applications' higher network utilization, as well as increased network use by emerging desktop applications, core business application are left vying, sometimes unsuccessfully, for adequate access to the network. These bandwidth-poor and latency-laden effects of an overburdened network have an even larger and more dramatic effect on mission-critical applications that may want to make use of new multimedia desktop technology: not only to they need access to the burdened network, they need more of the increasingly precious network resources, putting core business applications that use multimedia features in double jeopardy.
If such over-subscription of available network resources doesn't paint a bleak enough picture, the prevalence of Wide Area Networking can be added to the situation, introducing an even more critical and more precious bandwidth restriction at the WAN link. With this, the situation is exacerbated.
Thus to manage oversubscription of network resources, to regulate the allocation of available network resources, and to present network data in a means more friendly to a shared network environment (i.e., in a less bursty manner), mechanisms that help manage network activity from an end-to-end perspective are needed. These mechanisms are found in the blanket technology called GQOS.