2.2 Raising the QE Product of a Network
The utility of QoS mechanisms lies in their capability to raise the QE product of a network. This is illustrated in Figure 2.2.
){kind=link}
By applying increasingly sophisticated QoS mechanisms, it is possible to incrementally raise the QE product of the network. By raising the QE product of a network, the network can provide improved QoS more efficiently. In many cases, this means that improved service can be offered at a lower cost.
Figure 2.2 Using QoS Mechanisms to Raise the QE Product of the Network
2.2.1 Quality, Efficiency, and Overhead
In designing a network, the network manager is faced with the following questions:
-
Do I need to support high-quality services through my network?
-
If so, can I afford to support these services by overprovisioning the network?
-
If not, how much QoS mechanism am I willing to deploy in my network to provide the required QoS?
If the network manager wants to support high-quality services with respect to a certain set of applications, then the quality is fixed. Because Q = C/E, the manager must either reduce E or increase C until the required quality can be supported. In this case, the network manager is faced with the following options:
-
Maintain the same QE curve, but shift the operating point of the network toward higher quality. This requires adding bandwidth with a resulting decrease in efficiency (reducing E).
-
Apply QoS mechanisms to raise the QE curve of the network so that a higher quality can be provided without adding bandwidth (increasing C).
The first approach incurs the cost of increasing bandwidth. The second approach incurs the costs associated with the deployment of the QoS mechanisms selected. A variety of QoS mechanisms can improve the QE product to varying degrees with a corresponding savings in bandwidth and the associated cost. The marginal savings in bandwidth costs must be weighed against the marginal costs of deploying and managing the mechanism to determine just which mechanisms are worth deploying. In general, the more a specific QoS mechanism raises a network’s QE product, the higher its complexity and, therefore, the higher its cost.
In the remainder of this book, various QoS mechanisms will be discussed in the context of their impact on the QE product, their complexity, and their cost.
2.2.1.1 It Is Not Always Necessary to Raise the QE Product
Often, there may be no need to raise the QE product. In certain cases, the network manager may not be interested in providing high-quality services. In other cases, bandwidth may be so inexpensive that the network manager can afford to overprovision and to operate the network inefficiently. This is often the case with LANs.
2.2.1.2 QoS Mechanisms Are Local, But Their Impact Is Global
The choice of which QoS mechanisms to apply to a network rests in the hands of the manager of each network. However, these days most networks are subnetworks of larger networks and, ultimately, of the Internet. Consequently, if a network manager decides not to support high-quality services through a subnetwork, that decision may compromise any application traffic traversing the subnetwork. QoS mechanisms may be applied on a local basis, but their impact is global.
2.2.1.3 QoS Mechanisms Do Not Create Bandwidth
As described previously, QoS mechanisms do not create bandwidth. No QoS mechanisms will make it possible for a 28.8Kbps modem link to support high-quality HDTV. In certain cases, the network manager will have no choice but to increase the bandwidth of the network to provide high-quality services. However, QoS mechanisms do increase the efficiency with which existing resources are used. Thus, in many cases, the existing bandwidth may be sufficient to offer the required service if it is used efficiently. In these cases, QoS mechanisms eliminate or at least defer the need to add bandwidth to the managed network.
2.2.1.4 Quality Is Application-Specific
When deciding whether to support high-quality services and how much overprovisioning is required, the network manager must consider the issue in the context of specific application requirements. For example, while the Microsoft LAN may be sufficiently overprovi-sioned to offer high-quality service for IP telephony, it may offer poor service quality for HDTV streams. A single physical network often can be partitioned into a number of logical networks, each offering different QE products, and each targeted at different application traffic. This concept will be discussed in the “Sharing Network Resources: Multiple Resource Pools” section of this chapter.