- The Quality/Efficiency Product: The Reason to QoS-Enable a Network
- 2 Raising the QE Product of a Network
- 3 The Value of Different QoS Mechanisms in Raising the QE Product of a Network
- 4 Illustrative Examples
- 5 Sharing Network Resources: Multiple Resource Pools
- 6 Summary
2.5 Sharing Network Resources: Multiple Resource Pools
The general QoS-enabled network is required to simultaneously support applications with differing QoS requirements. Thus, in any part of the network, it must be possible to provide both low- and high-quality services. To this end, any physical subnetwork can be partitioned into a number of logical networks. The physical resources are allocated among the logical networks. Each logical network may be operated at a different point on the network’s QE curve or even on a different QE curve.
2.5.1 Isolation Between Traffic Types Requiring Different Quality Service
High-quality services typically are made practical via the use of signaling and explicit admission control. Low-quality services may be offered by push provisioning, with no explicit admission control. To support both service types simultaneously in a single physical network, policing is required. Policing refers to the capability to prevent traffic from seizing resources to which it is not entitled.
Traffic admitted through the process of signaling and explicit admission control allows itself to be more readily policed than that which is not. The admission control process informs the network of the routes that will be used by admitted traffic. Resource requests for traditional IntServ services also inform the network of the specific quantity of resources that will be used by admitted traffic along the indicated routes. Thus, signaling requests offer the network policing parameters for the signaled traffic. The network then can ensure that the signaled traffic does not claim resources along routes other than those on which it is admitted and that the signaled traffic does not claim excess quantities of resources. (Note that policing may be applied on a per-conversation basis or on an aggregate basis).
By contrast, traffic that is not allotted resources as a result of signaling and explicit admission control does not offer policing parameters to the network. The network manager allots resources to this traffic by pushing classifiers to network devices, which qualify the traffic to receive certain resources. However, because the traffic offers no hint as to where it will appear in the network and at what volumes it will appear, the network manager is hard pressed to select appropriate policing parameters. Without policing, any traffic that appears at a network device and matches the preconfigured classifiers is capable of seizing resources.
To maintain the integrity of high-quality services, the network manager must prevent this rogue traffic from seizing resources that are required to support the high-quality services. Because this traffic offers no policing parameters, the network manager is left with no choice but to subjugate it to the traffic that is being offered high-quality services. This means that at network devices, traffic associated with high-quality services is policed and given prioritized access to resources. Traffic associated with lower-quality services is not policed but is given lower priority in its access to resources. In effect, traffic is divided into two pools.
NOTE
Because traffic associated with lower-quality services is subjugated to traffic associated with higher-quality services, it is not necessarily true that applications requiring lower-quality services are treated with less importance than applications requiring higher-quality services. For example, many network managers would balk at the notion that SAP/R3 traffic is treated less importantly than video-conferencing traffic. In general, this would be unacceptable.
The overall treatment of the application’s traffic is determined not only by the priority granted to the application’s traffic in any particular queue, but also by the size of the resource pool available to the application’s traffic. Typically, only a small fraction of the resources available at network devices is available for reservation through explicit admission control, with the majority remaining available for traffic associated with lower-quality services.
Thus, although traffic associated with lower-quality services may briefly yield to latency-sensitive traffic associated with higher-quality services, the average amount of resources available to the lower-quality traffic is likely to be higher than that which is available to higher-quality traffic.
2.5.2 The Four Logical Networks
It is useful to recognize four logical networks within the general physical network. Each of these controls a certain (though not necessarily constant size) resource pool. Each may be operated on a different QE curve, and each offers a different general QoS to accommodate a different type of traffic. In general, traffic requiring higher qualities of service is policed so that it does not starve traffic requiring lower-quality services. The four logical networks can be described based on the type of traffic they serve, as follows.
-
Quantifiable traffic requiring high-quality guarantees—This type of traffic requires a specifically quantifiable amount of resources along specific routes. These resources typically are allocated as a result of RSVP signaling, which quantifies the amount of resources required by the traffic flow in each part of the network. The highest-priority queues in network devices are reserved for this traffic. This traffic is subjected to strict admission control and policing. Examples of this type of traffic include IP telephony traffic and other multimedia traffic.
-
Nonquantifiable persistent traffic requiring high-quality guarantees—This type of traffic requires resources that cannot be specifically quantified. However, it tends to be persistent in the sense that it consumes resources along a known route for some reasonable duration. Resources are allocated to this class of traffic as a result of RSVP signaling, which does not specifically quantify the resources required by the traffic flow. This signaling informs the network of the application sourcing the traffic, as well as the route taken through the network. This information facilitates prediction of traffic patterns, enabling reasonable quality guarantees. However, because resource requirements are not strictly quantified, resource consumption cannot be strictly policed, and this traffic is assigned to queues that are of lower priority than those available for quantifiable traffic. Examples of this type of traffic include client-server, session-oriented, mission-critical applications such as SAP and PeopleSoft.
-
Nonquantifiable, nonpersistent traffic requiring low- or medium-quality guarantees—This type of traffic is relatively unpredictable because its resource requirements cannot be quantified and because its route through the network is fleeting and subject to frequent changes. The overhead of signaling cannot be justified for this type of traffic because it can provide little information to assist the network manager in managing the resources allocated to it. Because the impact of this traffic is so unpredictable, this traffic is forced to use queues that are of lower priority than those used by signaled traffic. As a result, only low-quality guarantees can be offered to such traffic. An example of this type of traffic is Web-surfing traffic.
-
Best-effort traffic—This is all remaining traffic, which is not quantifiable and not persistent, and which does not need any QoS guarantees. The network manager must assure that resources are available in the network for such traffic, but no specific QoS must be provided for it. This traffic uses default FIFO queues and receives resources that are left over after the requirements of higher-priority traffic have been satisfied.
NOTE
Although it is implied that the resource pools are isolated from each other using strict priority queuing (see Chapter 3, “Queuing Mechanisms”), this is not necessarily the case. Other queuing schemes may be used as well, such as allocating relative shares of a link to different subsets of traffic, with no strict priority relationship between the queues.