Quality of Service |
Traffic patterns fall into two primary groups:
Elastic traffic adapts easily to change. When little bandwidth is available, elastic traffic delivery is slow. Delivery is faster when bandwidth is abundant. The data sender is automatically tuned to the rate of the network. Elastic traffic is usually generated by transaction-oriented applications, such as bulk data transfers.
Real-time traffic is generated primarily by real-time applications that require dedicated bandwidth, such as video conferencing. Real-time traffic is limited in its ability to adapt to changing network conditions, and delays can significantly reduce intelligibility and usefulness.
Traffic control supports four service levels to meet the needs of the two primary traffic pattern groups:
Best-effort is the standard service level of many IP-based networks. It is a connectionless model of delivery that is suitable for elastic traffic. Packets are sent with no guarantees for low delay or adequate bandwidth.
The next two levels are suitable for real-time applications, giving them preferential service:
Controlled load approximates the behavior of best-effort service in unloaded (not heavily loaded or congested) conditions. A flow receiving controlled load service at a network device can experience little or no delay or congestion loss. Any unreserved bandwidth or reserved bandwidth not currently in use remains available for other traffic. See the Internet Engineering Task Force (IETF)–defined Request for Comment (RFC) 2211 for detailed information about this service level.
Guaranteed service guarantees the maximum limit on delay. This is most useful if every host on the data path provides this level of service, including routers or switches that are compliant with QoS and RSVP. However, the impact of guaranteed traffic on the network is heavy, so it is not desirable for applications that generate elastic or best-effort traffic. Any unreserved or reserved bandwidth not currently in use remains available for other traffic. See the IETF-defined RFC 2212 for more information about this service level.
Qualitative service is designed for applications that require prioritized traffic handling but cannot quantify their QoS requirements in terms of a concrete flow specification. These applications typically send out traffic that is intermittent or burst-like in nature. In the case of qualitative service, it is the network that determines the treatment of qualitative flows. Mission-critical ERP applications generate this kind of traffic. For more information about this service level, see the IETF link on the Web Resources page at http://windows.microsoft.com/windows2000/reskit/webresources. See the IETF-defined Internet Draft titled "Specification of the Qualitative Service Type" for detailed information about this service level.
The QoS Admission Control policy determines which service level a user receives.