How QoS Works

This section describes how these components work together in a common QoS scenario. Figure 9.2 illustrates a common QoS deployment.

Enlarge figure

Figure 9.2 How QoS Works

1. A client on Network A requests QoS. The application used to transmit data is QoS-enabled. The application requests QoS from the RSVP SP.
2. The RSVP SP requests the RSVP service to signal the necessary bandwidth requirements, and notifies traffic control that QoS has been requested for this flow. Traffic is currently sent at a best-effort delivery level.
3. An RSVP message is sent to the QoS ACS server, requesting a reservation. Note that it is RSVP messages that are passed to the QoS ACS, not the data packets which are ultimately transmitted from sender to receiver.
4. The QoS ACS server verifies that enough network resources are available to meet the QoS level requested, and that the user has the policy rights to request that amount of bandwidth. The Local Policy Module uses the Kerberos ticket in the RSVP request to authenticate the user identity and look up the user policy in Active Directory. Note that the QoS ACS can verify resources for the sender, receiver, or both.
5. After verification is complete, the QoS ACS server approves the request and logically allocates bandwidth. The QoS ACS server forwards the request toward the receiver (client) on Network B.
6. When the RSVP request passes the edge router on Network A, the router keeps track of the resources (bandwidth) that are requested. The bandwidth is not yet physically allocated (RSVP is a receiver-initiated protocol and bandwidth can only be reserved by the receiver). The same process is repeated on the edge router for Network B.
7. The request is passed through each network device in the data path before it arrives at the receiver. The receiving client indicates it wants to receive the data and returns an RSVP message requesting a reservation.
8. When the receiver's request for bandwidth passes through the edge router on Network B, it already has cached the information about the requested bandwidth (from the sender's request). The router matches the receiver request with the sender's request, and installs the reservation by physically granting the bandwidth. The same process is repeated on the edge router for Network A.
9. The reservation is sent back to the sender. The layer 3 network devices (the edge routers) are capable of approving and allocating the physical bandwidth. The reservation simply passes through the layer 2 switch.
10. During this process, the traffic is sent by traffic control on the sender as best-effort. Upon receiving the reservation message, the traffic control on the sending host begins the process of classifying, marking, and scheduling the packets to accommodate the QoS level requested. The QoS Packet Scheduler performs the priority marking for RSVP, 802.1p for prioritization on layer 2 devices (illustrated here as the switch), and for Differentiated Class of Service for layer 3 devices (illustrated here as edge routers).
11. The QoS Packet Scheduler begins sending the prioritized traffic. The data is handled as priority by all devices along the data path, providing greater speed of throughput and a more successful transmission to the client on Network B.

Note that this example is a general description. Variations are possible depending on network topology as well as the presence of different network devices.

© 1985-2000 Microsoft Corporation. All rights reserved.