QoS is the capability of a network to provide differentiated service to selected network traffic over various network technologies. Configuring QoS does not increase the bandwidth of your network. It merely gives you more control over how the bandwidth you have is allocated to different devices on the network. Make sure you understand your traffic, the protocols involved, and the sensitivity of the application to network delays.
Latency, Jitter, and Loss
The quality of a network transmission is a result of three things:
■ Latency
■ Jitter
■ Loss
Latency is how long it takes for a packet to be received by the endpoint after it is sent from the source. Latency is also referred to as delay. Asymmetrical tunneling after a Layer 3 roaming event between controllers can introduce delay. Again, symmetrical mobility tunneling is the recommended configuration.
Delay can be broken into two parts:
■ Fixed delay: The time it takes to encode and decode the packets and the time it takes for the packet to traverse the network.
■ Variable delay: Caused by network conditions. If the network is highly utilized at certain times of the day, the variable delay would be higher at those times than others.
Jitter is the value that results from the difference in end-to-end latency between packets. If a packet takes 50 ms to traverse the network and the next packet takes 100 ms, you have a jitter value of 50 ms.
Loss is simply the ratio of packets that are successfully received by the endpoint to those that were sent by the transmitter.
Correct Packet Marking
You need to be familiar with three types of packet classifications:
■ AVVID 802.1p User Priorities (UP)
■ AVVID IP Differentiated Services Code Point (DSCP)
■ IEEE 802.11e UP
Depending on the traffic flow of a packet, these classifications are used to properly classify, or mark, that traffic on the network. There are two key concepts to understanding the traffic flow and the way the packets are marked:
■ With Layer 3, both wired and wireless traffic trust Architecture for Voice, Video, and Integrated Data (AVVID) IP DSCP end to end.
■ With Layer 2, wired uses AVVID 802.1p UP markings, and wireless uses IEEE 802.11e UP.
Because wired and wireless are using different Layer 2 classifications to mark the packets, the AP must convert between AVVID 802.1p UP and IEEE 802.11e UP. Table 11-2 shows the QoS conversion table that allows the AP to convert between the different classifications.
Table 11-2 Access Point QoS Translation Values
|
QoS Baseline IEEE 802.1 UP-Based Traffic Type |
AVVID IP DSCP |
AVVID IEEE 802.1p UP |
IEEE 802.11e UP |
|
Network control |
— |
7 |
— |
|
Inter-network control (LWAPP1 control, IEEE 802.11 management) |
48 |
6 |
7 |
|
Voice |
46 (EF2) |
5 |
6 |
|
Video |
34 (AF41) |
4 |
5 |
|
Voice Control |
26 (AF31) 25(CS3) |
3 |
4 |
|
Background (gold) |
18 (AF21) |
2 |
2 |
|
Background (gold) |
20 (AF22) |
2 |
2 |
|
Background (gold) |
22 (AF23) |
2 |
2 |
|
Background (silver) |
10 (AF11) |
1 |
1 |
|
Background (silver) |
12 (AF12) |
1 |
1 |
|
Background (silver) |
14 (AF13) |
1 |
1 |
|
Best Effort |
0 (BE) |
0 |
0, 3 |
|
Background |
2 |
0 |
1 |
|
Background |
4 |
0 |
1 |
|
Background |
6 |
0 |
1 |
1LWAPP = Lightweight Access Point Protocol 2EF = Expedited Forwarding
Figure 11-2 shows an example of voice packet markings and necessary conversions.
As you can see in Figure 11-2, the Layer 3 AVVID IP DCSP packet marking for the Voice Control packet remains the same between the wired and wireless medium. The Layer 2 markings, however, are different, and the AP has to convert between the wireless 802.11e UP value of 6 and the wired AVVID 802.1p UP value of 5.
Figure 11-2 Sample QoS Packet Markings and Conversion
In an LWAPP/Control and Provisioning of Wireless Access Points (CAPWAP) deployment, the AP and the controller exchange LWAPP/CAPWAP control and LWAPP/CAPWAP data packets. For LWAPP/CAPWAP control packets, QoS is simple in that the AP always sends control packets with an AVVID IP DSCP tag of 46 in the LWAPP/CAPWAP header. The controller also sets the AVVID IP DSCP tag to 46 for control packets to the AP. If the interface is tagged—that is, the controller network port on the switch is a trunked interface—it also sets the AVVID 802.1p UP to 7.
For LWAPP/CAPWAP data packets, the AP must convert the QoS markings as traffic is sent to and from the wireless clients. If a mistake is made here, it is important to realize that the mistake only affects wireless QoS because the AVVID IP DSCP tag of the payload is never altered. Figure 11-3 outlines the QoS tagging conversions for wireless-to-wired and wired-to-wireless traffic.
Figure 11-3 Traffic Flow QoS Classification Mappings
In Figure 11-3 the following occurs:
Wireless to Wired:
Step 1. The wireless client sends a packet that contains Layer 2 IEEE 802.11e, Layer 3 AVVID IP UP, and data.
Step 2. When the packet reaches the AP, because an AP never sends tagged packets, the 802.11e UP information is stripped without changing the original AVVID IP DSCP or data.
Step 3. The AP encapsulates the packet and uses the QoS conversion table (Table 11-2) to convert the IEEE 802.11e UP value from the packet to get the AVVID IP DSCP tag to correctly mark the LWAPP/CAPWAP header.
Step 4. The packet reaches the controller and the controller de-encapsulates the packet, removing the LWAPP/CAPWAP header and bridging the "naked" packet onto the wired network.
Step 4a. If the packet is sent out of the controller on a tagged interface, the controller uses the QoS conversion table to properly map the AVVID IP DSCP marking from the LWAPP/CAPWAP header of the packet to the correct Layer 2 AVVID 802.1p UP tag.
Notice that the AVVID IP DSCP from the client never changed throughout the whole process.
Wired to Wireless:
Step 1. A packet destined for the client enters the controller.
Step 2. The controller copies the AVVID IP DSCP marking from the incoming packet to use on the LWAPP/CAPWAP packet header.
Step 2a. If the packet arrives at the controller on a tagged interface, the controller also copies the AVVID 802.1p UP tag.
Step 3. The AP strips the LWAPP/CAPWAP header and uses the AVVID IP DSCP tag to correctly map the IEEE 802.11e tag before sending the packet to the client.
Again, the AVVID IP DSCP marking from the original packet was never changed throughout the whole process.
Note With controller code Release 5.1, Cisco introduced Traffic Classification (TCLAS) to ensure that voice streams are properly classified. Because LWAPP/CAPWAP data packets always use the same ports, 16666 and 5247 respectively, and the AP uses the outside QoS marking to determine which queue the packets should be placed in, using port-based QoS policies is inadequate. With TCLAS, even if the LWAPP/CAPWAP AVVID IP DSCP markings are incorrect, the traffic is tagged correctly.
Upstream and Downstream QoS
When discussing QoS, it is important to understand the terminology and direction of the traffic flow to and from the AP and the controller. You have both upstream and downstream QoS:
■ Radio downstream: Traffic leaving the AP and traveling to the WLAN clients.
■ Radio upstream: Traffic leaving the WLAN clients and traveling to the AP. Enhanced Distributed Channel Access (EDCA) rules provide upstream QoS settings for WLAN clients.
■ Network downstream: Traffic leaving the controller traveling to the AP. QoS can be applied at this point to prioritize and rate-limit LWAPP/CAPWAP traffic to the AP. Configuration of Ethernet downstream QoS is not covered in this topic.
■ Network upstream: Traffic leaving the AP, traveling to the controller. The AP marks the traffic according to the value set by the wireless client using the IEEE 802.11e UP to AVVID IP DSCP conversion table.
Figure 11-4 outlines the QoS traffic flow concepts.
Figure 11-4 QoS Traffic Flow
In Figure 11-4, the diagram is broken into two separate ranges: A and B. Range A specifies the network between the AP and the controller (the LWAPP/CAPWAP tunnel). Range B refers to the wired network after the controller. Keep these range designations in mind as you read through the rest of the topic because this concept will come up again.
Wi-Fi Multimedia
WMM is a certification that applies to both clients and APs. The features are taken from the 802.11e draft.
Using the eight IEEE-developed 802.1p QoS classifications, WMM maps the classifications into four access categories. The four access categories are mapped to the WMM queues required by a WMM certified device. Table 11-3 outlines the 802.1p to WMM mappings.
Each of the four WMM queues competes for the wireless bandwidth available on the channel. WMM uses Enhanced Distributed Coordination Function (EDCF) for handling the queue traffic. If more than one frame from different access categories collides internally, the frame with the higher priority is sent. The lower-priority frame adjusts its backoff parameters as though it had collided with a frame external to the queuing mechanism.
WMM prioritization helps minimize delays in wireless networks for time-sensitive applications such as voice and video. WMM is the default EDCA parameter set on the controller.
Table 11-3 802.1p WMM Mappings _
|
Priority 802.1P Priority |
Access Category |
WMM Queue |
|
Lowest 1 |
AC BK |
Background |
|
2 |
||
|
0 |
AC BE |
Best Effort |
|
3 |
||
|
4 |
AC_VI |
Video |
|
5 |
||
|
6 |
AC_VO |
Voice |
|
Highest 7 |
TSPEC
TSPEC allows an 802.11 wireless client to signal its traffic requirements to the AP. The client includes the TSPEC in the add traffic stream (ADDTS). The TSPEC from the client includes requirements for data rate, packet size, number of streams, and more. The 802.11e standard specifies TSPEC to provide the management link between higher QoS protocols and the channel access functions. Channel access functions are defined by the EDCA mechanism. TSPEC allows the AP to control access bandwidth to avoid traffic congestion. To enable TSPEC on the controller, you enable Call Admission Control (CAC).
ADDTS
The ADDTS function is how a WLAN client (STA) performs an admission request to an AP. Signaling its TSPEC request to the AP, an admission request is in one of two forms:
■ ADDTS action frame: This happens when a client originates a phone call associated to the AP. The ADDTS contains TSPEC and might contain a traffic stream rate set (TSRS) information element (IE) (Cisco Compatible Extensions v4 clients). A Cisco wireless phone actually performs two ADDTS because the codec used in the call is not known before the RTP stream is established.
■ Association and reassociation message: The association message might contain one or more TSPECs and one TSRS IE if the STA wants to establish the traffic stream as part of the association. The reassociation message might contain one or more TSPECs and one TSRS IE if an STA roams to another AP.
The ADDTS contains the TSPEC element that describes the traffic request. Apart from key data describing the traffic requirements, such as data rates and frame sizes, the TSPEC element tells the AP the minimum physical rate (PHY) that the client device will use. This allows the calculation of how much time that station can consume in sending and receiving in this TSPEC, thereby allowing the AP to calculate whether it has the resources to meet the TSPEC. TSPEC admission control is used by the WLAN client (target clients are Voice over IP [VoIP] handsets) when a call is initiated and during a roam request. During a roam, the TSPEC request is appended to the reassociation request.
When the traffic stream finishes, the STA must send a Delete Traffic Stream (DELTS) to release the CAC resources used for that stream.
QoS and H-REAP
When using voice on a Hybrid Remote Edge Access Point (H-REAP) AP, the traffic flow may change depending on the configuration. For WLANs that have data traffic forwarded to the WLC (that is, for centrally switched WLANs with WMM traffic), the behavior is the same as local mode APs. For WLANs that are locally switched, however, a different approach is taken. The AP marks the dot1p value in the dot1q VLAN tag for upstream traffic. This occurs only on tagged VLANs—that is, not native VLANs.
For downstream traffic, the H-REAP AP uses the incoming dot1q tag from the Ethernet side to queue and mark the WMM values on the radio of the locally switched VLAN.
The WLAN QoS profile is applied both for upstream and downstream packets. For downstream, if an IEEE 802.1P value that is higher than the default WLAN value is received, the default WLAN value is used. For upstream, if the client sends a WMM value that is higher than the default WLAN value, the default WLAN value is used. For non-WMM traffic, there is no class of service (CoS) marking on the client frames from the AP.
