The Need for Wireless LAN QoS

WLANs are growing in popularity. They are mostly implemented as extensions to, but are occasionally deployed as overlays to, wired LANs, or replacements for wired LANs. The difference between wired and wireless LANs is in the physical layer and in the MAC layer. Please note that Logical Link Control (LLC) and MAC are considered upper and lower sublayers of the OSI Layer 2 Data Link Control (DLC) layer, respectively. Upper-layer protocols and applications such as IP, TCP, and FTP run identically on both wired and wireless platforms. Figure 8-1 shows two access switches (Layer 2) connected to a distribution (multilayer) switch. The access switch on the right side has wired devices plugged into its access ports. The access switch on the left side, however, is connected to a wireless LAN controller (WLC), which in turn is connected to and controls two wireless LAN APs. Each wireless client on the left side of Figure 8-1 communicates with an AP by sending and receiving frames over the radio frequency (RF) link, and it gains access to the network.

Figure 8-1 Wireless LAN Extending the Wired LAN

Wireless LAN Extending the Wired LAN


Wired Ethernet uses carrier sense multiple access with collision detection (CSMA/CD) as its MAC mechanism. Wireless LAN (802.11), on the other hand, lacks the ability to read and send data at the same time; therefore, it cannot detect collision like its wired counterpart can. Hence, WLAN uses carrier sense multiple access with collision avoidance (CSMA/CA) as the MAC mechanism. Collision avoidance is accomplished by distributed coordinated function (DCF). DCF uses RF carrier sense, inter-frame spacing (IFS), and random back-off/CWs. Please note that random back-off/CWs are sometimes casually referred to as random wait timers.

To be able to offer end-to-end QoS, the wireless portion and components of a network must comply with and satisfy the QoS needs of the applications. Following are some of the main QoS needs of applications, such as voice and video:

■ Dedicated bandwidth

■ Controlled jitter and delay (latency)

■ Managed congestion

■ Shaped traffic (rate limited)

■ Prioritized traffic (with drop preference)

WLAN QoS Description

IEEE has provided the QoS extensions to WLANs in the 802.11e specifications. The ONT courseware refers to 802.11e as a draft for standardization, but at the time of this writing, 802.11e is already approved and is considered a new standard. IEEE defines 802.11e as the first wireless standard, adding QoS features to the existing IEEE 802.11b and IEEE 802.11a (and other) wireless standards, while maintaining full backward compatibility with them. While 802.11e was in the standardization process, Wi-Fi Alliance released a specification called the Wi-Fi Multimedia (WMM) for the interim period.

WMM is a subset of 802.11e; for instance, WMM reduces the eight priority levels of 802.11e to four access categories. Note that access category has the same meaning as priority level. Using the basic CSMA/CA-based DCF, each client generates a random back-off number between 0 and a minimum contention window (CWmin) and waits until the RF channel is free for an interval called distributed coordinated function inter-frame space (DCF IFS or DIFS). From that moment on, the channel is continuously checked; if it is free, the random back-off number is decremented by 1 until it becomes 0. At that time, the client sends the frame. If the channel becomes busy, the client has to wait until the channel is free, wait for a DIFS interval, and start decrementing the random back-off interval all over again.

The CSMA/CA-based DCF gives all devices the same priority, so it is considered a best-effort mechanism. WMM, on the other hand, provides traffic prioritization (or RF prioritization) by using four access categories: Platinum (or voice), Gold (or video), Silver (best-effort), and Bronze (background), in descending priority order. The four access categories are in effect four queues, each of which gets a higher probability of transmitting than the access priority (or queue) below it. If a specific type of traffic is not assigned to an access category, it is categorized as best-effort (Silver). The eight 802.11e priority levels are mapped to four WMM access categories, as shown in Table 8-2.

Table 8-2 Mapping of80211e Priority Levels to WMM Access Categories

WMM Access Category

802.11e Priority Level

Voice (Platinum)

6 or 7

Video (Gold)

4 or 5

Best-Effort (Silver)

0 or 3

Background (Bronze)

1 or 2

802.11e (and its subset WMM) uses Enhanced Distributed Coordination Function (EDCF) by employing different CW/back-off timer values for different priorities (access categories). If a client finds the RF channel available, it waits for a DIFS period, and then it has to wait for a random back-off period based on the CWmin associated with the priority of the traffic being submitted (more accurately, the queue that the traffic is submitted from). If the traffic is high priority, its CWmin is smaller, giving it a shorter back-off timer value; if the traffic is lower priority, its CWmin is larger, giving it a longer back-off timer value. Note that with EDCF, even though high-priority traffic such as voice is statistically expected to be transmitted before lower-priority traffic, it is not guaranteed to do so at all times; therefore, technically EDCF cannot be equated to a strict priority system. With EDCF, IFS (Inter Frame Space) is referred to as AIFS (Arbitrated IFS).

NOTE In the original ONT student course material, on the page titled "WLAN QoS RF BackOff Timing," SIFS is mistakenly used instead of DIFS. Short inter-frame space (SIFS) is used only before transmitting important frames such as acknowledgements, and it has no random back-off. SIFS is not used to transmit regular data frames. Data frames, on the other hand, must wait for a DIFS and then begin the random back-off procedure.

Split MAC Architecture and Light Weight Access Point

To centralize the security, deployment, management, and control aspects of WLANs, Split MAC Architecture (a part of Cisco Unified Wireless Network Architecture) shifts some of the functions traditionally performed on the autonomous AP to a central location (device). The main functions performed on legacy autonomous APs are shown in Table 8-3 categorized under two columns: real-time 802.11/MAC functionality, and non-real-time 802.11/MAC functionality.

Table 8-3 Real-Time and Non-Real-Time 802.11 MAC Functions

802.11/MAC Real-Time Functions

802.11/MAC Non-Real-Time Functions

Beacon generation

Association/disassociation/reassociation

Probe transmission and response

802.11e/WMM resource reservation

Power management

802.1x EAP

Table 8-3 Real-Time and Non-Real-Time 802.11 MAC Functions

802.11/MAC Real-Time Functions

802.11/MAC Non-Real-Time Functions

802.11e/WMM scheduling and queuing

Key management

MAC layer data encryption/decryption

Authentication

Control frame/message processing

Fragmentation

Packet buffering

Bridging between Ethernet and WLAN

To address the centralized RF management needs of the enterprises, Cisco designed a centralized lightweight AP (LAP or LWAP) wireless architecture with Split-MAC architecture as its core. Split-MAC architecture divides the 802.11 data and management protocols and AP capabilities between a lightweight AP and a centralized WLAN controller. The real-time MAC functions, such as those listed in the left column of Table 8-3, including handshake with wireless clients, MAC layer encryption, and beacon handling, are assigned to the LWAP. The non-real-time functions such as those listed in the right column of Table 8-3, including frame translation and bridging, plus user mobility, security, QoS, and RF management, are assigned to the wireless LAN controller.

Next post:

Previous post: