Adding 'Quality' to Wireless LANs

Wireless VoIP gives workers who aren't regularly at their desks more flexibility, and it can reduce operational costs associated with the use of cell phones and private radio walkie-talkie systems in the enterprise. Warehouse, distribution, retail and health-care operations have used wireless VoIP for a number of years, and it will become increasingly popular over the next several years as both VoIP and WLAN technologies mature and their prices decrease. In fact, the percentage of large enterprises deploying wireless VoIP is predicted to increase from the single digits today to 33 percent in 2006, according to Infonetics Research.

Wireless QoS can augment other applications as well, such as guest access, where you offer your clients, customers or other visitors wireless access over your WLAN while they're on site. QoS here gives your internal users higher priority than visitors, and likewise, some internal users take precedence over others.

Then there's the consumer market, where Wi-Fi has been a huge hit as a home networking technology for sharing printers, files and broadband Internet connections. QoS lets home users prioritize how bandwidth gets split among these kinds of operations. WLAN QoS also is necessary for multimedia home entertainment, where moving digitized multimedia content across home systems is gaining steam.

But QoS is more complex in the enterprise because applications are more varied and the physical scale of WLANs is greater. With QoS standards slow to emerge, some organizations have taken pragmatic approaches to ensuring their apps get the bandwidth they need. Some hospitals, for example, deploy multiband 802.11 a/b/g network infrastructures: They dedicate the 5-GHz 802.11a system to data applications and the 2.4-GHz 802.11b/g system to voice.Consider the Options

WLAN infrastructure vendors, including Airespace, Cisco Systems and Colubris Networks, offer proprietary wireless QoS. They link traffic prioritization to users and groups by assigning priority to either the user's authenticated identity or the 802.11 ESSID (Extended Service Set Identifier).

We recently tested technology from Colubris in Network Computing's Syracuse University Real-World Labs® that provides QoS by ESSID, using what Colubris calls "service-aware" Wi-Fi QoS. We connected three Dell Latitude notebooks with embedded 802.11b NICs to our lab Ethernet LAN over a Colubris CN320 access point. We used NetIQ Chariot to transfer files to each of the clients. Not surprisingly, each client shared the available bandwidth and achieved comparable throughput of about 1.6 Mbps. We then reconfigured the AP with multiple 802.11 ESSIDs and gave each a different priority assignment, resulting in higher performance for prioritized users (the chart on page 83 shows how each client performed).

Although this approach doesn't address all QoS problems, it's easy to implement. It's simple to assign ESSIDs and associated security policies to phone devices so that they get priority over conventional data traffic. The catch is that though AP-centric prioritization schemes let you prioritize traffic from the AP to the client and from the AP to the wired segment, they don't do QoS over the airwaves. So if the wireless network experiences severe congestion, VoIP clients may not gain predictable access to the network for upstream traffic. That's because all WLAN clients adhere to the same rules for gaining access to the medium--those of CSMA/CA (Carrier Sense Multiple Access with Collision Avoidance)--which are designed for fairness rather than prioritized access.SpectraLink Corp.'s proprietary SVP (SpectraLink Voice Priority) technology gets around this problem by making media access less fair. VoIP phones supporting SVP can jump to the head of the line. Most WLAN infrastructure vendors have implemented SVP in their devices.

SVP provides wireless QoS by specifying a back-off value of zero for voice packets. That gives voice packets priority over normal data packets. The downside is that if multiple SVP VoIP phones attempt to transmit at the same time, you'll get collisions. If too many phones exist in a given cell, the phones can monopolize available bandwidth and thus starve out conventional data users. To get around this, limit the number of handsets per cell to a dozen or less. For the nuts and bolts of how media access is controlled on 802.11 networks, see "How 802.11 Media Access Works," below.

On the wired side, SVP also adds priority queuing to the AP--using a unique protocol type in the IP header--to give voice packets priority.

SVP is the de facto QoS standard for WLANs, but SpectraLink is well-aware that vendor-neutral standards are key to widespread adoption. So the company is backing efforts by the IEEE to enhance the 802.11 MAC with QoS support. The IEEE's 802.11e task group, which has been working on this for about five years, is expected to release a final standard early next year.

The snail's pace of the IEEE QoS effort spurred the Wi-Fi Alliance to develop a prestandard subset of QoS functions for the short term that will work with the eventual 802.11e standard. This strategy is much like the Wi-Fi Alliance's effort with the WPA (Wi-Fi Protected Access) security scheme, a subset of the capabilities that eventually emerged in the 802.11i security standard.The Wi-Fi Alliance recently released its spec, WMM (Wi-Fi Multimedia). WMM is modeled after the IETF's DiffServ architecture, which provides four data-access categories that can be assigned different priorities.

In addition to providing priority queuing on APs, WMM is a more granular version of SVP's airwave QoS and mapped to DiffServ's four access categories. So a WMM-enabled wireless VoIP handset, for example, can be configured to use voice priority, and a WMM-enabled TV set to support video priority while deferring access to voice traffic as needed. In a PC with both a soft-phone and data applications, for instance, individual packets would get tagged to ensure priority for the voice traffic (see "Priority One: Wi-Fi Multimedia Access Categories," page 84).

Most wireless vendors will begin adding WMM capabilities to their products over the next six months. WMM certification began in September, but it's too soon to tell whether many vendors will spend the time and expense of gaining formal WMM certification. At press time, products from nine vendors (including reference-design products from all the major chip-makers) had received Wi-Fi WMM certification. A list of WMM-certified products is available at at Wifialliance.org. Get more info on WMM, including a white paper there as well.

Still To ComeIn addition to prioritized access, 802.11e will provide scheduled access, where applications can reserve bandwidth resources. This HCCA (Hybrid Coordination Function Controlled Channel Access) is a centralized scheduling mechanism that provides more deterministic network access. A VoIP app, for instance, would send a resource-reservation request to the AP. The AP would then provide an appropriate assignment based on predefined metrics, including data rate, PHY rate, packet size, service interval and burst size.

WMM and 802.11e are major developments for WLANs, but QoS alone isn't enough. No matter how sophisticated your QoS parameters, if the WLAN is poorly designed and lacks sufficient capacity, your apps won't perform and users will notice. Most enterprises need a multiband 802.11 a/b/g wireless infrastructure as well with densely positioned APs. It's this combination of smart design and QoS that will make wireless the right infrastructure for voice and other applications.

Dave Molta is a Network Computing senior technology editor. He is also assistant dean for technology at the School of Information Studies and director of the Center for Emerging Network Technologies at Syracuse University. Write to him at [email protected].

802.11 MAC relies on CSMA/CA (Carrier Sense Multiple Access with Collision Avoidance), a slightly more sophisticated variation of the Ethernet CSMA/CD (Carrier Sense Multiple Access with Collision Detection) protocol. When the pipe on an 802.11 network becomes available, a node accesses it by observing two timing parameters: the interframe space, a predefined dead time between packets; and the contention window (sometimes called the random back-off wait), the period when nodes contend for access to the medium.

Because even legacy 802.11 networks need to prioritize acknowledgment and management frames over normal data frames, the 802.11 spec defines multiple interframe spaces used by different types of packets. The shorter the interframe space a node needs to observe, the higher its priority. So because a frame acknowledgment's interframe space is shorter, it always gets higher priority access to the medium.A node can compete with other devices for access to the WLAN when its interframe space timer expires. Without prioritization, each device calculates a random back-off period before attempting to transmit. To ensure fairness, the back-off value also factors in prior unsuccessful contention attempts so a node that "lost" out to another node with a lower random back-off is more likely to "win" next time.

The original 802.11 MAC included a PCF (Point Coordination Function) that provided a shorter interframe spacing, and thus the capability to support airwave QoS. But it was inefficient and unpredictable, and therefore relegated to standards footnote status.

Meru Networks, an emerging WLAN infrastructure provider, offers a unique twist to QoS. Although Meru plans to support WMM and forthcoming 802.11e standards, it believes many enterprises require more sophisticated QoS.

Meru's Air Traffic Control technology coordinates communication across wireless cells. Each access point communicates with the centralized controller, maintaining a list of clients, their traffic flows and neighboring APs. The software dynamically load balances clients across APs and coordinates traffic transmission around all other Meru APs while limiting interference.

It also provides global resource management. All traffic flows get analyzed in terms of available resources and those in use. Air Traffic Control then specifies which flows receive priority: If a VoIP WLAN phone sends voice traffic upstream every 30 milliseconds, the software can set aside a time slot for that traffic so it's received reliably.How does it provide predictability? Its scheduling approach goes beyond deciding in which queue a packet should sit until it gets transmitted. Meru instead uses a TDM-like scheduling algorithm on a CSMA/CA network, with a custom media-access controller on its APs. This controls the timing for when each frame gets put onto the wireless network. The technology also uses a combination of physical- and virtual-carrier sense to control clients so that even the upstream traffic can be processed predictably. To learn more about this approach, check out Meru's white paper. --Frank Bulk