Certification Shelved for Voice QoS over Wireless

Some age-old networking arguments never die--they just get rolled into new technologies. Just ask the wireless vendors working on the 802.11e standard for wireless QoS. A battle between random versus predictable access has played out, sounding remarkably similar to the battles between Ethernet and Token Ring.

EDCA (Enhanced Distributed Channel Access), the more Ethernet-like contender, has largely trounced the HCCA (HCF Controlled Channel Access), the predictable, more Token-Ring-like alternative. HCCA was more suitable for high-density, Vo-Fi deployment than EDCA and its loss could set off alarms among IT pros. The Wi-Fi Alliance claims it's addressing the problem by enhancing the EDCA portion of 802.11e to support a high number of voice calls; it expects to complete this work in early 2008. But despite the Wi-Fi Alliance's ongoing efforts surrounding EDCA, that work won't entirely replace HCCA.

Directing Wireless TrafficThe Wi-Fi Alliance's WMM (Wireless Multimedia) program certifies EDCA-compliant products; the WMM-SA (Wireless Multimedia-Scheduled Access) program was to have certified HCCA-compliant products. Managing a large number of voice stations per access point was to be the primary benefit of WMM-SA.

EDCA and HCCA are based on the initial access methods stated in 802.11, which specifies two methods of performing collision avoidance. The first is based on DCF (Distributed Coordination Function); the second, on PCF (Point Coordination Function). HCF (Hybrid Coordination Function) incorporates components from each.

DCF uses a distributed, "listen-before-you-speak" method and back-off timers to give wireless stations their fair share of the wireless medium, while the optional PCF leverages APs for coordination. With PCF, stations ask permission from the AP before communicating. DCF became a commercial success; PCF never took off.

EDCA's basic premise is to give priority to some data types over others through four priority queues. With EDCA, time-sensitive applications have a better chance of communicating, but EDCA provides no assurances of delivery. Although EDCA uses HCF, it follows the DCF mode of operation in collision avoidance.

HCCA was intended to provide assured delivery using "scheduled access"--a way for a wireless AP to grant time slots to clients for contention-free periods. The AP would receive reservation requests from stations, check resource availability, and grant or refuse the access. This is "admission control" and is often referred to as "parameterized QoS," because it receives the station requests in the form of parameters--what priority is required, amount of time needed, and so on--then controls transmission time and duration. The AP coordinates all access and contention-free periods. Clients must be aware of HCCA for the system to work. HCCA gives time-sensitive apps, such as voice, guaranteed access to the wireless medium.

HCCA isn't well-suited for unlicensed spectrum. Technology based on licensed spectrum has an advantage: In a managed RF environment, it's possible to guarantee access without concern for third-party interference. With unlicensed spectrum, tight coordination within your domain is impossible because nothing prevents another operator or technology operating in the same frequency from impacting the planned coordination. In many ways, HCCA provides a false sense of reliability. HCCA isn't suitable unless you are in a campus environment with a strict airspace policy or in a warehouse isolated from foreign interference. Interference in the unlicensed spectrum is the biggest obstacle to "guaranteed" QoS over wireless: The 802.11e standard states that "in unlicensed spectrum, true guarantees are often not possible. However, gradations of service are always possible, and in sufficiently controlled environments, QoS guarantees can truly be made."

The other hurdle with HCCA is that its implementation requires more memory and processing power on the AP, which increases costs; and, it's difficult to retrofit existing equipment. There must be a significant market driver for companies to implement HCCA; it's typically not a simple upgrade except on the newest hardware with sufficient memory and processing power. Additional intelligence is also required in roaming standards to handle HCCA, since it's controlled by the AP; unfortunately, that newer fast-roaming standard (802.11r) isn't ratified yet.

Stalled Certification

The Wi-Fi Alliance stated in 2004 and 2005 that it would begin to certify products based on HCCA in 2006, but that hasn't happened. The group quietly dropped WMM-SA in 2006. That means that the announced testing of WMM-SA has effectively been shelved.All is not lost for supporting high-density voice on wireless LANs. The Wi-Fi Alliance is implementing the admission-control portion of the 802.11e standard, which will make easier to support a high number of voice calls. In 802.11e,

admission control ensures that the AP can allocate the resources to handle more latency-sensitive applications. Some of the large vendors, such as Cisco Systems, already support some type of admission control, though these implementations are still proprietary.

Besides admission control, vendors have tried to proactively manage the impact of interference using automated tools to manage the radio frequency environment. Unfortunately, all standards-based Wi-Fi solutions in the foreseeable future will remain "best effort," and won't replace what HCCA was meant to deliver.

Jonn Martell is president of Martell Consulting. He has 20 years of experience in networking and it. He also teaches wireless, networking and security courses at the University of British Columbia. Write to him at [email protected].