The promise of increased user mobility and productivity has driven wider corporate adoption of WLANs, and one of the most alluring—yet most elusive—uses of the technology is voice over wireless.
Because VOW clients will quickly expose flaws in underdeveloped wireless networks—and because of the relatively limited number of client devices available—eWEEK Labs believes VOW deployments are not quite ready for prime time. However, administrators should begin to add VOW-specific criteria to their plans for both wireless and voice networks.
One of the biggest hurdles to wide adoption of VOW is the current dearth of Wi-Fi-enabled devices.
During the next year, we expect to see a sharp increase in the number of SIP (Session Initiation Protocol)-based VOW devices available in the United States. We also expect to see broader standards support: The current batch of devices supports only the slower IEEE 802.11b wireless standard.
An uptick in the number of available devices will increase choice for users and IT administrators, but implementers must take care to establish a core requirement of telephony features that any device must support. They must also ensure that devices can provide acceptable performance for their organizations mobile users.
Voicing support
Many wireless networks were designed for data access only, so organizations looking to deploy VOW must first determine whether their WLAN (wireless LAN) will adequately support voice for a large user base.
The WLAN will need to provide sufficient RF (radio frequency) coverage to minimize dead spots and be able to accommodate heavy user densities in many locations. Scalability, security and QOS (quality of service) features will also be of paramount importance.
Unlike wireless devices designed for data, VOW phones do not require much bandwidth. The phones weve seen to date generally support the G.711 a-law and u-law and G.729 codecs, so administrators should expect only 8K- or 64K-bps (unidirectional) throughput per call.
Handoff times and roaming performance will be critical metrics by which wireless phones will be judged. Handoff times can be measured by the amount of time it takes between data transmissions through different access points—the time between the last packet sent through access point A and the first packet sent through access point B.
During this interval, a wireless client is probing for an access point with a stronger signal and then associating and authenticating to this access point.
The human ear can detect latency and delay issues that take longer than 50 to 80 milliseconds, and extended handoff times (500 ms or more) can lead to dropped calls.
Gearing up for this wireless phone test, we performed a number of practice tests on wireless clients designed for data access. In these informal tests, we saw excellent sub-30-ms handoffs from some high-end adapters and sluggish times (well over 1 second) from some consumer-grade and integrated wireless devices.
Our final tests showed similar variability in our small sample group, as well as conclusive evidence that a very common brand of wireless adapter makes for a poor roaming soft phone.
To the test
eWEEK labs tested the roaming performance of a pair of VOW phones: Zyxel Communications Corp.s P-2000W v2 and Hitachi Cable Ltd.s WirelessIP5000. We invited Zultys Technologies and SpectraLink Corp. to submit products for testing, but we were unable to acquire and install their products in time for this project.
Using testing equipment provided by Azimuth Systems Inc. , we tested each devices roaming performance in a Cisco Systems Inc. Aironet-based wireless network. For comparison, we also tested the roaming capabilities of a Dell Inc. Inspiron 600m notebook computer configured with Intel Corp.s Pro/Wireless LAN 2100A wireless adapter and the X-Lite soft phone, a free SIP-based telephony application from CounterPath Solutions Inc. (formerly Xten Networks Inc.).
We were pleasantly surprised by the outstanding roaming performance turned in by Zyxels P-2000W v2. The P-2000W v2 far surpassed the other devices in our tests, turning in an average roam time of just more than 17 ms. According to Zyxel officials, this outstanding performance is due to a proprietary implementation of several operations, including the phones association with the access point, authentication and prekeying, as well as a quick-scan mechanism that finds the best access point to connect to. (The prekeying capabilities did not affect our testing because we tested without encryption.)
You cant knock the Zyxel phones performance, but because it is so hasty in looking for stronger signals from other access points, we have some concern about the excessive amounts of client beacons that would be generated with a large number of phones in the same space.
Hitachi Cables WirelessIP5000 has a nice ergonomic design and a strong feature set, making it among the most highly recommended of the limited number of Wi-Fi phones on the market today. The WirelessIP5000 is also the first SIP phone certified with ShoreTel Inc.s ShoreTel 6 telephony platform.
During our tests, the WirelessIP5000 performed acceptably, averaging 163 ms per roam. This throughput could cause noticeable delay in a standard wireless network as a user roams, but the performance we saw was in line with Hitachi Cables expectations: The WirelessIP5000 systems engineer manual estimates 100 to 300 ms for a Layer 2 handoff, depending on the type of access points used.
We also performed a base-line test on a PC with an integrated Intel Pro/Wireless LAN 2100A wireless adapter and the popular X-Lite soft phone application. Known for being a little sticky in roaming situations (Intel adapters have a tendency to stay associated with a single access point, regardless of signal strength), the Intel wireless adapter turned in horrible performance, with average roam times of 1.7 seconds and a maximum time of 3.4 seconds. Fortunately, its the rare user who will wander office corridors talking on a laptop-based soft phone.
Network issues
The numbers we present here reflect performance in a wireless network configured without encryption. Administrators should expect increased delays with security enabled, and these numbers may vary depending on the type of security selected.
The 802.1x protocol that provides strong port-based authentication to the WPA and WPA2 WLAN security specifications may add significant delay when a client initiates a roam between access points. In a worst-case scenario, the wireless client device must re-authenticate to the RADIUS server for each handoff.
The IEEE 802.11i security standard, ratified last year, includes a few optional parameters to help speed roaming in a secured environment. Implementers should check with their wireless vendor as to whether preauthentication to neighboring access points or proactive key caching is supported.
Many wireless switch vendors also include proprietary features to help load balance the wireless network and ensure QOS for users, particularly voice clients.
For example, Meru Networks Inc.s WLAN System works to remove the burden of roaming from the client devices almost entirely. Using a single MAC (media access control) address and one radio channel for an entire wireless network, roaming decisions are instead initiated by the central switch. Client devices view the entire network as a single radio cell, which minimizes delay in handoffs because the client does not need to reassociate and reauthenticate.
Technical Analyst Andrew Garcia can be reached at andrew_garcia@ziffdavis.com.
Check out eWEEK.coms for the latest news, views and analysis on voice over IP and telephony.