We tested five VoWLAN (voice over wireless LAN) offerings--from Cisco Systems, SpectraLink Corp., Symbol Technologies, TeleSym and Vocera Communications--and were impressed with their capabilities. These systems provide mainstream telephony services over an existing WLAN infrastructure, including access through gateways to PBXs and the PSTN. Some of the products offer innovative capabilities, including multiparty intercom features and voice recognition. In our tests we found voice quality very good under lightly loaded network conditions, but network congestion can lead to unpredictable performance with some products, even if the platforms include basic quality and class-of-service guarantees.
For the immediate future, these products have the greatest appeal to vertical markets, such as health care, education, manufacturing, distribution and retail, but broader enterprise deployments are inevitable, particularly if next-generation WLAN infrastructure makes its way into a larger percentage of enterprises in 2004. System costs are still high, but we expect to see prices drop dramatically in the next two years. And we hope to see vendors make the move from proprietary protocols to standards like SIP (Session Initiation Protocol), a change that will enhance interoperability and push down costs.
Standard methodologies for testing VoWLAN systems have not yet been developed, and many of the tools that can be used for wired VoIP systems cannot be used with wireless because they rely on wired Ethernet connectivity. Our primary goal in testing VoWLAN systems was to assess call quality under typical office conditions and under an environment with considerable background noise. We evaluated performance under both normal network conditions and congested conditions.
Our wireless test network used Cisco 1200 APs (access points), except in the case of SpectraLink and Cisco, which recommended the use of 350 APs. To measure voice quality, one wireless VoIP phone was placed near a speaker, where we played a professionally recorded audio speech. We then assessed sound quality using another phone from a different room, averaging the results of two independent testers listening to a minimum of one minute of source material. We graded each product on a 10-point scale. During these tests, we used the internal microphone/speaker as well as a headset for listening.
To simulate less-than-ideal network conditions, each device (source and destination) was assigned a unique SSID and associated with a different AP. The APs were connected to a Shunra Storm network simulator, where we introduced a variety of network problems. The first of these conditions was an aggregate round-trip packet loss of 10 percent, 20 percent, 35 percent and 60 percent. We also evaluated the effect of out-of-order packets, creating a model where packets had a 50 percent chance on either network port of being out of order. Finally, we simulated connection drops, where the device is still associated but wired network connection is not available.
In a slightly more real-world test, we set up 13 WLAN clients (a mix of Toshiba and Dell laptops using a combination of Cisco 350, Orinoco and onboard wireless NICs) performing NetIQ Chariot long-file transfers to simulate a condition of heavy WLAN traffic load.
To realize the benefits of cloud-native software, an application must actually be cloud-native—designed to run in the cloud, interacting with disaggregated cloud services, fully manageable as code—to deliver the expected benefits.
By combining resources and expertise, the AI-Enabled ICT Workforce Consortium will offer a blueprint for how industries can adapt to an AI-dominated future.
IT leaders should heed the guidance of cybersecurity insurance providers who think businesses should prioritize security education, incident preparedness, regular internal audits, and ongoing vulnerability scanning and patching.
