Voice Scores Big in Triple Play Results
May 09, 2005
The University of New Hampshire's tests of voice, video and data in a heterogeneous, multi-vendor network revealed that VoWi-Fi may just be a technical success.
Enterprises seeking answers on how best to integrate voice and video with existing wireless data networks should be encouraged by results released last week of the first-ever wireless "triple play" tests. After testing products from 35 different Wi-Fi vendors, the University of New Hampshire's InterOperability Laboratory (UNH-IOL) announced that while some technical issues remain, there is "a positive outlook for enterprises looking to combine the power of triple play services with the versatility of a converged network."
The Wi-Fi triple play tests were a first of their kind, according to UNH-IOL. Unlike homogenous one-vendor scenarios, the UNH-IOL's goal was to create a more realistic testbed, analogous to a heterogeneous converged enterprise wireless network. (It is only fitting that the results were released last week during Interop 2005, a tradeshow dominated by interest in VoIP.)
The UNH-IOL found that Wi-Fi access points were able to process twice as many VoIP calls as previously thought. The VoIP portion of the lab's week-long tests revealed that an individual AP could handle between 14 and 16 IP-based calls before quality began to suffer. Previously, researchers thought each AP could care for just seven high-quality VoIP calls. The new findings may indicate that enterprises will require half as many APs for VoIP services as previously believed.
Although service quality was "seriously impaired," the UNH-IOL results show that a single Wi-Fi AP can actually control as many as 24 simultaneous VoIP phone calls. At the extreme, the network was loaded with 24,000 calls. The VoIP portion of the triple play test spanned 17 hours, pushing "1.7 terabytes of voice traffic over the network, both wired and wireless."
"We certainly saw results that we would not necessarily have predicted," said Gerard Goubert, manager of the UNH-IOL wireless and VOX (Voice over Anything) consortia.
Goubert hinted at some of the potential results before the testing organization released its ultimate findings. In late March, Goubert told Wi-Fi Planet that IT departments must know how each network service may impact one another. Those comments were reflected in the final report.
"There will be specific IT issues that enterprise administrators will need to be aware of as they deploy triple play on their networks," said Goubert. "As the results show, no one really knew how many VoWLAN calls each wireless access point could handle, for example, or how the data, voice and video streams would all behave on the network."
Some 13 companies were part of the multi-vendor testing regime created by the UNH-IOL team. 138 light APs were included in the triple play tests highlighting the importance of data and voice behaving well together.
"Due to the high demand of VoIP services in professional settings, it is imperative to enterprise adoption that sensitive VoIP streams and data coexist within a WLAN switch system," according to the UNH-IOL white paper.
Indeed, because of the importance placed on voice, the data portion of the UNH-IOL tests centered on features that "improve co-existence between VoIP and wireless devices."Along with Aruba, Cisco, Dell and IBM, equipment test vendors Spirent, Azimuth, ClearSight, Empirix, Ixia, and VeriWave played roles in the wireless triple play research.
"It is absolutely necessary for systems to be voice aware," according to the researchers. Such awareness includes session-aware WLAN switches which separate data and voice traffic, preventing VoIP phones from reaching the data side of a network and data streams from interfering with voice traffic.
While separating voice and data is ideal, it is often more practical to mix the two, creating a converged network. For converged networks, the UNH researchers recommend that WLAN switches be session-aware, able to distinguish between VoIP and data traffic.
But awareness of voice traffic on a wireless network is not enough. Enterprises should employ enhanced RF management. "Voice-aware functionality tests proved that, consistent with their intended purpose, enhanced RF scanning did not have any noticeable effect on voice quality," researchers wrote.
The UNH-IOL researchers urge that the lab's Quality of Service (QoS) testing methodology be adopted by vendors. "Integration of these features into an overall QoS network architecture would improve the networks' ability to prioritize traffic types from end to end," according to the final report.
While the features of VoIP gear tested varied somewhat, the biggest difference was call capacity. Each call actually consists of a caller and recipient. "Each call typically is counted as two actual calls," said researchers. "This introduces a certain ambiguity into gauging the capacity of various systems." For example, one system was able to originate 120 calls and terminate 120 calls for a total of 240 concurrent VoIP calls. Another system originated 144 calls and terminated 144 calls for a maximum of 288 IP-based conversations. Yet another was capable of 24,000 VoIP calls, 12,000 on each side, according to the white paper.
Of all three parts of the wireless triple play, "video proved most fallible." The key to video quality is combining the right transportation method (wired or wireless) with a correctly-configured client. During the video testing portion, several video streams ran concurrently—all from the same device and video server. Using efficient Ipv4/Ipv6 multicasting, clients received the video streams using the free video LAN software. On older laptops using 802.11b networks, only a single video stream was received with "less than optimal" video playback due to dropped packets and their retransmission.
New laptops with plenty of memory and fast processors, using 802.11g or 11a, "easily maintained up to four simultaneous DVD-quality video streams using either wired or wireless connections," wrote the UNH researchers.
Why conduct video tests using old hardware? "Voice and video performance on less-than-optimal hardware configurations is key when testing triple play services, as rarely would every system attempting to use the triple play services be of sufficient specifications to guarantee consistent behavior," according to the white paper.
Maybe more surprising than some of the wireless triple play test results was the fact that the UNH-IOL was able to get competing Wi-Fi vendors to cooperate.
At first, participants didn't understand the reasons for the tests, says Goubert. But once they were told that the tests would determine which products from different companies could work together, wireless vendors endorsed the idea. "This was a collaborative effort," says Goubert.
The UNH-IOL is seeking input from service providers and carriers for the next triple play test, which will include a broadband/DSL version of the test. It will take place sometime in the next six months. [Corrected on May 10, 2005.]