TWL Office Telephone Systems - voice and data services - VOIP - MPLS ∓ Data Networks - maintenance Packages - Wales, Cardiff, Newport

Whereas many tests try to simulate a wireless network using wires, the Novarum benchmarks used real over-the-air testing in a building with voice and data clients. The 4,000-square foot test environment was designed to include multiple obstacles to RF propagation, necessary for accurate analysis of the multipath effects on which 802.11n relies.

Meru's Wireless LAN Virtualization system was designed from the ground up to scale to the very high data rates needed by dense users, so it isn't surprising that Meru leads in overall system capacity. But Meru also trounced competitors in every other metric that Novarum measured: perceived voice quality, energy efficiency and airtime fairness in both mixed-mode and 802.11n-only networks.

TCP Throughput
The simplest performance metric is total network throughput at the TCP layer. Using only one 5 GHz. radio, a single Meru access point with ten data clients consistently offered total throughput in excess of 170 Mbps, with peak speeds reaching higher than 180 Mbps - near the theoretical maximum of 802.11n draft 2.0. Meru's average throughput was 29% higher than Aruba's and 68% higher than Cisco's.

802.11n Fairness
A high average throughput isn't necessarily useful if the bandwidth is unevenly distributed among clients. The lowest throughput of the ten 802.11n clients is perhaps an even more important metric, as this measures predictability and consistency - something vital if 802.11 is to replace Ethernet.

Of the clients on the Meru network, the slowest ran at 15 Mbps, with all ten showing a standard deviation from the mean of only 9%. This demonstrates that Meru can both reach a high data rate and distribute capacity fairy across clients. The Cisco network had its slowest client running at 9 Mbps and a standard deviation of 14%, representing fairness but low capacity. Aruba's network was much less consistent than either Cisco or Meru, with its slowest client at 7 Mbps and a standard deviation of 32%.

Mixed Mode Fairness
To test performance in a mixed network of 802.11n and legacy clients, two of the 802.11n laptops were replaced by 802.11g laptops on each vendor's wireless LAN. The important metric here is airtime fairness, not throughput fairness, as 802.11a clients are naturally slower. If properly implemented, airtime fairness lets each client negotiate its own data rate with the access point and transmit at its own speed. The network should appear to be 802.11n to 802.11n clients, 802.11a/g to 802.11a/g clients.

Meru again led here, both in overall system capacity and fairness. The total throughput of the mixed mode Meru network was 157.7 Mbps, compared to 126.2 Mbps for Aruba and 92.1 Mbps for Aruba. Airtime fairness results were similar to the 802.11n-only test, with a standard deviation of 10% for Meru, 9% for Cisco and 43% for Aruba. As with 802.11n-only network, Cisco distributes its capacity fairly but offers a low overall throughput, while Aruba has a higher average data rate but much less consistency. Only Meru achieves high data rates and predictable performance.

The tests also demonstrated that the Aruba and Cisco access points had a clear bias towards 802.11n over legacy clients, giving 802.11n more of the network's airtime and thus an even bigger performance advantage over legacy clients than justified by 802.11n alone. This means that performance for legacy users connected to a Cisco or Aruba 802.11n access point clients is on average worse than would be expected if they were connected to a legacy access point with the same number of clients.

Voice and Data Quality in Converged Network
Voice and data performance are equally important in a converged network. Novarum's test used ten 802.11a voice and ten 802.11n data clients, with voice quality measured using the Mean Opinion Score (MOS) - a scale of 1 to 5 on which 4.0 generally wireline toll quality and 3.0 represents intelligibility.

All three wireless LAN systems scored MOS higher than 4.0 for upstream traffic, meaning callers from an 802.11 phone could be heard clearly. However, only Meru equaled this score for downstream traffic, with a MOS of 4.33. Aruba and Cisco both scored less than half of this number, meaning that users of 802.11 phones would have great difficulty understanding what a caller is saying.

Power Consumption
Novarum's final tests measured access point power consumption in three states - idle, ready and active. Meru again won this contest, with its active power consumption of 8.8 W less than either Cisco's or Aruba's idle power consumption of 9.2 W and 94 W respectively. In overall performance per watt terms, Meru achieved 21.6 Mbps/W, nearly twice Aruba's score and three times that of Cisco.

Summary
The tests demonstrate that 802.11n can scale to very high throughputs while handling a mix of applications and client types. However, not all 802.11n access points are equal. The Meru AP320 beats both the Aruba AP125 and the Cisco AP1250 in overall data rate, airtime fairness and power consumption, ensuring high and consistent performance for both voice and data applications while saving on energy bills.

Story By: Editor

Date : 08-04-2009