The Network Impact of 802.11n

Hi Steve,

Thanks for your questions. I'll number my answers to match your questions.

1) Yes, 802.11n has significantly more protocol overhead than Ethernet. With 802.11n, as with 802.11a/b/g, there are three types of frames: management, control, and data. Ethernet only has data frames, so management and control frames are overhead, when compared to Ethernet. Additionally, the 802.11 protocol uses CSMA/CA as a collision avoidance protocol, which introduces additional overhead that wouldn't be experienced by Ethernet. Collisions, random backoff, protection mechanisms, larger PHY and MAC headers, RF interference or poor signal strength causing retransmissions, and all sorts of other protocol overhead is also present in 802.11 that isn't present in Ethernet. Small frames (aka packets) are a throughput killer in Ethernet and Wi-Fi, but it's even more apparent in Wi-Fi because the protocol overhead is so large. For this reason, Aerohive has implemented dual-core CPUs in its 300 series APs, which keeps packet rates high when lots of small packets are on the network.

2) The ROI of VoWiFi is tremendous, which is why there is such a variety of enterprise-class VoWiFi handsets and badges these days. Segregation of VoWiFi is already happening because handsets typically use PSK for security due to a lack of fast/secure roaming mechanisms for 802.1X/EAP. For that reason, they are often placed on their own WLAN Profile (SSID) and are heavily filtered to maintain a strong security posture. This, however, does nothing toward helping with the small packet problem though because all of the packets from all of the WLAN profiles and client types are on the same RF medium.

3) VoIP itself doesn't really have "high overhead" per se. It uses small packets that are sent bi-directionally at regular intervals. You could, however, say that voice packets have RELATIVELY high overhead because the relationship of overhead-to-data is high. 802.11n has frame aggregation, called A-MPDU, which will allow a transmitter to send a set of voice packets together, but the problem with doing this is latency. Having the packets arrive in a timely manner is far more important than sending them efficiently when it comes to the quality of the voice conversation. Additionally, use of A-MPDU would only apply to 802.11n phones, which for now is mostly limited to soft phones (running on computers like laptops). There are relatively few 802.11n phones on the market right now. Holding up some voice packets while awaiting others is a bad idea, even if both sides of the wireless link is 802.11n capable. One good thing about using A-MPDU, in one particular circumstance, is that if the voice client is also a data client (like a laptop using a soft phone), then both data and voice packets (uplink and downlink) will be sent bundled together (since they are destined to the same receiver). If you're using an 802.11n voice handset, then uplink packets will not be bundled together due to 20-30ms timing intervals.

4) Right now, the Wi-Fi Alliance is STILL working on the Voice-Enterprise certification, which is long overdue. It is a very important step for the industry, and we're hopeful that the Wi-Fi Alliance will then produce a vendor-neutral VoWiFi implementation best practices whitepaper. So far, everything on the market is vendor-specific and focused either around a specific handset or a specific vendor's infrastructure. Another place to look would be CWNP.com, as they are a vendor-neutral education organization.

Hope this helps,

Devin