Nov 08, 2012 (05:11 AM EST)
Tablets Cause Wi-Fi Stress: Truth And Fiction

Read the Original Article at InformationWeek

Wi-Fi wireless LANs (WLANs) have been displacing wired Ethernet as a preferred means of client access for years. Two years ago, we asked readers about wireless plans: Our InformationWeek Wireless LAN survey found about 40% of respondents predicting that they would largely replace wired access infrastructure within five years. Fast forward and it's likely that today's torrent of mobile devices rushing into the enterprise, whether resulting from formal, front-door BYOD policies or furtive, or back-door infiltration, is accelerating the timetable.

Worse yet for early adopters, the changing device mix means existing WLAN installations could face significant stress, if not outright gridlock, unless network managers make some architectural changes. Yet enterprises could learn a lot about WLAN design from educational institutions, which have been among the leaders in wireless deployments and have some of the highest device densities of any environment. As Ohio University's CIO, Brice Bible, is quoted in discussing his campus' recent WLAN upgrade, "Wireless is by far the most popular access method on our campus and students are bringing more mobile devices to campus than ever before."

The challenges start with a dramatic increase in numbers, a situation that new devices like the iPad Mini and Nexus, as profiled in our latest research report, will exacerbate: The client count per employee could double or triple as users augment their company-issued laptops with smartphones and tablets. There are also inherent Wi-Fi hardware design limitations imposed by mobile devices optimized for portability and battery life, not network performance.

WLAN equipment vendors have been fond of scaring customers with a Gartner report claiming that "enterprises deploying iPads will need 300% more Wi-Fi." (See a PDF of the entire report here).

The 3x figure is derived from a simplistic extrapolation of differences in transmit power between the Wi-Fi radio in the iPad 2 and that of a typical laptop, which the paper claims to be 6 decibels (10 dBm vs. 15-17 dBm). Aside from being out of date, which the online copy of the paper now acknowledges with this disclaimer from Gartner: "(Note: This document has been archived; some of its content may not reflect current conditions,)" it's overly simplistic.

First off, the newest iPads and iPhone 5 use new wireless chips with output power comparable to a laptop. Secondly, the 300% number is derived by taking a simple power ratio, using basic math any first year electrical engineering student would know, without accounting for other physical or technical factors.

A 6 dBm difference translates to four-times the power, i.e. 3dB equals double the power, meaning theoretically you would need to space access points (APs) much more closely when using tablets to achieve the same average signal strength versus an all-PC environment. But this makes several assumptions that may not be true in practice, nor have they been demonstrated, at least in this paper, by actual testing: (a) that all the APs are already operating a maximum power (probably a safe assumption, but not necessarily true since doing so in some indoor environments may lead to excessive cross-channel interference); (b) that Wi-Fi performance is directly proportional to signal strength and that an iPad with half (or worse) the signal strength will have lower performance than a PC; or (c), that signal strength is the most important factor in tablet Wi-Fi performance. According to testing done by Aerohive in high-density, tablet-rich classroom environments, this is almost certainly not the case.

You see, the bigger limitation of tablets and smartphones isn't their radio power, but channel capacity. Perhaps the most important feature of 802.11n is MIMO (multiple-input, multiple-output) radios, namely the ability to support multiple spatial radio streams for a single connection. But this requires multiple antennas and more power-hungry, multi-stream Wi-Fi chips, two design requirements at odds with small, thin form factors and long battery life.

Thus, every current smartphone and tablet is a 1SS (single spatial stream) implementation, although things are a bit better for dual band devices like the iPad and iPhone 5 as they support a single stream on both the 2.4 and 5 GHz frequency bands. But with 1SS clients, everyone is still trying to share the same airtime on a given channel -- kind of like truckers on CB radios -- which leads to a massive RF traffic jam when a classroom of them are trying to talk at the same time.

Here's a typical example courtesy Aerohive Chief Wi-Fi Architect Devin Akin. He starts with some basic design facts, namely that an iPad needs 2 Mbps of sustained throughput to run multimedia (e.g. video streaming) applications and that there are 30 of them in the average classroom. Furthermore, a dual-radio (2.4 and 5 GHz) AP can process around 60 Mbps, or 30 Mbps per channel while using 80% or more of the available airtime on a single channel; any more leads to airtime contention (multiple clients trying to talk over the same frequency at the same time.) Thus 30 iPads times 2 Mbps per client nicely matches the throughput of a single AP, assuming you can steer half the clients to the 5 GHz band and keep them there.

Fortunately, band steering is a common feature of today's enterprise APs. As long as your 30 clients get a good enough wireless signal to maintain streaming throughput, adding power doesn't help; the primary benefit of more densely packing APs is to provide more RF time slices, since to avoid interference, adjacent APs are on different Wi-Fi channels.

Of course, if your environment isn't as client dense as a school, you might not saturate airtime even using widely distributed Aps. And here the Gartner paper does point out another potential problem: namely that the iPad (and now, iPhone 5) will aggressively 'downshift' to the 2.4 GHz band if 5 GHz performance drops off -- behavior I have witnessed many times. This normally wouldn't be a big issue if it was equally aggressive about 'upshifting' when the 5 GHz signal improves, but this isn't the case. This means that older, slower 2.4 GHz devices could get crowded out by fast-talking iPads looking for a better signal in areas of sparse coverage.

In sum, the influx of Wi-Fi tablets and smartphones into the enterprise undoubtedly means enterprise WLANs will need more APs, but the scaling factor is more likely proportional to the number of new devices, not their power output. So unless you're doing new mass iPad deployments (say in schools or hospitals), the amount of new WiFi you'll need is probably much less than 300%.

From SDN to network overlays, emerging technologies promise to reshape the data center for the age of virtualization. Also in the new, all-digital The Virtual Network issue of Network Computing: Open Compute rethinks server design. (Free registration required.)