One of the hottest topics on the mobility track at Interop last month was the new Wi-Fi radio link standard, 802.11ac. I've written about the technical capabilities of 802.11ac before, but the big question on everyone's mind at Interop was, "When is this going to happen in the enterprise, and how much is it going to cost me?"
From a standards standpoint, 802.11ac is essentially here--despite the fact that it will be several months before the actual IEEE standard is ratified. However, Wi-Fi is a strange environment when it comes to standards. Normally CIOs are rightfully cautious about making investments in new technologies before the standards are set in stone. Wi-Fi is different because we have the IEEE that develops the "standard", and the Wi-Fi Alliance that "certifies interoperability."
Given my experience, I'm not particularly worried about the status of the IEEE standard so long as the Wi-Fi Alliance's certification is in place. My friend Craig Mathias of Farpoint Group, who chairs the Interop mobility track, puts it this way:
"We've been down this path before with 802.11n. The Wi-Fi Alliance's certification for the "Draft Standard" for 802.11n was available about a year before the standard was ratified. Thus vendors were producing products to the draft standard and the Wi-Fi Alliance was certifying they would interoperate; for the most part, that's the assurance users should be looking at. The Alliance also committed that any changes from the draft standard to the final standard would be addressable in firmware or software."
At the end of the day, 802.11n products started to sell (because there was a real market need); they all interoperated; and the actual standards ratification was essentially a non-event.
However, 802.11ac will be an even bigger step forward and one that will have repercussions that ripple back through the entire LAN infrastructure. The theoretical top speed for 802.11n was 600 Mbps using a 40 MHz channel and 4-stream MIMO. In reality, the vast majority of 802.11n products only got to 3-streams on a 40 MHz channel, which yields a maximum data rate of 450 Mbps.
The 802.11ac standards describe a more efficient coding system, channel bandwidths up to 160 MHz, and potentially 8-stream MIMO--but we're not getting that on day one. According to Matthew Gast, Director of Product Management at Wi-Fi vendor Aerohive and author of an upcoming book on 802.11ac, the initial products will support 80 MHz channels with 3 streams from the AP, for a maximum data rate of 1.3 Gbps. Many of the first clients to hit the market will be single-stream mobile devices, however, in which case the maximum data rate will be limited to 433 Mbps, which is close to today's practical 3-stream 802.11n data rates.
In planning for the 802.11ac migration we need to be thinking about 1 Gbps wireline Ethernet to the access point; if we're looking long term, that's probably two 1 Gbps Ethernet connections to each access point. Now some may be saying, "The maximum defined transmission rate in 802.11ac on a 160 MHz channel using 8-streams is 6.93 Gbps, how are you going to handle that on two 1-Gbps Ethernet uplink connections?" Fortunately there are a couple of things playing in our favor.
* First, Wi-Fi provides a half-duplex connection while all of our wired connections are full-duplex, so whatever speed we're talking about on the wireless side, we can divide it by two when we get to the wired side (well, not exactly, given that the upstream-downstream ratio is not 50-50, but close enough).
* Secondly, the wireless protocol involves considerable overhead compared to the wired network--given Wi-Fi's control messages, collisions (Wi-Fi is a shared-media environment), pre-transmission waiting intervals, acknowledgements, etc. So from an efficiency standpoint, if Ethernet throughput is "apples", Wi-Fi is "crabapples."
* Finally, the industry never delivered the full maximum throughput of 600 Mbps for 802.11n, so what makes us think we'll ever see 6.93 Gbps on 802.11ac?
In any event, you also have to consider the impact on the infrastructure from the APs back to the data center, and the equipment's ability to route "intelligently" versus routing every message back to the data center. If we're going 1 to 2 Gbps to the station (i.e. the access point), what kind of capacity are we going to need on the backhaul connections to the data center?
Frankly we don't know, because the backhaul capacity requirement is different in cases where not all traffic needs to get back to the data center. If traffic is flowing client-to-client, as it would in a person-to-person video connection, we could potentially do the switching at the point closest to the source and destination. We haven't seen a lot on distributed LAN switching for Wi-Fi traffic up to this point, but Wi-Fi has been a relatively low-capacity traffic source, and for the moment, traffic is going from a client to either a server or an Internet connection, either of which is in the data center. There are use cases (e.g. increased user-to-user video connections) that could change that--but is that what we're going to see in the future?
Most enterprises are not upgrading their wired and wireless networks to 802.11ac in the near term (many of my consulting clients haven't finished their 802.11n upgrade as yet), but it is time to start talking to vendors about their migration plans. Make no mistake about it, 802.11ac is coming, so it's time to start getting our ducks in a row.
The obvious question is, "Will I need new APs?" For the most part the answer is "Yes." According Lisa Phifer of Core Competence, of the WLAN vendors she's reviewed, only Cisco and Xirrus are talking about APs with plug-in modules that would allow users to yank out the 802.11 radio and plug in a 802.11ac module. However, if those APs do not support two (or more) Gigabit Ethernet interfaces, you might be upgrading the AP anyway. Given the impact on the wired backhaul, Ms. Phifer also suggests looking at the possibility of using Wi-Fi mesh technology that will eliminate the need for cable runs to every AP.
The other issue will be planning your wired network. In terms of keeping the switching point as close as possible to the two endpoints, today virtually everything passes through the data center, but if that changes, there are great efficiencies to be gained by distributing the switching function. However, every environment is different, and if you can foresee applications that could alter your traffic patterns, this is an area you may want to explore.
So 802.11ac is coming and even if you don't see it in your short term plans, it will most certainly have an impact, both in the wired and wireless environments, in the longer term.
Follow Michael Finneran on Twitter and Google+!
@dBrnWireless
Michael Finneran on Google+
