Fiber-to-the-AP? Have you lost your mind?

Your homework assignment:
http://en.wikipedia.org/wiki/Category_6_cable
http://en.wikipedia.org/wiki/Category_7_cable#CAT7

…and let me tell you a brief story…

Once upon a time…in a land only 3 hours from my home…there lived a healthcare provider.  This particular healthcare provider has recently asked some VARs in the area to bid on their business. The network is a mess, to put it nicely.  So, it’s time for an upgrade, and one of the VARs has recommended running fiber to every AP.  How do I know? I was asked by the powers-that-be within the healthcare provider to evaluate whether this is a good idea or not.

If you just experienced a new facial tick, convulsions, or nausea at the thought of the cost of running fiber to the AP (FTTA), join the crowd…so did I.  I honestly didn’t know whether to laugh or be sick at the thought.

If you’re one of those people who like to argue just for the sake of arguing and think that you may convince me that FTTA or multiple copper runs are good ideas for anything other than extreme corner cases, please do not waste your breath. I’m writing this blog because the entire notion of FTTA in 99.9% of use cases is stupid, borderline criminal, and someone has to speak out against it. I will also illustrate how multiple copper runs to an AP are completely unnecessary as well.

As of today, here are the only exceptions that I’ve been able to come up with, and each are corner cases.

     * Equipment.  There is exactly one equipment exception to my assertion about FTTA, and that is large Xirrus arrays. Xirrus’s large, multi-AP arrays can connect many hundreds of clients across an array housing up to 16 radios. Combined with the fact that they need proprietary PoE for these large arrays anyway, you may could justify some of the additional costs associated with fiber. For 2 and 3 radio APs, you need copper, and only ONE copper run per AP.

     * Interference. If the copper cable must traverse areas where there is an extreme amount of EMI that could be picked up along the length of the cable, then either strongly shielded cable (Cat7) or fiber would help. If there’s so much EMI that it can significantly penetrate Cat7 STP, then I personally wouldn’t want to work in that facility. #JustSayin  If it were a nuclear power plant or something, then perhaps, but that’s an extreme corner case.

     * Distance. If you had one wiring closet for a huge building, and you needed cable runs that far exceeded 100m, then fiber might be a good option, but a VERY expensive one. I’ve never seen such a scenario personally, and again, I believe that would be an extreme corner case. Of course the same is true for outdoor APs.

     * Lightning. If you have an AP in an area where it could easily get struck by lightning, and you want a way to prevent an electrical surge from getting back into your wired infrastructure, then fiber will do the trick.  This isn’t for all APs mind you…but again, only for the extreme corner case.

The justification behind the MSP’s suggestion was “future-proofing”, though it’s pretty obvious that they know not what they speak of, which is a clue as to why the network is currently in shambles.

OK, let’s start at the beginning…

     * Cat7 and Cat6a cable supports speeds up to 10Gbps (of throughput)
     * Cat7 and Cat6a cable allows use of PoE+ to the AP now and will accommodate future versions of PoE down the road
     * Be aware that with Cat7 cable, you need GG45 connectors for backwards compatibility and that Cat7 cable will cost more than Cat6a
     * Cat6a is, by far, the best bang for the buck when future-proofing. 10Gbps data rates without a reduction in distance, it’s about half the price of Cat7, and it’s available everywhere.

By going with high-quality copper connections to the AP, you avoid the following costs:

     * The cost of the fiber itself, running the fiber, and maintenance of the fiber (overall more expensive than copper and more skilled labor and tools required)
     * Fiber connectors (as much as $15 each, with TWO needed per fiber run of course)
     * The cost of fiber/copper media converters (typically over $1k each).  TWO needed per fiber run (one at each end)
     * The cost of two 10Gbps XFP transceivers per media converter (for a total of FOUR per fiber run)
          A. Example costs of 10Gbps Media Converters and 10Gbps XFP optical transceivers
     * The cost and man-hours of replacing SFP+/XFP connectors and/or 10Gbps media converters when they fail

General cost per AP breakdown:
     * AP costs ~$1k
     * Fiber run
     * Two 10Gbps media converters (~$1k+ EACH = ~$2k+ per fiber run)
     * Two 10Gbps XFP transceivers per media converter (~$1k+ EACH = ~$4k+ per fiber run)
     * Two power runs: One for the AP & media converter at the AP end. One for the media converter at the switch end. (~$400+ EACH = ~$800+ per fiber run)
          A. As an alternative, you can leave out one media converter and power run by buying all XFP switches with optical transceivers, but the cost of those switches, with their associated optical transceivers is obscene.

Yet another drawback of FTTA is the inability to reboot the AP remotely via PoE unless media converters are manageable via IP.

The reality of Wi-Fi for the next 10+ years is:

Now

Even though today’s APs have two radios, with theoretical data rates of 1.3Gbps for 11ac on 5GHz and 450Mbps for 11n on 2.4GHz, those data rates can realistically only be used in homes.  Nobody in their right mind designs for or uses 80MHz channels in the enterprise (which are required for 1.3Gbps data rates) because there’s not enough license-free spectrum to allow for a reasonable channel-reuse plan. Additionally, 256QAM modulation (also required for 1.3Gbps) is only reasonably possible for <50 feet in a perfect RF environment, and thereafter falls off to 64QAM and lower orders of modulation quickly. Enterprise-class 4×4:4 systems won’t be released into the market until well into 2015, so we’re stuck with 3×3:3 (current units) for the large part of another year as our top-end APs. When you take into consideration that most clients (phones, tablets, cheap laptops), aren’t capable of doing more than 2 spatial streams, and that’s even if they have 11ac at all (which most devices don’t), then you quickly realize that all of this hoopla around needing multi-gigabit backhaul from the AP is straight-up hype.

Using the highest-performance AP in the market, when the AP is loaded with 11ac 3×3:3 capable laptops, and using both 2.4GHz and 5GHz simultaneously, and using 80MHz channels (which you don’t do in the enterprise), you would be luckyif all laptops are moving giant files and are practically sitting on top of the AP, to see 900Mbps of throughput.  This means that no phones, tablets, or cheap laptops can be connected to get this kind of throughput. It also means that no devices can be further from the AP than about 30-40 feet and the RF environment has to be amazingly perfect.  In my 15+ years in Wi-Fi, I’ve never seen any scenario happen like this.  In a densely-populated, heavily-loaded environment, you might see an average of 10-12 Mbps of throughput across an AP…and that rarely happens in the enterprise. Even the 10-12Mbps scenarios are mostly branch office environments where one AP is serving 50-100 heavy users.

Next Gen

Starting in early 2015 and going through the next 5 years, 4×4:4 MU-MIMO 11ac devices will be available. 4×4:4 MU-MIMO devices will either have one 11ac radio and one 2.4GHz 11n radio (like now) or will have two 802.11ac radios. If you take the worst case of two 11ac radios, and configure them for 40MHz channels (maximum in the enterprise due to current lack of spectrum), then you still only have two radios times 800Mbps theoretical data rate. When you subtract ~40% to yield actual throughput, that’s a total of ~950Mbps across both radios combined – BEST CASE POSSIBLE. Everyone seems to agree that client devices will continue to be limited to a maximum of 4×4:4 in the future (due to battery life), so the next best feature will be serving multiple downlink clients simultaneously.  How much value will this add?  Who yet knows in the real world, but my guess is <30% overall downlink gain, as it mostly helps make 1SS devices more spectrum efficient.

Next Next Gen

Even if the AP manufacturers could figure out how to make 8 spatial stream capable devices (and they won’t because of the AP size requirement to house all of those antennas), then the maximum you’ll ever see throughput attain, even theoretically in a perfect scenario, would be ~4Gbps because the maximum possible data rate for 802.11ac is <7Gbps.

Even when designing everything wrong (e.g. 160MHz channels, standing so close to the AP that you could cable to it, and so on), you won’t get throughput values that require fiber or multiple gigabit-capable copper runs. It’s highly likely that 4×4:4 11ac will be 11ac’s last generation, and we will start moving to 802.11ad (60GHz at up to 7Gbps data rate) and/or similar technologies for in-room connections while continuing to use 11ac for inter-area roaming and access.

In none of these scenarios, which cover 10+ years, will you ever need fiber. Cat6a and Cat7 will last you between one and two decades for wireless connectivity, even if every worst-case scenario happens.  Don’t forget that even with today’s de facto standard of Cat5e, we can do 1Gbps, and I’ve never seen any AP (outside of an extreme corner case like a stadium) that was sustaining more than 20Mbps during business hours, and it’s typically <10Mbps in most very busy environments.

Summary

Buy Cat6a copper cable. 

Fiber-to-the-AP….Are You Completely Insane?