With 11ac, The WLAN Industry Owes Customers A New Kind Of Network Switch

I realize I’m beating the 11ac thing up pretty good lately, but I think I finally hit on what bugs me about the way the new hot technology is being brought to market. What I’m about to describe is more of a BAN issue (BAN=BigAss Network, where APs are counted in the hundreds or thousands) and not so much of concern for smaller environments.

802.11ac is being delivered in rather bizarre (for the customer) “waves”.

  • Wave 1: Data rates to 1.3 Gbps. You’ll do fine (for most new first wave APs) with a single Gig uplink, and many new APs will work on 802.3af POE, not yet requiring .3at. Fine, good. No real squawks.
  • Wave 2: You get the joy and cost of recabling your environment to add a second Gig uplink, doubling the number of switchports in use for the WLAN and configuring Etherchannels, and depending on what vintage switches you have- upgrading them for latest POE standard, all to help get to data rates likely to realistically be between 2 and 2.5 Gbps best case.

And this is where I say “time out”. I’d like the WLAN makers to bear some of that Wave 2 logistical pain. And I want them to get creative to take the onus off of the customer. Here’s what I want:

  • In simplest terms- I don’t want to use two cable runs. And I don’t want the complexity and risk of 4000 more Etherchannels for my APs. But I still want the benefits of 11ac Wave 2.
  • I would like the WLAN vendors to put their brilliant minds (and that I do mean sincerely- these guys and gals accomplish amazing, amazing stuff) to work to come up with a new switch or mid-span injector. Here’s the requirements:
    • No feature bloat. May not even need to be VLAN aware.
    • Provides lots of PoE
    • Somehow puts 2 Gbps of uplink to an AP on a single UTP run without requiring me to configure a port channel
    • Cost effective (by customer standards), no licensing BS, and ultra-reliable

Spare me the lecture that there is no such thing as 2 Gig Ethernet, and that what I’m asking for would be based in no existing standard. The WLAN industry has long since turned it’s back on standards and interoperability, which is why vendor lock prevails. Other than PoE and what comes out of the antenna (and even that can be a dubious discussion), the mention of standards is a joke in the WLAN industry as each vendor authors their own technical magic. So be it- I just want new magic and don’t care that it’s not exactly Ethernet in the middle.

I’m OK feeding this new component a 10 GB uplink that it then divvies up into auto-configured 2 Gbps AP uplinks of some proprietary protocol. Or feeding it 2 single-gig ports on my wireless management VLAN that it then magically muxes into a 2 Gbps, big powered uplink that connects via a single wiring run (of excellent quality, of course) to each AP. At that point, all of MY work was done in the closet, and I didn’t run a slew of new wires for my wireless network.

If we don’t get something disruptively creative on the wired side to go along with 11ac, pretty much any TCO discussion on new 11ac ownership presented by WLAN vendors will be incomplete at best, and poppycock at worst. I’ve seen both announced and unannounced 11ac products- and the prices are pretty steep (well, except for Ubiquit-). But we’re supposed to believe that 11ac lets us draw down the wired network considerably, and so be willing to buy into a higher premium for wireless. But… adding lots of new switchports and cabling runs (not trivial in many environments,  can add hundreds of dollars in cost to real TCO for each AP) has to be considered.

As a customer, I feel OK asking- because the customer is always right (well, except when they’re wrong). So… when will my new non-standards-based 2 Gbps mega-PoE switches arrive?

15 thoughts on “With 11ac, The WLAN Industry Owes Customers A New Kind Of Network Switch

  1. Frank

    At least one baby step in the right direction would be demanding proper LACP support. This means both supporting LACP from the initial boot firmware up in the AP, and equally important, ensuring that the edge switches hosting the APs are able to scale up LACP to that degree. I’ve seen too many switches that support LACP, but have such weak CPUs that they run out of LACP processing time long before they run out of physical ports. For wireless deployment, I’d ideally like to see those switches capable of running LACP actively on every physical port simultaneously.

    That obviously wouldn’t help with the physical cabling problem, but for those lucky enough to have extra runs to their AP locations, it would at least take a little of the sting out of the configuration part.

  2. John

    Great post and thank you for your other 802.11ac posts as well.

    Another thought crosses my mind….

    If I have a 24x port switch (all 24 = PoE) and I have 12 phase 2 802.11ac APs… but each AP only requires 1 Drop for power….but 2 (1gb) drops for traffic….. I guess I just wind up using/wasting 1 PoE drop per AP. This may not be an issue to a lot of folks but not everyone has VoIP throughout their network, not everyone has PoE switches deployed throughout their network. I cringe at the thought of purchasing a 24port PoE switch and patching 12 Non-PoE drops into it or a 48port and patching 24 non-PoE drops into it.

    or can I expect some sort of LACP/Etherchannel magic to work across 2 different switches… IE my AP Eth0 is patched into the PoE SW1 for 1gb link and PoE and the APs Eth1 is patched into non-PoE SW2 for the additional 1gb link but not PoE (thus not wasting a second PoE port on PoE SW1).

  3. wirednot Post author

    Thanks, John and Frank. The complications of wiring for 11ac kind of remind me of the early days of VoIP. You’d be presented with incredible TCO estimates from vendors that showed staggering savings over legacy telephony- as long as you didn’t include power concerns. Add in PoE and backup UPS to come anywhere close to the power-out uptime of traditional Telcos, and the optimistic estimates quickly deflated. That didn’t mean VoIP wasn’t a good move, but it did mean that vendors were not being complete with the “and this is what it will REALLY cost you” projections. And as I mention here, small environments certainly won’t get hit as hard as larger ones, who will definitely feel the effects of wiring and licensed feature “taxes” because of their size.

  4. A

    I believe this whole multi-gig thing is a bit over-hyped.

    Achieving those speeds (not rates, real speeds) is not easy. Consider that the max RATE with 80MHz channels and 4SS is abot 1700Mbps. I do not see enterprises casually using 160MHz channels, and we have yet to live and see 8SS APs and chpsets. Ok, let’s add another 300 Mbps for the 2.4Ghz .11n radio with 4SS@20Mhz for a total of 2000Mbps RATE, recall the “rate to real throughput” ration, and move on…

    Even with 100% ratio, I doubt that APs will be able to transfer 2Gbps+ worth of real traffic with in-line encryption, firewalling and all other fancy stuff embedded in modern distributed WLAN architectures. With pure bridging back to controller – maybe. But then you need terabit controllers 🙂 Do note the price wars in the enterprise AP segment, as the scale of deployments grows. We still have to wait and see cheap überAPs 🙂

    And finally, who in the world will consume such bandwidth (and will depend on it so critically, to justify all the wired/wireless upgrade costs)? Even with 8 spatial streams you will be able to transmit to 8 clients at a time. Given that modern wireless networks gravitate towards ultra-portable laptops, tablets and smartphones, I doubt they will be able to consume (not to say, generate) any decent amount of data to fill up the entire 1G uplink. We’re talking access layer, not backbone or entire datacenters going wireless, right? Look at the whole “Gigabit to desktop” thing – who’s using the whole gig?

    So, IMHO, even a single 1G wire for your typical AP (not a bridge or any other special case) will last long beyond the Wave 2 induction date, or even further. Unless some new traffic-crazy application is invented. 🙂 The whole point of 802.11ac in the Enterprise in much more about cell capacity, than sheer performance..

    1. wirednot Post author

      Well said, and quite my sentiments as will be reflected in my NWC column very soon. I wish the WLAN industry would tone it down on the multi gig message, but it wouldn’t market well…

  5. Mike R

    “The WLAN industry has long since turned it’s back on standards and interoperability, which is why vendor lock prevails. Other than PoE and what comes out of the antenna (and even that can be a dubious discussion), the mention of standards is a joke in the WLAN industry as each vendor authors their own technical magic.”

    I disagree with this completely. Yes, there are extensions and modifications to standards in WLAN, but not “turned it’s back” or “a joke”. There is plenty of data to show that standards are adhered to (WiFI Certification!!!). There is backward compatibility/interoperability built into each new standard!

    Perhaps you are talking about things not part of the standard, such as mainboard power supply? Whether its 24V, 48V or .af? Again, not part of the WLAN standard. Other standards such as 802.3 and 802.3af and what not are still standards that manufacturers are following.

    Where issues usually arise are: 1) standards did not sufficiently cover a given implementation 2) standards are behind the NEED/DESIRE for a particular thing 3) proprietary modifications for technological improvements. 4) stupid patent trolls

    I am not aware of any vendor lock for standard WiFi equipment (Enterprise stuff with channel blankets and other such technologies are clearly intended to solve issues that standards cannot, and its not claiming to follow standards, AFAIK). Non-WiFI WLAN like Wimax does have proprietary vendor locks, but not due to standards, but due to commercial reasons.

    Perhaps if you provided a few examples of what you mean, I might better understand your position. But its certainly not to the degree you’re insinuating that WLAN makers do their own “technical magic”. Even early stuff, will be “based on draft blah”.

    Lastly, on a different note, why would you go and actively replace out 2000 AP’s at once? Unless there is an absolute need to squeeze out every bit of Mb (which should have just gone and wired up the building in the first place) (video editing, SAP database, etc), just replace as the equipment ages/dies out. Any new development/projects can go 11ac, but leave existing infrastructure alone unless its TRULY NEEDED. This isn’t like going from 11b to 11n. This is going to be a marginal improvement over 11n when all things considered (onlt the extremely close clients with no noise and interference will get 11ac rates). In other words, it’s a “nice” to have, not a “need to”/”must” have.

    I think most everyone is forgetting that with Ethernet, you have DUPLEX. You can have 1Gbps in one direction, or 2Gbps total from both directions. It’s not reasonable to light up 2000 AP’s with 2+ Gbps links as you will have tons of other infrastructure that would need to be upgraded as well. You need a TON of CPU to handle 2+Gbps in one direction, and it’s just not realistic or smart to actively implement this for people who don’t NEED it.

    1. wirednot Post author

      Thanks for the input, Mike. Simply put- Can you attach a Cisco AP to an Aruba controller? Can I use an Aerohive AP on my AirTight dashboard? From my Meraki manager, can I manage and monitor Ruckus controllers and APs? The answer of course is no, on all accounts. Hence, we have vendor lock. Different brand Ethernet switches and routers can be stitched together with Ethernet, BGP, etc, but no such luck with WLAN hardware. So, I have to disagree with whatever leads you to a different conclusion. Every vendor does do it their own way, and every vendor’s way is the best (just ask ’em).

      I don’t disagree with “need” versus “nice to have” at all, and in no way imply that you “have” to do anything. At the same time, when you have a large environment it can get ugly for support when you have a checkerboard of technologies to keep track of- just a bit more work and some environments don’t want the complexity especially if it means different code versions for new versus old APs.

      The industry so over-fed us the hype on 11ac that they forgot that people need to wire for wireless and sometimes need to plan for 10-25 year refresh cycles, that’s where much of this blog comes from.

      1. Mike R

        Just saw the 10-25 year refresh cycles thing. I will say that almost no one will be thinking of wireless as a 10-25 refresh. Technology is moving too fast. 11b is 14 years old and hasn’t really been sold as a new product for like 8 years. Wireless is getting CHEAP, and cheap doesn’t mean well engineered, long lasting. It means throw it away and replace every ~5 years.

        From a manufacturer point of view, they want to sell you new gear over and over, and not have to support old, non-money making gear that gets more expensive and harder to support over time (parts go EOL in shorter times than say 10 years ago). So make things cheap and easier to replace (quantity over quality).

      2. wirednot Post author

        Mike, I’m not suggesting WLAN systems refresh in 10-25 year cycles- but in historic/complicated buildings, THE WIRING (which I mention) might. The point is the WLAN Industry (made up of WLAN vendors, by my estimation) cant get their 11ac stories consistent on uplink. After Wave 2, will I need more than 2 UTP? Will I need fiber? Something else? For hose of us who only get to wire large, old spaces every couple of decades, it’s an important question. Some technical perspective on the reality of true throughput capabilities suggest that regardless of whether 11ac ever lives up to it’s 6+ Gbps promises allowed by standard, you’ll still never REALLY need more than 2 1-gig uplinks. But other voices disagree. Regardless, my original point is that even for 2-Gig uplinks, I’d like to not do port channels. They suck when you have thousands of them. I’d like to see a new switch that puts 2 Gig on a single UTP- and in private switch vendors will tell you it’s possible. I’m asking for practical innovation, brother. You don’t have to agree- that’s what makes America great.

  6. Mike R

    Controller’s are not part of the 802.11 standard. At least not when I search. I would consider a controller to be part of a single product, but with the mainboard portion with the radio portion being separate. So an Aruba mainboard with software isn’t going to work with a Cisco radio being ‘half’ of two different products. They are sold to manage THEIR products, right?

    It is completely your choice to use specific products from specific companies to provide you with features which are not part of the standard. Its certainly not a must that every company is going to be locked into. I find those controller systems to be limiting and costly (while they do have their benefits).

    Based on this controller point, I would say your request for a vendor solution should come from those vendors supplying you hardware controllers. It’s not the ‘WLAN industry problem’ to deal with, it’s “WLAN controller makers” that need to deal with it. They make their own solutions, they should fix it themselves.

    You are locking yourself into vendor gear for controlling/management, which is not the same as WLAN or standards.

    But I see what your point is. I’m not much of an enterprise indoor guy, so your point didn’t come to mind at the time (though on re-read, it should have been obvious to me). I guess where we disagree, is I don’t think this is an “industry” problem for all WLAN mfg’s to solve, it’s those stupid hardware controller makers, who are a limited group, to come up with solutions for their own products that they made and sold outside of any standard.

    As for vendor solutions, given that you have 1Gbps in each direction, the controller would have to scale significantly enough that it would be way more expensive than benefit. Most likely, they will have a solution where you continue to use a single gigabit port and single Ethernet, given actual good throughput is likely not going to exceed 1 Gbps in either direction.

    Not all 11n devices have gigabit ports despite being capable of handling over 100Mbps.

    Lastly, I’ve got the pessimistic view that wifi will be way too polluted to be able to do 160MHz channels with 4×4+ antennas outside of a lab (I’m in an apartment building, so my experience is all channels in use). 4×4 won’t be purchased in volume and they won’t come down in price like 3×3 stuff will (IMO). Maybe 3×3 @ 80MHz will be realistic for high volume to get cost down, but higher than that, major costs for size, antenna, CPU, etc.

    Rather than increasing speeds, work needs to be done on the protocols to reduce excessive spectrum from being used on retries and other management (searching for AP’s, etc), and getting rid of 11abg devices. Until that happens, getting these very high rates and modulations isn’t going to be a regular occurrence and not something that I think these hardware controllers are going to address anytime soon, as they are not yet needed.

    1. wirednot Post author

      Given that the top 3 WLAN vendors all use controllers, it’s close enough to call an industry issue regardless of philosophy on controllers. But even where you don’t have controllers, you still need AP uplinks from PoE switches. I’m saying I’d like to see evolution- 2 Gbps and PoE on single UTP on a new switch. I’m not arguing your analysis of wide channels or anything else- I’m wanting help from the industry on future-proofing my very expensive cable plant made up of thousands of runs of UTP by coming up with a new switch. Its OK to ask for evolution, isn’t it?

      1. Mike R

        RE: Wiring. Yeah, I missed your point there. I was thinking of wiring specifically for purpose, but I neglected to think of the pre-wiring that goes into the walls when buildings are constructed for future purposes and the incredible cost to re-cable after the fact. I see you what you mean by that point. (Again, I’m totally forgetting indoor perspective).

        Unfortunately, the 2 Gigabit problem is not just limited to WLAN but Ethernet, and is more difficult to solve due to needing to pass power over two of eight wires in addition to cross talk avoidance. If someone solved that problem, patented it, they’d be super rich in 2 seconds flat. It’s actually better that it applies to more than just a limited industry, as major adoption is needed to bring it down to commodity pricing.

        I’m really glad to have come across this Blog today to learn more about this. A few years ago, when I wanted to increase my home LAN from 1Gbps to higher, I walked away thinking that to get to 10G, you needed fiber and it was expensive, and Infiband was the “cheaper” solution, which was still extremely high and not worth the few times it would be beneficial at home.

        After a bit of googling, I haven’t seen any showstopper reason why 10GBaseT couldn’t be leveraged, aside from cost and mass production. Having a team go into a Fortune 500 wall sure negates a few hundred dollar NIC. Since the deployment of 10G is in the millions per year and 100M/1000G is in the billions, the difference is cost is like sky high. But as more people get fiber to the home, and 4k TV’s, it will help drive the demand for cheaper 10G gear.

        I was surprised to read on wikipedia a max range of 55 m (cat 5e or 6), 100 m (cat 6a or 7) for 10GBase-T. I imagine a building will often have drops closer to 100m than 55m, but since you mentioned that wiring was for wireless, which would have been in the last 15 years, it may have cat 6 in some cases, which will probably be iffy, given cable would meet and/or beat the minimum frequency specification (like when cat5 works for 1000BaseT). The vendor solution might be to use a more robust modulation that doesn’t get 10G, but if 2G is all that is needed, maybe a modulation/algorithm change can handle 100M…

        So for existing building with less than Cat 7, I’d suggest sticking with wave 1 for the next while. For new buildings, putting in cat 7 should future proof (at least for a decade) and allow for better wave 2 upgrade path. Perhaps a repeater device with additional power where possible…

        Well, after today I’m more inclined to believe there will be a vendor provided solution for you in the future to reuse existing cat5, I just think it might be a few more years than you’d like before it makes economical sense (resisted urge to say “cents”).

  7. Mike R

    Another thought… will as many AP’s still be needed after the retrofit? Greater power, beam steering, better receive sensitivities vs 5 year ago AP’s, etc. This could help leverage existing wiring as well, especially if wiring ran to locations where two or more AP’s were installed (ch1/6/11 installs).
    Also, some older installs might be wired for 2.4G only and 5GHz only separately, and new AP’s would be dual band to reduce legacy AP count and reclaim a cable run.

    So you’ve convinced me that this issue (get 2-10Gb over 100m of cat5) should be solved and made into a standard. 😉

  8. Pingback: 11ac- Unintentional Noise | wirednot

  9. Pingback: 3/4/2015: Sticky Clients, New Tools, Help Manage the Message (and the Environment) | W to the 4th

Tell me what YOU think.

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s