New Cisco 802.11ac 3800?

By rfenster posted Oct 03, 2013 03:48 AM


Don't worry, you didn’t misread the title. The reality is, there’s a lot we can read into Cisco’s recent Aironet 3700 series 802.11ac clip-on AP announcement. ‘Clip-on’ being the operative word here.


It turns out that neckties and APs have a lot more in common than I suspected. I’ll never admit to it in person, but the pure utility and simplicity of a clip-on tie carries a certain allure. Kind of like fanny packs and child-leashes. We all secretly like the story, but it only takes a split second to realize it’s a bad idea.


Now you’re probably thinking: What does this have to do with wireless APs? Well, you may not realize it, but by announcing the 3700 clip-on AP today, Cisco also pre-announced its successor, the 3800. Let me explain.


It helps to do a little history lesson and we only have to go back a couple months to the Cisco 3600 to see what’s in store for us with this new AP.


  • Act 1: June of 2012 Cisco announces the 3600 three-stream 802.11n AP that is “future proof” for 802.11ac using a clip-on module.
  • Act 2: August 2013 Cisco makes the 802.11ac clip-on module and supporting software available.
  • Act 3: October 2013 Cisco announces the 3700 three stream 802.11ac AP to replace the 3600 as the primary 802.11ac phase 1 AP.

Wait….Why did Cisco announce another three stream 802.11ac AP, when they just fulfilled their promise of future-proofing the 3600 for 802.11ac? It turns out Cisco didn't have much of a choice.


It all becomes clear when you look at how an older clip-on AP performs after you attach a new antenna module to it. I'll take you through the gory details.


First we can look at what happens when you try to connect a bunch of 802.11ac clients. We recently tested this using Veriwave, showing UDP downstream encrypted traffic.


Veriwave Multiclient S.jpg


You can see that Cisco’s 802.11ac radio gets absolutely crushed when it goes beyond 50 clients. The Aruba AP 225 keeps humming up to 255 clients. The older CPU and inadequate memory of the Cisco AP simply can’t handle the load of all these clients.


Now, let’s look at single-client TCP performance as measured by third-party research firm, Miercom, in their most recent test report.


Meircom 5 GHz 11ac 3ss v2.jpg


Miercom measured over 50% better rate-vs-range performance for 802.11ac with the Aruba purpose-built AP 225 compared to the Cisco clip-on 80211ac AP. This report validates the performance numbers I showed in a recent blog.


Bottom-line, you can’t teach an old AP new tricks. The same will be true in 2015, when the next wave of 802.11ac functionality comes along and Cisco tries to clip it on and make it work on the 3700. What’s more likely is that Cisco will do some hand waving and try to get you into a new 3800 802.11ac AP two months after the 802.11ac Phase 2 module is available.


Here’s the reality. A clip on architecture is riddled with compromises.




  1. Antenna placement is limited to the surface area of the clip-on module. Not ideal for maximizing RF performance.
  2. Design compromises lead to power inefficiency and require non-standard POE. Cisco has already stated that the 3700 will not run full 802.11ac functions on standard POE.
  3. The network uplink is not designed for the future requirements of new clip-on modules. The 3700 assumes that the next wave of 802.11ac will only need one GigE port uplink. This is not a safe bet.
  4. The CPU and memory that are designed into the AP today will not fully support the performance of future iterations of 802.11ac.  Did we forget about Moore’s Law? You only have to look at the 3600 to see how that movie ends.
  5. You still have to go in the plenum, remove every AP, retrofit each one with a new clip-on module and then re-install. All that work for poor performance and a poor user experience.

The “hand-tied” purpose-built difference

What you want is a design that isn’t distracted by having to accommodate clip-on components.  The most cost-effective and highest performance APs will always be purpose built.


While a clip-on necktie is merely a fashion faux pas, a clip-on AP is going to leave a lasting impression on your network and your checkbook.


Still don’t believe me? See how shattered glass and helicopters give a fun take on the argument in this video



Nov 05, 2013 07:47 AM

Hi JYL, The Miercom report contains more detail. It's posted online and available for you to review Miercom is well connected within Cisco and validated with them that the 3600 was properly configured with the latest publically available release at the time.


The Veriwave multiclient test is decidedly more simple because it's not over the air - it simply connects to the antennas and simulates multiple clients downloading encrypted UDP. The goal here was just to get AP performance capacity...kind of like dyno'ing a car.

Nov 03, 2013 11:25 PM

To prove that the testing was really done on the real lab can you share the conifg so people on this community can validate what was testd? I think everybody can claimed that their product is superior but without the real config and define the condition it was tested its all the same to me.


If we believed that our product is superior we should put all the config on how it was tested.




Oct 06, 2013 07:31 PM

Sorry, I didn't see these numbers you guys are claiming.

Oct 04, 2013 05:19 PM

You know...all vendors do this. However, we remind our customers why purpose built is a superior mechanical design to modular. I like to think of Aruba as the Daily Show or Colbert Report of the wireless industry. Cut through all the BS to deliver real facts. We prove it with test reports and full configs with both vendors.

Oct 03, 2013 11:48 AM

WiFivomFranMan, Thank you for your comments and I can certainly appreciate your dislike of product comparisons (I change the channel when the iPad vs Microsoft Surface commercials come on). The objective was to present new performance testing data for the AP 225 and draw parallels to Cisco’s approach.  You’d be surprised how often we get asked for this information and the testing is a critical part of our standard qualification practices anyways.


Also we absolutely welcome your recommendations on improving the administrative experience! “Brutal Intellectual Honesty” is what we try to live by at Aruba. I encourage you to open a discussion in the forums:

Oct 03, 2013 09:25 AM

It is bad that Aruba trash talks Cisco so much.  Why can you just make great products?   You sling gear by trash talking.  You speed test use clients that aren’t even released.  What kind of crap is that?  “if you had this client (which you can’t get) you would get these great speeds.. way better than Cisco!!!”    Come on, get real. 


I put the 220 and the 3600ac to a head to head with real clients.  You know what?  They both sucked.  I couldn’t get more than 290mbps out of the Aruba.  An Apple Airport out performed both the Aruba and the Cisco. 


While I am here, I’ll also say the Aruba GUI is horrible.  The CLI is a real good copy of Cisco (I wish the Cisco WLC was as good of a copy).  However a simple config comes to 800 lines.  Good grief…  Hide the defaults or improve how to you show run with some sub commands.  You can’t hand the keys over to a NOC because supporting it (GUI) requires an engineer.


You guys should give up the trash talk and work on making your products not suck.