The Tyranny of the Controller

Michael Vizard

Intelligence, much like information, wants to be distributed as much as possible. So when it comes to wireless networks, the question is, why is so much intelligence consolidated inside the wireless controller?

We know that controlling the flow of information is power. In the case of the wireless controller, it also means creating an artificial requirement to buy an additional piece of network hardware.

Devin Akin, chief Wi-Fi architect for Aerohive, argues that the intelligence required to manage a distributed network should be distributed to the access points, thereby effectively eliminating the need for the controller. Aerohive has been quietly making that argument for the past several years. But as wireless networks become a mainstay of corporate networks, chief technologists should start asking themselves what kind of wireless network architecture they want.

Key considerations in the process should be how much throughput is ultimately going to go over this wireless network and, based on that number, does the controller represent a probable traffic bottleneck? The other thing to consider is how much network bandwidth intelligence is going to reside on the next generation of clients. If the clients are smarter about signalling their bandwidth requirements, doesn't that mean the access point needs to be smarter about serving those requests?

Controllers were a necessary evil when wireless networking started out. But increasingly, it's looking like they have outlived their usefulness unless, of course, you happen to be the company that makes them.

 



Add Comment      Leave a comment on this blog post

Apr 8, 2010 2:04 PM Anonymous Anonymous  says:
Good article I can make strong arguments for both intelligence at the AP as well as centralized control. THere are two models Dumb low end AP's with centralized congestion control, this works well when your end users are fixed in place. Not so well with roaming users, but it can still be done. Smart AP's with distributed congestion control as you mention. The cost factor must be considered and you also can get locked in to one vendor. There is also often a need for centralized congestion control for the aggregate pipe. Art Reisman CTO www.netequalizer.com Reply
Apr 8, 2010 10:04 PM Joe E Joe E  says:
Your article is spot-on in that there are two models for managing wireless LANs. The distributed architecture referenced in the article has actually been around for some time (since before the Aironet days). The controller-based model was developed to address the mobility and authentication issues brought on by this model, and ended up subsuming it. Having a controller absolutely does not force you to route the data through it; APs can send traffic locally just as the controller-less model can, and flexibly as well. However, the primary issues plaguing enterprise wireless networks today are overall stability and performance, not controller overload. IT is tasked with providing a stable and predictable experience for the glut of high-value applications that are being sent across the WLAN, such as video and VoIP. The real question IT should be asking is not whether or not a controller is required; they should be looking at the WLAN solution that is in place and determining if it can provide the density and seamless mobility needed to run these apps successfully over the WLAN. Architectures based on microcell technology are just not equipped to provide this level of performance. Joe E Meru Networks Reply
Apr 22, 2010 6:20 AM Devin Akin Devin Akin  says: in response to Joe E
I agree with some of what you're saying big Joe, but local forwarding of data at the AP without applying security and QoS policy (as a minimum) is ill advised. "Distributed forwarding" in no wise equals "controller-less" for many reasons, including application of stateful security policy to uplink and downlink traffic flows through the AP. We've done a number of high-density tests, across a wide variety of vendors' 11n gear, including Meru's, and in multi-AP deployments with a proper site survey, multiple channel architecture always handles more users due to having more channel space and more cumulative airtime to work with. Perhaps Meru had an advantage with high-density handling at one point in time, but that advantage has disappeared over the last year due to several vendors (not just Aerohive) adding Airtime Fairness mechanisms, and several protocol & channel optimization features (e.g. bandsteering). Physics says that if you optimize channel use and have more spectrum to work with, you can get more work done. I think you'll agree that if you build it, they'll come...in droves...and they'll bring every application known to man (and some that aren't yet known). Now is the beginning of the decline of controllers, which is strongly indicated in watching the market's #2 player recently push both forwarding and firewall features to their APs due to controller overload (per their own documents) and the market's #3 player push both firewalling and QoS into their APs along with local forwarding. Devinator Reply

Post a comment

 

 

 

 


(Maximum characters: 1200). You have 1200 characters left.

 

 

Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.


 

Resource centers

Business Intelligence

Business performance information for strategic and operational decision-making

SOA

SOA uses interoperable services grouped around business processes to ease data integration

Data Warehousing

Data warehousing helps companies make sense of their operational data