Dynamic Frequency Selection of channel assignment for APs in high density areas

  • 1
  • Question
  • Updated 3 years ago
We have Ruckus ZF7263's at our hotel. Our vendor installed the APs with the auto configuration for channel settings. After doing a wireless survey I noticed that there were several adjacent APs assigned to the same channel. It seems from this observation that the auto assignment of channels doesn't always do a good job of assigning the channels. Is there a reason why I would have 6 adjacent APs would be assigned to the same channel? It seems that Dynamic Frequency Selection isn't the ideal way to setup APs in high density locations. Is it your recommendation to manually configure the channels instead of relying on DFS?
Photo of Gunnar Amundson

Gunnar Amundson

  • 6 Posts
  • 0 Reply Likes

Posted 3 years ago

  • 1
Photo of John D

John D, AlphaDog

  • 497 Posts
  • 136 Reply Likes
Do you mean DFS or ChannelFly or Background Scanning? DFS isn't really a channel selection algorithm — it's a method for using channels with radar restrictions.

In my experience, it's not unusual even with ChannelFly to have adjacent AP's using similar channels if there's not a lot of network load — each AP observes good capacity on the channel. However, once an AP starts heavily using the channel, the others sharing the same channel will observe the sudden decrease in capacity and switch away from the channel.
Photo of Gunnar Amundson

Gunnar Amundson

  • 6 Posts
  • 0 Reply Likes
I'm talking about the dynamic channel assignment (channels 1-11) for the 2.4GHz frequency. So that would be DCA, not DFS. You are correct, DFS detects radar interference on the 5GHz frequency.

What I'm seeing at a lot of our hotels is the vendors are using the auto setting for channel assignment. Then on my surveys I see the technology doesn't really work that well as I'm see adjacent APs assigned to the same channel instead of a non-overlapping channels. Perhaps the auto setting is good for some environments, but not all. I'm getting mixed results, especially in high density AP deployments.
Photo of John D

John D, AlphaDog

  • 497 Posts
  • 136 Reply Likes
Ok, so it sounds like you're using ChannelFly to dynamically assign channels. ChannelFly is a statistical analysis algorithm based off actual observations of throughput on your network, not based off conventional heuristics (such as assign adjacent AP's non-overlapping channels).

In my opinion while it has its limitations and caveats, it does tend to be quite effective, especially in the real world where rogue AP's and other external circumstances generate 2.4GHz interference outside your control.

I'd recommend dumping the support info for your AP's to inspect the historical channel capacity statistics, to see if you agree with ChannelFly's observed throughputs on other channels. It might not be as disadvantageous as you might think for adjacent AP's to use the same channel, if the load is light and the other channels have even worse throughput.

The ChannelFly algorithm always chooses the channel with the highest observed throughput potential.
Photo of Monnat Systems

Monnat Systems, AlphaDog

  • 717 Posts
  • 151 Reply Likes
Ideal way to setup APs in high density locations is to use channelfly algorithms however using this features has its own caveats as this feature runs real time and make real quick decisions when it sees a another channel with a significant performance increase. this triggers a wave of client disconnects. As per my experience, Legacy devices get stuck, computer need to rebooted to bring WNIC to life or WNIC need to disabled or enabled. this happened to me, users slow WNIC can't see SSID as AP is channel hopping using channelfly

take a middle path where you increase the MTBC of channelfly feature where "you" dicate how often you want AP to change channel rather that random which creates more trouble than solving.

PS: i get quite a few support call for AP running channel fly in non-high density environments.
Photo of Gunnar Amundson

Gunnar Amundson

  • 6 Posts
  • 0 Reply Likes
Thanks for your comment. Very helpful!
Photo of John D

John D, AlphaDog

  • 497 Posts
  • 136 Reply Likes
Using ChannelFly in run-stop mode is a decent middle ground too and worth thinking about as an alternative to high MTBC. If you have sensitive clients, high MTBC will just delay the inevitable complaints from clients about wifi bouncing or dropping out due to clients not keeping up with channel switching.

I have found that low MTBC combined with 120 minute run-stop is a good combo.
Photo of Gunnar Amundson

Gunnar Amundson

  • 6 Posts
  • 0 Reply Likes
From reading these comments it seems to me that the advantages of Channelfly would be:
1. simplicity of configuration
2. Dynamic channel assignment for when there is a change in the environment where a channel change would benefit to avoid interference
3. to select the best channel for better throughput

The reason I posted here was because I saw the results of Channelfly assignments on our APs were contrary to everything I read and learned about channel assignments causing interference between adjacent APs.

At out hotel, the biggest problem is with supporting hand held wireless devices that by default have poor receptivity and are prone to interference and poor performance. Signal strength is critical to the performance of these devices. To enhance the signal strength I am considering deploying more APs. In our 5 story 177 room hotel we currently have 22 ZF7363 APs. Our guests that use smart phones are having issues with getting their connections dropped and receiving poor bandwidth performance. Other devices like laptops are not having these problems. Throughput is important, but in our hotel, reliable connectivity and signal strength are more important.

It seems to me that by manually assigning the channels to avoid interference of the signals on overlapping channels would benefit these hand held devices better than automatically assigning the channels for better throughput. I'm also wondering how much throughput are we looking at? A significant amount or just a little better? And is that amount of throughput really benefiting the hand held devices with connectivity issues.
Photo of John D

John D, AlphaDog

  • 497 Posts
  • 136 Reply Likes
The general problem with handheld devices is that they are very sensitive to the signal being weak, mainly because the user's grip can attenuate signals, the user tends to move around a lot which makes bad roaming decisions really devastating, and they are generally 1 stream so they'll see 1/2 or 1/3 of peak throughput anyway.

You may find that having more AP's is helpful if signal strength is an issue, or SmartRoam if you observe smartphones are not connecting to the most nearby AP. IMO both of these will benefit you more than channel assignments.

Especially in an environment like a hotel, dynamic noise from your guests bringing in rogue devices can often make conventional channel selection guidelines suboptimal. You can see from support info what ChannelFly measures every channel's capacity to be. The difference sometimes is huge.
Photo of Gunnar Amundson

Gunnar Amundson

  • 6 Posts
  • 0 Reply Likes
I can see the advantage of using Channelfly, but what advantages or conditions are there for configuring the channels manually?

Here is a comment about the results of having common channel assignments on adjacent APs.

http://www.wireless-nets.com/resource...

"setting access points near each other to the same channel will degrade capacity, and performance will be significantly lower as traffic on the network increases. As a result, it’s best to assign non-overlapping channels to nearby access points."

When I set the channel assignment to auto I noticed several adjacent APs assigned to the same channel. If the above statement is true, how could the automatic assignments to the same channels result in better performance or connectivity? Wouldn't those 6 APs signals have an effect on each others signals as they would with any other AP?

I'm thinking that Channelfly may be useful under some, but not all environments. I think the same may be true for manual assignments too. Even though one channel setting may have little impact, I'm looking at all aspects of a wireless network. Number of APs, location/placement of APs, channel assignment, and any other configuration that will improve our guest network.

I get comments about slowness, disconnections, connectivity problems to name a few. My tests do not show these results, but I haven't yet tested during the peak hours between 10PM and midnight. I do my tests using multiple devices iPhone, iPad, laptops (OSX and Windows), I have done numerous speed tests, wireless survey, heat map, etc. So even though channel assignments seem less significant, I have to look at every aspect to narrow down these issues.
Photo of John D

John D, AlphaDog

  • 497 Posts
  • 136 Reply Likes
The problem with using the manual 1, 6, 11 setup is that it's the same way everyone thinks, and those might turn out to be the most crowded channel. Also, suppose I'm running a baby monitor or microwave oven that spews non-802.11 noise down near channels 1-4. If that's the case, using channel 1 is a horrible idea and will encounter nothing but interference. In fact, in a clean environment, if you deploy 3 Ruckus's with ChannelFly next to each other, you'll find that they will settle on picking channels 1, 6, and 11. Just in the real world, you'll find that some channels simply are no good.

Setting access points near each other to the same channel probably has few issues if they are all lightly used. If they are heavily used, the "losing" AP(s) will notice a sudden decline in channel capacity (due to the winning AP emitting a lot of traffic on that channel) and they will switch away.

The bottom line is, ChannelFly's problem is not capacity or not being able to avoid interference — quite the opposite — that's the strong suit. ChannelFly is the equivalent of you switching channels and doing speed tests for all 11 channels (and more on 5GHz) every few hours and choosing the best channel.

The weakness with ChannelFly is that some clients react badly to channel switches, and either disconnect from the wifi temporarily or even stay disconnected and assume the wifi network is down.

Are you using standalone AP's or a ZoneDirector? You should be able to have logs from your clients that correlate reports of connectivity problems with what the clients are actually doing (e.g. is there a disconnection right after a ChannelFly switch? Are clients disconnecting regardless of ChannelFly? Are clients picking distant AP's instead of nearby ones?)

The right way to approach this problem is by looking at the logs and possibly asking clients for their MAC address or host name for when they are having trouble.
Photo of Gunnar Amundson

Gunnar Amundson

  • 6 Posts
  • 0 Reply Likes
We have a ZoneDirector. I will reference the logs, do some tests with and without ChannelFly, Thank you for your help.