5

I have a wifi deployment with about 60 access points, currently limited to channels 1,6, and 11. This is a college campus, and the residence halls can get quite noisy, from an rf perspective. We typically have about 30 devices per access point, but the kind of traffic the college students expect to use — hd streaming video/netflix, low-latency gaming, and even some p2p file sharing (though we have severe traffic shaping for that at the gateway) — really eats up air time on the access points, and adding more density or pushing users to the 5Ghz space is proving difficult.

I'm considering breaking with tradition, and using either channels 1,4,7, and 11, channels 1,5,8, and 11, or channels 1,4,8,and 11. The idea is to carefully place it so a channel 1 AP is never near a channel 4/5 AP, a channel 4/5 AP is never near a channel 7/8 AP, and a channel 7/8 AP is never near a channel 11 AP. The extra distance should help reduce the chance of collisions on the over-lapping channels. If done correctly and in conjunction with a few additional access points, it would result in a theoretical 33% increase in available airspace.

Has anyone else tried this? Did it work well? If so, which set of channels did you go with?

Joel Coel
  • 12,910
  • 13
  • 61
  • 99
  • 1
    Probably no chance of moving to 5Ghz? :) – jscott Aug 29 '11 at 22:18
  • Is this 802.11g or n? – hookenz Aug 29 '11 at 22:44
  • What issue are you trying to solve that has led you to think of this? – hookenz Aug 29 '11 at 22:47
  • Is the problem reception? Provided you layout the 4 access point channels with some nice distance between competing channels you will be fine. I can think of no reason why you couldn't, make sure your not stealing too many channels from your neighbors though :) – Silverfire Aug 29 '11 at 23:52
  • The problem is contention/throughput. People get 3 to 5 bars (they can see and connect), but can't do much once connected because of the amount of rf noise from too many clients talking at once... It's the kind of traffic college students expect to use: all the hd streaming video, low-latency gaming, and even some file sharing that really eats up airtime more than just simple web browsing. – Joel Coel Aug 30 '11 at 15:09
  • Are these all individual access points or are you using a controller? – SpacemanSpiff Sep 01 '11 at 00:14
  • @spaceman We actually have 63 individually managed access points. At $100 plus no controller costs and using existing infrastructure, the whole system ran $6000 to cover the entire campus. There are some obvious drawbacks (less reporting, having to manually tweak issues like this) but it really works pretty well, for about 1/10 what a controller-based system would cost. – Joel Coel Sep 01 '11 at 00:51
  • Kudos to you for doing it on a budget. The reason I asked is that a nice feature of the controller based solutions is transmit power balancing automatically. – SpacemanSpiff Sep 01 '11 at 01:07

2 Answers2

1

You hsould not try to overlap channels. You will destroy rates for both overlapped sides.

Instead, you should try to reduce power at the access points to limit their reach (and thus the area where they occupy airspace). Then you can add additional APs in the halls. The goal is to make sure that each AP only handles a manageable amount of devices, i.e. two-digit numbers when using good APs.

It requires some amount of testing how the reach of the APs is. Preferably, you should do this during actual realtime load as people dampen the signals and thus influence the reach of the APs.

Holger Just
  • 3,315
  • 1
  • 16
  • 23
  • I understand reducing power on the APs, but that doesn't help when all the clients are still talking at full volume. I was hoping for something that addressed client noise as well. – Joel Coel Aug 29 '11 at 22:04
  • Well, clients are supposed to lower their sending power as well. Actually, it is the only way to manage airspace. Also, if you'd assume that clients send at full throttle, your idea of having ovberlapping channels would make even less sense as they would consume even more actual airtime, on the two channels that is. – Holger Just Aug 29 '11 at 22:08
1

I can't quite see how what you're proposing will work. Maybe I haven't quite understood it properly. I know that channels 1, 6 & 11 are the best channels to use because they don't overlap each other. While say channel 4 will use some of the frequency range of channels 1 & 6 etc. There is also a channel 14... don't use it. It's a narrower frequency that is limited to about 2Mbit/s or something. I did try it once and couldn't get it to perform.

I can only assume that you're having performance issues which is why you're looking at doing this right?

Well, having been down this track on a much smaller scale I implemented QoS and it made an enormous difference. I followed the hints from a guy on the net nicknamed toastman. He has large deployments in appartment buildings and hotels that he manages, and QoS has made what he has set up perform really well.

Realise this, that through QoS you can force your clients to be less chatty as their internet chat is slowed by the QoS giving other users a chance. Also, Bittorrent is a big cause of internet crawl. His QoS settings have helped here. Even though throughput for an individual is slowed, overall performance is increased for everyone and certain protocols which operate in bursts will still feel snappy.

Following his tips has really made a huge difference to my own set up, and even my home network.

Have a look at and read this: http://www.linksysinfo.org/index.php?threads/using-qos-tutorial-and-discussion.28349/

If in fact there are some interference problems, this link is very interesting. http://www.techedbackstage.net/2009/07/15/diagnosing-and-resolving-extremely-high-rf-utilisation/

hookenz
  • 14,132
  • 22
  • 86
  • 142