Over the last year I’ve installed or updated about a dozen networks to use AC technology. Some have worked great from the beginning and others have been a misery and have never quite worked right. It took a while but I think I’ve got a solution.
All the networks that have had issues have had a few things in common. They were multi AP networks (using roaming with one single SSID) and used mostly Apple iOS clients, but not exclusively.
Over the last six months the issues have become worse, in the offices where they existed. I’ve researched and tested a lot of possible solutions including hardware (Cisco WAP371, Ubiquiti UAP-AC, Asus RT-56U and open-mesh.com), settings and learned more then I ever thought I’d need to know about how roaming works.
The main symptom is this: client devices refuse to roam to a stronger AP even when a stronger AP in the same group is available. This can cause poor performance since a user might walk from one floor to another and their device will keep using the AP that it was associated to when it was on a different floor.
So far, the issue seems like a bread and butter issue for a wifi admin. The smart people behind 802.11 have even created protocols that are supposed to prevent this exact problem.
- 802.11k helps your device measure the signal of other APs in the area without disconnecting from the current AP.
- 802.11r helps your device roam quickly from one AP to another. This feature is sometimes called “fast roaming”.
- 802.11v helps your device determine which AP, of the useful APs, has the least load on it and is therefore the best option.
The problem we in the small business wifi world have, is that only 802.11r is implemented in small business wifi hardware. And even when it is, the point at which a fast roam is triggered, isn’t something we can control. I’ve tried finding the signal level that common devices use to start looking for a better AP but I haven’t found a good source for this data yet.
So all we have to work with is 802.11r, fast roaming… Ok fine. This means that we need to setup access points so that their signals overlap but drop quickly when a device moves closer to a teamed AP. And this is where everything goes to pot.
The convential wisdom is that 5.0GHz wifi is faster then 2.4 but passes through material poorly when compared to 2.4. In practice I’ve found that the signal slows down dramatically but it is still usable and not low enough to cause a roam to start. So since a 802.11r fast roam does not get triggered, performance suffers.
And here is where I learned why there is a such a thing as a “power output” control to lower the strength of the signal. Using this setting might allow you to match the reach of the signal with the designed coverage area. At one site this week, lowering the signal on three existing AP’s to 12% and adding two AP’s (also at 12%) solved the issue.
And one last bit of hard earned knowledge. SMB access points might not give you control of output power or only give you rough settings like high and low. Especially when you are reusing existing wiring, you might not be able to build a single SSID network with good performance and reliable roaming without a Ruckus, Aruba or Cisco’s Aironet controller. This is what I found at another site this week. In this case, the bleed between floors was so bad that I could not lower the output power enough to prevent bleed through to the floors above and below and still have enough power to penetrate multiple plaster walls on the same floor. So I gave up and setup three SSIDs. The users will have to remember to switch when they move between floors but at least the network will be reliable.
If you have anything to add, I’d love to hear it.
Recent Comments