Roaming over Wi-Fi

…guest post from Ahmad Nassiri

A lot has been written and discussed about roaming but why it is hard to get the message across that ‘WLAN roaming is a decision made by a client device and not the wireless medium’?

Yes there are options available to help manipulate this decision but still, it is that individual client that makes the final decision!

In simple terms roaming is when a wireless client leaves one BSSID and authenticates/associates to a new BSSID. Many factors affect this transition, including client’s wireless card characteristics, receive sensitivity, environment, wireless APs placement, signal quality/availability, signal overlap, secondary coverage, type of application, authentication method and last but not least interference. Client devices monitor the condition of their current connection and as the quality of signal (RSSI) degrades for any reason, they start the probing phase to move to an AP with better RSSI. The exception is every client does this differently thus making things complicated.

Mobility or simply roaming is at the core of wireless technology. The ideal wireless network is to have a seamless roaming experience without any significant packet drops affecting the connection or worse disconnect completely. For time sensitive applications like voice & video roaming becomes critical while other applications may not be as much susceptible to roaming interruptions. Therefore it becomes very important to incorporate roaming requirements of different devices on the wireless network. Designing to the weakest client’s capabilities are always recommended.

I have come across designs where one wireless device’s characteristics considered initially, AP on stick survey used a different wireless device (usually the recommended wireless card by the survey software vendor!). And finally when the network goes into production, different varieties of wireless devices are associated to the network. That is when design shortcomings, incompatibility and environmental factors affecting roaming are identified. Unfortunately if the purpose of the design has been to provide coverage (green heat maps) in the first place, then expect problems!

A while back I undertook a design for a healthcare facility with multiple sites. The network was designed to the best practice principles. Results from the AP on Stick survey was as expected, the same laptops and tablets in-scope for the production network was used for testing. When the network went live in the pilot site all services including roaming was working fine. Staff using WOWs (Workstations on Wheels, formerly known as COWs Computer on Wheels) without any issues and moving them across the facility without any disruption.

The roll out for remaining sites took a while, and staff raised concern about roaming when using the WOWs. Reports included slow response and even disconnection when moving the units within the facility. Interestingly this issue was not experienced on the tablets using the same WLAN and application! Upon some basic analysis it was discovered the laptops were not the same model tested and deployed in the earlier sites. Client had made another deal with a different vendor in between and new laptops were deployed. These laptops did not roam as the other ones and sticking to the original APs for longer. The wireless card driver was updated to no avail, enabled ‘roaming aggressiveness’ and ‘power saving mode’ to Maximum Performance that is when we noticed roaming happening. While we did not enable any of these options in the original laptops! An assumption was made by the client that any laptop will work just fine but as we learned this was not the case in this scenario. Even though the receive sensitivity of both laptops were almost identical they roamed differently!

Every vendor determines their wireless chipset cell boundaries, receive sensitivity and RSSI cut-off range.

This makes it very important to obtain data sheets and study them thoroughly to avoid issues. It is also highly important and recommended to have an open communication channel with the client during the project life cycle to understand any deviation from original scope and its effects on deliverables and to avoid situations like above.

General Recommendations:

  • A wireless design undertaken by a professional wireless engineer is the key here, considering and incorporating end device capabilities, applications and environmental factor requirements
  • Design based on the less contentious 5.0GHz band requirements
  • Adjust a higher signal for 5.0GHz band enabling devices to prefer using the 5.0GHz signal over 2.4GHz
  • Allow for signal overlap between APs (vendor dependent)
  • Depending on the end devices requirements, target an RSSI of -65dbm to -67dbm
  • Optimal AP transmit power, channel selection, QoS settings and minimizing interference (CCI/ACI) helps for better roaming
  • Disable the support for 802.11b
  • Disable lower data rates, this along with optimal transmit power and channel selection will minimize the sticky client situations, where client tend to stick to one AP even though other APs with better RSSI are available. Sticky clients not only affect roaming but the entire wireless experience for other clients as well
  • With client devices capable of dual band support the 2.4GHz can be disabled, forcing clients to use the only available band (5.0GHz)
  • Wireless clients learn about available APs by passive or active scanning. To obtain a list of target APs in time is very important to make the roaming decision and becomes critical for clients that handle real-time applications like voice and video
  • One way of helping some clients achieve faster roaming is to configure the device to scan only the available/selected channels in the WLAN. But unfortunately this option is not available on all devices
  • The following IEEE standards supports roaming and are recommended to be enabled in the wireless core infrastructure, exception is if the client device supports these standards!
    • 802.11v Wireless Network Management
    • 802.11k Radio Resource Management
    • 802.11r Fast Basic Service Set (BSS) Transition (FT)
      With 802.11r enabled the client can pre-authenticate to a target AP before roaming and reduces the authentication time and minimizes interruption
  • Configuring client limit thresholds on each AP or aggressive load balancing is risky and must be enabled in a very controlled and robust network. While it helps distribute/load balance clients equally across the APs, it can affect a roaming client targeting to roam to a particular AP but fails due to the AP reaching its configured threshold or capacity
  • Similar to client limit threshold, some wireless vendors allow for tweaks to the RSSI thresholds. This forces clients to abide by the limits or get disconnected. Again this must be enabled carefully taking into consideration all other devices associating to the network
  • On the windows clients there are certain settings that can enhances the device roaming behavior, like ‘aggressive roaming’ and ‘power saving mode’. This has to be a very calculated change as WLANs prone to interference may not be a suitable fit to enable this feature
  • Apple recommends turning on 802.11k as it saves the client time to look in each channel.
  • While we are focusing on better roaming, what if we do not want to roam as often? Roaming too often is not a good sign and highlights some underlying issue. Interference is one of the factors that causes unwanted roaming. If the medium is congested the wireless device keeps on trying to roam to a better AP repeatedly causing disruptions
  • The best recommendation is to test the roaming behavior of the intended client devices in a POC. Sometimes even the product data sheets does not include enough information regarding roaming.

And finally I hope the wireless infrastructure makes roaming decisions one day, just as the cell towers in cellular industry does!