CCNP Wireless CUWSS 642-732 Summary Notes 21 : Infrastructure requirements Part 2

802.11n APs with 100% duty cycle:

• the 2,4 band when oversubscribed with clients communicating at 144 Mbps may consume up to 70 – 75 Mbps on wired side.
• the 5 GHz band when oversubscribed and using 40 MHz channels with clients communicating at 300 Mbps may consume up to 160 – 170 Mbps on wired side.
• Both bands on the Ethernet Side will consume (75 + 170) = 245 Mbps per AP

Planning:

• Limit number of APs per WLC to make sure that overall traffic is less than port bandwidth
• use a 20:1 oversubscription ( 20 APs to 1 Controller)
• At RF level, the distance between the APs should be such that if one goes down, RRM can increase power of surrounding APs and cover the hole.
• Controller redundancy should be in place so that if one controller goes down, APs can join another. Redundancy models:
  • N+1:
    • one controller backs up n controllers
    • Only 1 additional controller
    • Cheap redundancy
    • only 1 WLC can fail at a time. the redundant controller could become oversubscribed with access points if there are multiple primary controller failures,
    • When a controller reaches the maximum number of joined access points, it accepts no more Lightweight Access Point Protocol (LWAPP) join requests.
    • Each access point is configured with a primary controller and all access points point to the next single redundant controller as secondary.
  • N+N:
    • N controllers back up N controllers
    • Each WLC loaded to half its license capacity
    • More expensive model
    • If network fails, all APs  move to the other WLC
  • N+N+1:
    • N controllers back up N controllers as secondary, and
      one controller backs up all N controllers as tertiary.
    • Intermediate but more common model
    • In this design, it is important to load-balance the access point capacity across both controllers. WLC are loaded to a variable perecntage of their licence (eg 66%).
    • to logically group access points on controllers to minimize inter-controller roaming events.
    • Additional backup is available in NOC
    • When one or several WLCs fail, the APs can join the backup or other controllers in the network
    • There should be enough access point and bandwidth capacity on each controller to handle a failover situation.
    • the plus one controller at the NOC or data center to no longer have to be a member of the same mobility group. It can now be referenced by its IP address versus being part of the same mobility group.
    • Most redundant but most expensive
• AP deployment Model
  • Standard deployment (not advised)
    • All APs connect to same controller
  • Salt and Paper Model
    • Neighbouring APs connect to different controllers
    • In one WLC gets disconnected, only half have to fail over
    • Better redundancy than standard model
    • Downside: when all controllers are available, interAP roaming leads to intercontroller roaming leading to added delays when roaming
    • Recommended only if applicatiosn are not time sensitive
    • Determine the roaming paths during the walkthrough visit. deploy APs along the path so that roaming in that path occurs on the same controller
    • If WLCs do not connect to same switched network, their WLAN – VLAN mapping may be different which means that we will have local to foreign roaming that increase wired bandwidth consumption
• Controller location
  • Distributed model
    • WLCs in each building close to the APs.
    • Negatively affects roaming efficiency. Intercontroller roaming even represents 20 ms delay when both controllers are in same subnet and 30 ms when in different subnets
  • Centralized model
    • All WLCs in same location
    • Capacity management is simpler
    • Simpler to create high availability controller topology
    • All packets must be sent to the controller

References:

1. CCNP Wireless  (642-732 CUWSS) Quick Reference Guide by Jerome Henry
2. IPexpert’s CCNP CUWSS Wireless Voice on Demand (642-731)
3. CUWSS Student Guide 1.0