Installation guide

84
Wi-Fi Location-Based Services—Design and Deployment Considerations
OL-11612-01
Deployment Best Practices
considered normal for the test environment (an internal WLAN development facility where there are
many access points and clients being tested independently of one another), such an incredibly high level
of rogue activity is certainly not what one would expect in a routine business environment.
Notwithstanding the unusually high level of rogue traffic, Appendix C—Large Site Traffic Analysis,
page 116 clearly shows why configuring too short a polling interval would not be recommended via
overburdened or slow WAN links. Enabling polling for only asset tag and wireless clients shows that the
amount of traffic being transmitted back to the location appliance from each controller per polling cycle
is on the order of ~300,000 bytes. Polling for statistical information as well in this test environment
would add ~125,000 bytes from each controller per polling cycle. This would amount to approximately
425,000 bytes of traffic from each controller to the location appliance per polling cycle.
Discounting the high amount of rogue traffic detected by a factor of 75 percent, it is reasonable to
anticipate that polling for more typical volumes of rogue information would add between 100,000 and
200,000 bytes per controller per polling cycle for a total traffic volume of approximately
500,000–600,000 bytes per controller per polling cycle. With reasonable polling intervals on
modern-day high speed WAN links and campus LANs, more than ample bandwidth would be available
to handle this amount of polling traffic in addition to servicing other traffic.
Traffic Between the Location Appliance and WCS
Although the location appliance and the WCS are in routine communication with one another, peak
traffic flows tend to occur during the synchronization (Location > Location Servers > Synchronize) and
backup/restore processes. Traffic during network design synchronization can also be a concern if the
WCS and the location appliance are separated by slow or congested WAN links. Delays resulting from
such congested links may result in very long synchronization times or in extreme cases, the inability to
propagate updated network designs and calibration models between WCS and the location appliance.
Thus, whenever possible the location appliance and WCS should be co-located on a high-speed LAN.
Keep in mind as well that unlike polling, the synchronization process can be initiated either on-demand
or scheduled as a routinely occurring event. Thus, for example, synchronization can be scheduled to
occur during an off-peak period on a daily basis. This allows the administrator of the LBS system to
better take advantage of periods where there is a lull in traffic from other network users.
During the network design synchronization process, the more up-to-date partner (either WCS or the
location appliance) shares updated network design and calibration model information with the other
partner. In the typical case of WCS possessing a more recent version of a network design or calibration
model, WCS issues commands to the location appliance via SOAP/XML to verify the network designs
and calibration models it contains. After WCS confirms its knowledge of what is contained within the
location appliance, it issues the appropriate commands via the API interface to update it.
Lab testing was performed for a relatively simple configuration consisting of a single campus network
design (two buildings, single floor each, 14 access points total) and four calibration models (two site
calibrations plus the two simple default calibration models). TCP traffic between WCS and the location
appliance was observed during network synchronization of an updated network design and two
calibration maps. This analysis indicated 840 packets being transmitted from the WCS to the location
appliance totaling 790,144 bytes and 879 packets transmitted from the location appliance to WCS
totaling 895,367 bytes for a grand total of 1719 packets and 1,685,511 bytes.
In larger location-aware deployments, it is logical to expect the traffic levels between WCS and the
location appliance to increase because of the potential for an increased number of buildings, floors, and
access points present per floor. Other factors that increase the amount of data exchanged during a
synchronization are the presence of any walls or other obstacles defined within network designs, or if
any coverage areas are defined via the Map Editor. All these constitute additional information that is
contained within the network design that is exchanged between partners during synchronization.