1. Eduard Garcia-Villegas
Dept. of Network Engineering
eduardg@entel.upc.edu
Just some common sense rules
put together in a nice set of
colorful slides
EFFICIENT Wi-Fi
deployments
The basics
Where’s that
fu*#}ng Wi-Fi ?!?
Download this presentation (PDF) from:
http://ocw.upc.edu/sites/default/files/materials/15016145/efficientwi-fideployments-5709.pdf
2. Contents
EFFICIENT Wi-Fi deployments
o Big vs. small
o Analyze requirements
o #STAs and #Needed radios
o Available channels
o Reuse factor
o Dimensioning cells
o Optimization
Efficient Wi-Fi deployments 2
by Podere Casanova
3. Wi-Fi deployments: intro
Efficient Wi-Fi deployments 3
In the era of ubiquitous Internet…
Wireless internet access can be a traumatic
experience due to
o Many concurrent users (dense scenarios)
o Coexistence (older/slower devices, other
technologies sharing the band, etc.)
o …
o POOR DESIGN
4. Wi-Fi deployments: big vs. small (1)
Efficient Wi-Fi deployments 4
Coverage-driven design
o In the past: maximize cell size && minimize costs
• Optimize AP location and increase cell size less APs
needed (lower cost)
• Problems:
– more devices per AP (lower per STA throughput)
» Reduced efficiency due to higher collision probability
5 STAs x AP <2 STAs x AP
= STA
VS.
= AP
5. Wi-Fi deployments: big vs. small (2)
Efficient Wi-Fi deployments 5
= STA
= AP
LAME!
Coverage-driven design
o In the past: maximize cell size && minimize costs
• Optimize AP location and increase cell size less APs
needed (lower cost)
• Problems:
– Longer distances AP STA mean worse signal quality and,
hence, more robust (slower) PHY rates are used
» Capacity of the whole cell is reduced
» Longer tx time more power consumed and more collisions
6. Wi-Fi deployments: big vs. small (3)
Efficient Wi-Fi deployments 6
= STA
= AP
HIDDEN NODES!
CAN’T REACH ITS AP!
Coverage-driven design
o In the past: maximize cell size && minimize costs
• Optimize AP location and increase cell size less APs
needed (lower cost)
• Problems:
– More hidden nodes more collisions
– Power mismatch: AP (high tx power) and STA (low tx power)
» STA can hear the AP, but the AP can't hear the STA
» If you want a big cell, increase the antenna gain, not the tx power!
7. Wi-Fi deployments: big vs. small (4)
Efficient Wi-Fi deployments 7
Coverage-driven design
o In the past: maximize cell size && minimize costs
• Optimize AP location and increase cell size less APs
needed (lower cost)
• It has problems in present (dense) deployments.
Other key aspects
o KPI requirements
o Client and AP capabilities
• Are modern ≥ 11n capable (how many antennas)?
Coexistence with 11a/b/g? Dual band?
o Propagation phenomena
• Outdoor/indoor? APs mounted on ceiling, walls or floor?
o User density
8. Efficient Wi-Fi deployments
The basics
Analyze requirements
ANALYZE REQUIREMENTS
Per user Total
#STAs per
RADIO
#RADIOS
NEEDED
AVAILABLE
CHANNELS
REUSE
FACTOR
DIMENSION
CELLS
OPTIMIZE/
TROUBLESHOOT
9. Wi-Fi deployments: requirements
Efficient Wi-Fi deployments 9
The first thing is to identify key performance indicators (KPI)
o Minimum bandwidth required to satisfy supported
applications
o Maximum latency tolerated
o Expected Min-Avg-Max number of active devices
Examples (per-user requirements):
o School
• BW: <3Mbps (video
streaming; desktop/file sharing)
• Delay tolerance: low
(video streaming; intranet login)
• Users: Min-Avg-Max =
up to 30 per classroom
o Convention center (1500 att.)
• BW: <1 Mbps (web browsing;
e-mail)
• Delay tolerance: Medium
• Users: “educated guess”
– 70% will connect Wi-Fi device
– 50% simultaneously
– 1500 x 0.70 x 0.5 = 525
10. Wi-Fi deployments: #STAs and radios
Efficient Wi-Fi deployments: the basics 10
Capacity-driven design (rule of thumb)
o Example 1 school (<3Mbps x 30 users per classroom):
• 20 STAs per AP each classroom served by two radios (two
APs or one dual band AP)
– Assume homogeneous (IT-controlled) 11n 2x2 devices
– Good signal quality (high rates available) STAs achieve
~80Mbps of net throughput (isolated)
– Allow future growth: AP utilization ≤ 75%
75/(100*3Mbps/80Mbps) = 20 STAs per AP
o Example 2 convention center (<1Mbps x 525 users)
• 32 STAs per AP 525/32 = 16 – 17 radios
– Assume heterogeneous (BYOD) devices
– Diverse signal quality STAs achieve ~40Mbps of net
throughput
– AP utilization ≤ 80% 80/(100*1Mbps/40Mbps) = 32 STAs/AP
11. ANALYZE REQUIREMENTS
Per user Total
#STAs per
RADIO
#RADIOS
NEEDED
AVAILABLE
CHANNELS
REUSE
FACTOR
DIMENSION
CELLS
OPTIMIZE/
TROUBLESHOOT
Efficient Wi-Fi deployments
The basics
Available channels
12. Wi-Fi deployments: channels (1)
Efficient Wi-Fi deployments: the basics 12
Capacity limited by the scarcity of available spectrum
o 2.4GHz ISM band
• Only three non-overlapping channels (1,6,11)
• Four (almost) non-overlapping channels (1,5,9,13) where available
Baseline capacity
Three channel scheme:
Baseline x3
Ch1
Ch11
Ch6
Three channel scheme:
Baseline x3.05
Ch1
Ch11
Ch6
Ch11
Four channel scheme:
Baseline x3.9
Ch1
Ch5
Ch9
Ch13
13. Wi-Fi deployments: channels (2)
Efficient Wi-Fi deployments: the basics 13
Capacity limited by the scarcity of available spectrum
o 2.4GHz ISM band
• Only three non-overlapping channels (1,6,11)
• Four (almost) non-overlapping channels (1,5,9,13) where available
– Not available in all regulatory domains (e.g. North Americas)
– Many devices default to Americas config. will see coverage
gaps in the areas served by APs in Ch13.
• Highly congested: coexistence with WPANs, cordless phones,
baby monitors, microwave ovens…
o 5GHz ISM band
• 15-21 non overlapping channels in different sub bands
• Highly variable from one regulatory domain to another
– Some channels only for indoor use, others require DFS
– Different tx power limits …
14. ANALYZE REQUIREMENTS
Per user Total
#STAs per
RADIO
#RADIOS
NEEDED
AVAILABLE
CHANNELS
REUSE
FACTOR
DIMENSION
CELLS
OPTIMIZE/
TROUBLESHOOT
Efficient Wi-Fi deployments
The basics
Reuse Factor
15. Wi-Fi deployments: reuse factor
Efficient Wi-Fi deployments: the basics 15
#Radios Needed
Reuse Factor =
Available Channels
o If Reuse Factor ≤ 1 LUCKY YOU!
o Otherwise, each channel is shared among Reuse
Factor APs INTERFERENCE!
• Minimize interference by.
– Carefully dimensioning cells
– Smart channel management
16. ANALYZE REQUIREMENTS
Per user Total
#STAs per
RADIO
#RADIOS
NEEDED
AVAILABLE
CHANNELS
REUSE
FACTOR
DIMENSION
CELLS
OPTIMIZE/
TROUBLESHOOT
Efficient Wi-Fi deployments
The basics
Dimension the cell
17. Wi-Fi deployments: dimension cells (1)
Efficient Wi-Fi deployments: the basics 17
What is the cell radius?
o Max distance at which frames
can be decoded
• Pt is tx power
– Decreases with MCS (to avoid distortion)
• Sr is receiver sensitivity
– Increases with MCS
– Rr reception range
– d is the distance tx rx
– α is the path loss exponent
o Different radius depending on targeted
MCS
𝑷 𝒓 ≈
𝑷 𝒕
𝒅 𝜶
⟶ 𝑹 𝒓 ≈
𝑷 𝒕
𝑺 𝒓
𝟏
𝜶
VERY FAST
SLOW
R1 R2 Rn
18. Wi-Fi deployments: dimension cells (2)
Efficient Wi-Fi deployments: the basics 18
How to set cell radius for Wi-Fi small cells?
o Reduce AP’s tx power
• Reduces interference over
other cells
• Avoids AP/STA power
mismatch
• Reduces suitable rates
NOT SO
FAST
SLOW
19. Wi-Fi deployments: dimension cells (3)
Efficient Wi-Fi deployments: the basics 19
How to set cell radius for Wi-Fi small cells?
o Reduce AP’s tx power
• Reduces interference over
other cells
• Avoids AP/STA power
mismatch
• Reduces suitable rates
o Increase min tx rate of the cell
• Reduces performance anomaly
and allows higher average rate
– Avoid “sticky” STAs
• Possible unsupported devices
– Accept, at least, 802.11b@11Mbps?
OUT!
NOT SO
FAST
20. Wi-Fi deployments: dimension cells (4)
Efficient Wi-Fi deployments: the basics 20
BUT…interference goes beyond the cell edge
o Carrier Sense Range (Rc)
• Max distance at which frame
preamble can be detected and,
hence, prevent concurrent
transmissions in the same channel.
– Only 3dB SNR is enough! (>200m
outdoors)
– Behavior improved in IEEE
802.11ax
o Beyond Carrier Sense Range
• Transmitted frames are just noise
LEAVE ME
ALONE!
21. Wi-Fi deployments: dimension cells (5)
Efficient Wi-Fi deployments: the basics 21
Coverage strategy for maximal densification
o Reduce reuse distance
• Low gain directional antennas
• AP placement
– Overhead: AP installed on the ceiling/lamp posts facing down
– Side: AP installed on walls/pillars
– Floor: under floor/under seat (stadiums or auditoriums)
– Even consider mounting APs behind walls/obstacles and avoid LoS
(enriches multipath diversity leveraged by MIMO)
120º
vs.
60º
coverage
reuse
Ch1 Ch1Ch1Ch1
reduced
reuse distance
22. ANALYZE REQUIREMENTS
Per user Total
#STAs per
RADIO
#RADIOS
NEEDED
AVAILABLE
CHANNELS
REUSE
FACTOR
DIMENSION
CELLS
OPTIMIZE/
TROUBLESHOOT
Efficient Wi-Fi deployments
The basics
Finishing touches
23. Wi-Fi deployments: channel plan (1)
Efficient Wi-Fi deployments: the basics 23
Ch11
Ch11
Ch6
Ch1
Ch6
Ch11
Ch1
Ch6
Ch11
Ch1
In your
dreams
Reality(t)
Dynamic and unpredictable spectrum utilization
o License-free bands!
Intelligent channel assignments are required
24. Wi-Fi deployments: channel plan (2)
Efficient Wi-Fi deployments: the basics 24
NO INTERFERENCE!
Automatic and dynamic channel assignments aimed at
reducing interference maximizing performance
o APs gather information of the environment
• Number of APs detected
• Power received from neighboring APs
• Portion of time the channel was reported busy/idle by CCA
Ch. X is free!
Ch. X is free!
25. Wi-Fi deployments: channel plan (3)
Efficient Wi-Fi deployments: the basics 25
Automatic and dynamic channel assignments aimed at
reducing interference maximizing performance
o APs gather information of the environment
• Number of APs detected
• Power received from neighboring APs
• Portion of time the channel was reported busy/idle by CCA
o Ideally, client STAs too (and report via IEEE 802.11k)
Ch. X is free!
Ch. X is free!
26. Wi-Fi deployments: channel plan (4)
Efficient Wi-Fi deployments: the basics 26
Automatic and dynamic channel assignments aimed at
reducing interference maximizing performance
o APs (ideally, STAs too) gather information of the environment
• Number of APs detected
• Power received from neighboring APs
• Portion of time the channel was reported busy/idle by CCA
o Distributed approach (autonomous APs)
• Each AP periodically (and asynchronously) scans the medium and
chooses the least congested channel local optimum
• Alternatively, APs collaborate (exchange information) to produce
better decisions
o Centralized approach (controller-based)
• APs send periodic reports to a controller
– Knowing the whole picture and having more resources (i.e. CPU,
memory, etc.) controller runs a sophisticated optimization algorithm
global optimum
28. Wi-Fi deployments: load balancing (1)
Efficient Wi-Fi deployments: the basics 28
Wi-Fi users are quasi-static and tend to concentrate in space
& time hot spots
o Clients (i.e. traffic) unevenly distributed among APs
• Some APs (channels) congested and some others underutilized
o Load Balancing techniques could increase ability to satisfy QoS
requirements
• Load Balancing techniques widely
used in cellular networks
• Take advantage of overlapping areas
between neighboring cells
– Clients can be served by several BSs
– System decides the best BS for a
client depending on BSs’ loads
• Not directly applicable to Wi-Fi WLANs
– Clients decide association and
roaming, not the network
BA C
D E F
29. Wi-Fi deployments: load balancing (2)
Efficient Wi-Fi deployments: the basics 29
Load balancing with client-driven association in WLANs
o Typically, client STAs decide best AP based on RSSI
measurements (i.e. strongest Beacon or Probe Response Frame)
• Uneven distribution of users uneven distribution of load
o Some APs broadcast load information (BSS Load element) and
some clients do care about it
o Network-oriented client-driven load balancing
• Band steering: encourage utilization of the 5GHz band
– If AP or controller detect a STA sending Probe Requests in the two bands
do not send responses through 2.4GHz radios, only through 5GHz
• Disassociation/blacklisting
– Network decides STA’s best AP the rest of APs ignore that STA
requests (if already associated, current AP sends Disassociation frame)
• Cell Breathing: adapt size of the cell
– Congested APs reduce tx power of Beacons and Probe Responses
underutilized APs do the opposite
30. Wi-Fi deployments: load balancing (3)
Efficient Wi-Fi deployments: the basics 30
Example of cell breathing
o Reduce power of Beacons and Probe Responses
• do not reduce power of data frames since this will reduce suitable
rates and increase error rate
Cell A Cell B
1
2
3
4
31. Efficient Wi-Fi deployments: the basics 31
Wi-Fi deployments: load balancing (3)
Example of cell breathing
o Reduce power of Beacons and Probe Responses
• do not reduce power of data frames since this will reduce suitable
rates and increase error rate
Cell A Cell B
1
2
4
3
32. Wi-Fi deployments: The End
Efficient Wi-Fi deployments: the basics 32
Don’t forget the wires!
o Data/power wires to APs
• If not…multihop or mesh-based
wireless distribution system
o Uplink pipe
• Imagine all this headache for just
a DSL WAN connection…
33. Some references (1)
Load balancing
o Garcia-Villegas, E.; Vidal, R.; Paradells, J. (2006, June). “Load Balancing in WLANs through
IEEE 802.11k Mechanisms,” in 11th IEEE Symposium on Computers and Communications,
ISCC 2006.
o Garcia-Villegas, E.; Vidal, R.; Paradells, J. (2008, July). “Cooperative Load Balancing in IEEE
802.11 Networks with Cell Breathing,” in 13th IEEE Symposium on Computers and
Communications, ISCC 2008.
o Garcia-Villegas, E.; Ferrer, JL.; Lopez-Aguilera, E; Vidal, R.; Paradells, J. (2009). “Client-
driven load balancing through association control in IEEE 802.11 WLANs”. European
Transactions on Telecommunications, ETT vol. 20, no. 5, pp. 494-507. John Wiley & Sons.
Sensitivity control
o Afaqui, MS.; Garcia-Villegas, E.; Lopez-Aguilera, E.; Smith, G.; Camps-Mur, D. (2015)
“Evaluation of Dynamic Sensitivity Control Algorithm for IEEE 802.11ax,” IEEE Wireless
Communications and Networking Conference, WCNC 2015, pp. 1072-1077
o Afaqui, MS.; Garcia-Villegas, E.; Lopez-Aguilera, E.; Camps-Mur, D. (2016) “Dynamic
Sensitivity Control Algorithm leveraging adaptive RTS/CTS for IEEE 802.11ax,” in IEEE
Wireless Communications and Networking Conference, WCNC 2016
o Afaqui, MS.; Garcia-Villegas, E.; Lopez-Aguilera, E.; Camps-Mur, D. (2016) “Dynamic
Sensitivity Control of Access Points for IEEE 802.11ax”, in IEEE International Conference on
Communications, ICC’16
Efficient Wi-Fi deployments: the basics 33
34. Some references (2)
Channel management
o Garcia-Villegas, E.; Vidal, R.; Paradells, J. (2009). “Frequency assignments in IEEE 802.11
WLANs with efficient spectrum sharing”. Wireless Communications and Mobile Computing,
WCMC vol. 9, no. 8, pp. 1125-1140. John Wiley & Sons
o Mengual, E.; Garcia-Villegas, E.; Vidal, R. (2013, September). “Channel management in a
campus-wide WLAN with partially overlapping channels,” in The 24th IEEE International
Symposium on Personal, Indoor and Mobile Radio Communications, PIMRC 2013
o Deek, L.; Garcia-Villegas, E.; Belding, E.; Lee, S-J.; Almeroth, K. (2011, December). “The
Impact of Channel Bonding on 802.11n Network Management,” in 7th International
Conference on emerging Networking EXperiments and Technologies, CoNEXT’11
o Deek, L.; Garcia-Villegas, E.; Belding, E.; Lee, S-J.; Almeroth, K. (2014). “Intelligent Channel
Bonding in 802.11n WLANs,” IEEE Transactions on Mobile Computing, vol. 13, no. 6, pp.
1242-1255
o Deek, L.; Garcia-Villegas, E.; Belding, E.; Lee, S-J.; Almeroth, K. (2013, June). “Joint Rate
and Channel Width Adaptation for 802.11 MIMO Wireless Networks,” in IEEE Conf. on
Sensing, Communication, and Networking, Secon’13, pp. 167-175 (Nominee for the Best
Paper Award)
o Deek, L.; Garcia-Villegas, E.; Belding, E.; Lee, S-J.; Almeroth, K. (2015). “A practical
framework for 802.11 MIMO rate adaptation,” Computer Networks, vol. 83, pp. 332-348
Efficient Wi-Fi deployments: the basics 34
35. EETAC - UPC
Master's degree in Applied Telecommunications
and Engineering Management
IoT & Ubiquitous IP
Course offered at:
Notes de l'éditeur
100*expected throughput / maximum possible throughput expected % utilization per STA
BYOD = bring your own device
NOTE THAT THOSE ASSUMPTIONS WORK ONLY FOR THIS EXAMPLE
Set of supported rates is announced in beacons
Sticky STAs: they keep association with current AP until completely out of reach (with worst MCS), even though there are better APs
Set of supported rates is announced in beacons
Explain that experiment run on the campus XSF-UPC (evident benefits when our channel management –considering partially overlapping channels – was in charge)