ieee 802.11ax: wireless networking in high-density...
TRANSCRIPT
IEEE 802.11ax: Wireless Networking in High-density WLANs
Boris [email protected]
March 8, 2017Wireless NetworkingGroup Webinarhttp://wnrg.upf.edu/
2
About this Webinar
● About IEEE 802.11ax details, there are many excellent white papers / tutorials / slides that can be found just browsing the web
● In this Webinar, we will focus only in the following features:
– Dynamic Spectrum Access / Channel bonding
– MU transmissions
– Dynamic Sensitivity Control / BSS coloring / SR Opportunities
● These slides can be found at http://wnrg.upf.edu/
Will IEEE 802.11ax improve the user experience in dense WLAN deployments?
3
About this Webinar
High-density WLANs
Dynamic Channel Bonding
MU transmissions
Dynamic Sensitivity
Control & BSS coloring
Introduction
Final remarks
4
WLANs
● Broadband Internet access (few Gbit/s)
● ISM bands (2.4, 5 GHz)
● Decentralized Management
● Chaotic & Dense deployments
● Interference limited performance
● CSMA/CA
5
Internet
DATA
t
AP
A
B
SUCC. TX
ACK
DATA
ACK
0
SUCC. TX COLLISION
DATA
0
0
07
77
remaining backoff
14 11 8
slotted backoff countdown
DCF = CSMA/CA (slotted backoff) + Stop & Wait ARQ
6
High-density WLAN deployments
Bellalta, Boris. "IEEE 802.11 ax: High-efficiency WLANs." IEEE Wireless Communications 23.1 (2016): 38-46.
7
More traffic, and more heterogeneous
● Previsions show a clear increase in mobile / wireless traffic
– High-definition video traffic will become dominant
● The network must offer high and constant transmission rates● Traffic / content aware mechanisms, including traffic differentiation
– IoT traffic
● Many small and miscellaneous devices● High aggregate traffic
Flexible Resource Allocation
8
Next PHY/MAC amendment: IEEE 802.11ax-2019
Next steps in the WLAN evolution, < 6GHz
● Improve performance & spectrum efficiency
– High-order modulations, channel coding, spatial multiplexing
– Wider channels ~ channel bonding
– Multi-user transmissions (MU-MIMO, OFDMA)
– Interference cancellation (including full duplex)*
● Improve spatial reuse
– Dynamic channel selection
– Transmit power control*
– Dynamic sensitivity control*, BSS coloring + Spatial Reuse Opportunities
* Not likely to be included in 11ax
9
IEEE 802.11ax
Feature IEEE 802.11ac IEEE 802.11ax
Frequency bands [GHz] 5 GHz 2.4 and 5 GHz
Channel widths [MHz] 20, 40, 80, 160, 80+80 The same
OFDM 64 tones / 20 MHz 256 tones / 20 MHz
OFDM symbol duration 4 μs (GI = 0.8 μs) 16 μs (GI = 3.2 μs)
Modulations BPSK, QPSK, 16-QAM, 64-QAM, 256-QAM
Adds 1024-QAM
Forward Error Coding Convolutional, LDPC optional LDPC mandatory
MU transmissions DL MU-MIMO DL & UL MU-MIMO (in RUs > 106 subcarriers)DL & UL OFDMA
OFDMA (RU sizes, subcarriers)
- 26,52,106,242,484,996,2x996
MU-MIMO stations Up to 4 Up to 8
10
Comments
● OFDMA, and UL-MU-MIMO are the most disruptive features included in 11ax compared to previous high-throughput amendments (IEEE 802.11n/ac)
– From single, or few packets transmitted at each channel access, to many
11
About this Webinar
High-density WLANs
Dynamic Channel Bonding
MU transmissions
Dynamic Sensitivity
Control & BSS coloring
Introduction
Final remarks
12
Dense WLAN deployments
● AP density (APs/m2)
– 1 AP every 25 m2
● STA density
– 1 STA every m2
– In the figure, 1 STA every 6.25 m2
– Higher if we talk about IoT devices
● Observations
– Each color represents a different channel
– Presence of walls, objects, people, and other “obstacles” should be taken into account
– The figure is optimistic, as the radius of the carrier sense range can be several times larger
5 m
5 m
carrier sense range
What are the most common patterns in terms of overlapping
WLANs?
13
Collisions
● Collisions with nodes inside the carrier sense range
– Two or more nodes finish their backoff at the same time
● Unnecessary channel contention pauses due exposed nodes
● Collisions with nodes outside the carrier sense range (i.e., hidden nodes)
5 m
5 m
A
14
Exposed Nodes
● Collisions with nodes inside the carrier sense range
● Unnecessary channel contention pauses due exposed nodes
– Transmission A prevents
transmissions B, C, D
– Area throughput reduction
● Collisions with nodes outside the carrier sense range (i.e., hidden nodes)
5 m
5 m
A
B
C
D
15
Hidden Nodes
● Collisions with nodes inside the carrier sense range
● Unnecessary channel contention pauses due exposed nodes
● Collisions with nodes outside the carrier sense range (i.e., hidden nodes)
– Transmission A suffers interference from transmissions B and C
– SNR for A seems high enough to not be much affected, though…
– Does the RTS/CTS mechanism work? 5 m
5 m
A
C
B
16
Comments
● OFDMA, and UL-MU-MIMO are the most disruptive features included in 11ax compared to previous high-throughput amendments (IEEE 802.11n/ac)
– From single, or few packets transmitted at each channel access, to many
● High-density WLANs stress some coexistence issues, but not everything is negative, and there is room to find innovative solutions
17
Improving performance / efficiency
Activity periods
Contention periods
Improving the efficiency means lower temporal spectrum occupancy, so, we also improve coexistence with other
networks, reduce collisions, exposed and hidden node problems
t
t
WLAN 1
WLAN 2
WLAN 1
WLAN 2
18
About this Webinar
High-density WLANs
Dynamic Channel Bonding
MU transmissions
Dynamic Sensitivity
Control & BSS coloring
Introduction
Final remarks
19
Wider Channels
● Allows for higher transmission rates
● Motivation: Wider channel, more data subcarriers
● Drawbacks: Increases contention with neighboring WLANs
RU Width11ax Data
Subcarriers
Rate [Mbps](64-QAM, ¾, 16us,
1 SU-SS)
Gain(vs 20 MHz)
20 234 65.812 1
40 468 131.625 2
80 980 275.625 4.1880
160 1960 551.250 8.3761
20
Wider Channels
5 m
5 m
P
● Dynamic Channel bonding➔ Maximize channel width➔ Minimize contention
160 MHz
20 MHz
40 MHz
80 MHz
160 MHz
f
21
Dynamic Channel Bonding
22
Wider Channels
Faridi, Azadeh, Boris Bellalta, and Alessandro Checco. "Analysis of Dynamic Channel Bonding in Dense Networks of WLANs." IEEE Transactions on Mobile Computing (2016).
23
Comments
● OFDMA, and UL-MU-MIMO are the most disruptive changes in 11ax from previous high-throughput amendments
– From single, or few packets transmitted at each channel access, to many
● High-density WLANs stress some coexistence issues, but not everything is negative, and there is room to find innovative solutions
● The use of wider channels only improves the spectrum utilization and allows for higher transmission rates in low-density WLAN scenarios
– Channel width adaptation
24
About this Webinar
High-density WLANs
Dynamic Channel Bonding
MU transmissions
Dynamic Sensitivity
Control & BSS coloring
Introduction
Final remarks
25
MU transmissions
STA A
STA B
STA CSTA D
STA A
STA B
STA C STA D
Downlink MU Uplink MU
● Downlink and Uplink
● OFDMA and MU-MIMO
● Centralized and Decentralized
● New frames
– Trigger, MU-RTS, MU-ACK
● Required info at the AP:
– Channel Reports
– Buffer State Reports
26
MU transmissions
t
fIndep. transmissions
a) MU uplink b) MU downlink c) MU downlink
OFDMARU
MU-MIMO(RU > 106 subcarriers)
SU-MIMO
● RUs = {26,52,106,242,484,996,2x996} subcarriers● Up to 8 MU-MIMO beams● Up to min(Map,Msta) SU-MIMO streams
W
d) SU uplink
P
Scheduling?
padding
27
DL MU transmission
W
RU1
RU2
RU3
STA A
STA BSTA C STA DSS
t
MU-RTS
CTS BACK
STA A
STA B
STA C STA D
DL-MU-PPDU
AIFS
TDL-MU
AP
STA A
STA B
STA C
STA D
SIFS
UL MU transmission
28
t
RTS
CTS BACK
PPDU
AIFS
TDL-SU
AP
STA A
STA B
STA C
STA D
RTS
CTS BACK
PPDU
TDL-SU
4 TDL-SU > TDL-MU ?
W
RU1
STA ASS
W
RU1
STA DSS
… vs 4 SU transmissions
29
MU
SU
30
4 TDL-SU > TDL-MU ?
TDL-SU = 0.9415 ms, → 4 TDL-SU = 3.7660 ms
TDL-MU = 0.97350 ms
What about throughput?
MU-DL: 49.307 Mbps
SU-DL: 12.746 Mbps
31
Map
= 8 antennas
Msta
= 4 antennas
W = 160 MHz
Bellalta, Boris, and Katarzyna Kosek-Szott. "AP-initiated Multi-User Transmissions in IEEE 802.11 ax WLANs." arXiv preprint arXiv:1702.05397 (2017).
32
Channel Sounding
~Number of stations
t
Trigger
AIFS
TCS
AP
STA A
STA B
STA C
STA D
1/λcs
CS reports
Trigger
CS reports
NDPA NDP
t
33
Channel Sounding
~Number of stations
t
Trigger
AIFS
TCS
AP
STA A
STA B
STA C
STA D
1/λcs
CS reports
Trigger
CS reports
NDPA NDP
t
Bellalta, Boris, and Katarzyna Kosek-Szott. "AP-initiated Multi-User Transmissions in IEEE 802.11 ax WLANs." arXiv preprint arXiv:1702.05397 (2017).
34
UL MU transmission, AP-initiated
W
RU1
RU2
RU3
STA A
STA BSTA C STA DSS
t
MU-RTS
CTS
MU-ACK
STA A
STA B
STA CSTA D
UL-MU-PPDU
AIFS
TUL-MU
AP
STA A
STA B
STA C
STA D
Trigger
35
UL MU transmission
● We may have collisions between UL MU transmissions and UL SU transmissions
– STAs included in UL MU transmissions may also start a transmission at the same time
● Prioritize AP transmissions (for instance using an special EDCA Access Category)
Bellalta, Boris, and Katarzyna Kosek-Szott. "AP-initiated Multi-User Transmissions in IEEE 802.11 ax WLANs." arXiv preprint arXiv:1702.05397 (2017).
36
Buffer State Reports
● To schedule UL MU transmissions, the AP requires to know the buffer state of the stations
● 11ax supports two approaches:
– Solicited BSR: each station explicitly delivers its BSRs in any frame sent to the AP as a response to a BSR Poll send by the AP.
– Unsolicited BSR: user stations implicitly report its BSRs in the QoS Control field of any frame sent to the AP.
~Number of STAs
1/λbs
Tbs t
37
MU transmissions in 11ax, vs 11ac?
Bellalta, Boris, and Katarzyna Kosek-Szott. "AP-initiated Multi-User Transmissions in IEEE 802.11 ax WLANs." arXiv preprint arXiv:1702.05397 (2017).
38
Comments
● OFDMA, and UL-MU-MIMO are the most disruptive changes in 11ax from previous high-throughput amendments
– From single, or few packets transmitted at each channel access, to many
● High-density WLANs require some new considerations for coexistence, but not everything is negative, and there is room for innovation
● The use of wider channels only improves the spectrum utilization and allows for higher transmission rates in low-density WLAN scenarios
– Channel width adaptation
● MU transmissions allow for reducing overheads, though new ones are required: Is there any optimal tradeoff?
● 802.11ax may only outperform 802.11ac in scenarios with many STAs assoc. to a single AP
39
About this Webinar
High-density WLANs
Dynamic Channel Bonding
MU transmissions
Dynamic Sensitivity
Control & BSS coloring
Introduction
Final remarks
40
Exposed Nodes
● Collisions with nodes inside the carrier sense range
● Unnecessary channel contention pauses due to exposed nodes
– Transmission A prevents
transmissions B, C, D
– Area throughput reduction
● Collisions with nodes outside the carrier sense range (i.e., hidden nodes)
5 m
5 m
A
B
C
D
41Bellalta, Boris. "Throughput Analysis in High Density WLANs." IEEE Communications Letters (2016).
42
Dynamic Sensitivity Control
● Goal: reduce the number of exposed nodes
– STAs will transmit more often → Higher throughput
● How: Increasing the CSth
● Drawback: the number of hidden nodes may also increase
5 m
5 m
A
B
C
D
43
Dynamic Sensitivity Control
● STA A is transmitting to AP A
● STA B detects the channel as busy as it has CSth = -82 dBm, and the received power is -67 dBm
● STA B defers any transmission attempt for the duration of the on-going transmission
-35 dBm
- 75 dBm
AP A
STA A
AP B
STA B
- 67 dBm
CSth = -82 dBm
44
Dynamic Sensitivity Control
-35 dBm
- 67 dBm
- 85 dBm
- 91 dBm
- 58 dBm
- 43 dBm
SNR=56 dB
SNR=42 dB
AP A
STA A
AP B
STA B
Adjusting CSth
we are
effectivelydoubling the throughput
CSth = -62 dBm
● If STA B CSth = -62 dBm, it would have detected the channel as idle, and it would have initiated a transmissions to AP B
● SNR at both sides seems high enough
45
Dynamic Sensitivity Control
t
t
Beacon (Ptx = 20 dBm)
-45 dBm -38 dBm -51 dBm -47 dBm
Initial measuring period
1) The STA calculates E[RSSI], and2) Updates CS
th = max(CS
th,min, min(CS
th,max,E[RSSI] + Margin))
Continuous CSth adaptation Initial measuring period
window
CSth
update CSth
update CSth
update
Avoid HN from stations in the same WLAN
Guarantees an extra exclusion area
20 dBm
E[RSSI]
CSth,max
CSth
Margin(dB)
CSth,min
46
Dynamic Sensitivity Control
Afaqui, M. Shahwaiz, Eduard Garcia-Villegas, Elena Lopez-Aguilera, Graham Smith, and Daniel Camps. "Evaluation of dynamic sensitivity control algorithm for IEEE 802.11 ax." In Wireless Communications and Networking Conference (WCNC), 2015 IEEE, pp. 1060-1065. IEEE, 2015.
● Using only DSC, gains of 10 % in throughput are obtained
● Gains of 20 % when combined with channel allocation and rate control
● Higher number of hidden nodes may prevent further gains
47
BSS coloring
t
ACK
ACK
DATA
DATA
Tx detected
Frommy
BSS?
yes no
Defer
Ignore
RSSI>CS-
OBSS’
yes no
Defer
● The CCA’ can be set following the previous DSC algorithm● We avoid Hidden Node problems between STAs of the same WLAN
48
Spatial Reuse Opportunities
● Identify Spatial Reuse (SR) opportunities by acquiring knowledge from OBSSs
– By using the BSS color
● Instead of CSth, adapt the transmission power of the “secondary” user to minimize any harmful interference to the “primary” user
● The duration of the SR opportunity is lower than the duration of the transmission from the primary user
t
t
OBSS A
BSS
OBSS A Tx
Duration of the “primary” transmission
tOBSS B
OBSS B Tx
Duration of the “primary” transmission
Duration of the “primary” transmission
49
Comments
● OFDMA, and UL-MU-MIMO are the most disruptive changes in 11ax from previous high-throughput amendments
– From single, or few packets transmitted at each channel access, to many
● High-density WLANs require some new considerations for coexistence, but not everything is negative, and there is room for innovation
● The use of wider channels improves the spectrum utilization and allows for higher transmission rates in low-density WLAN scenarios
– In high-density WLAN scenarios, the extra channel contention is harmful
● MU transmissions allow for reducing overheads, though new ones are required: Is there any optimal tradeoff?
● 802.11ax may only outperform 802.11ac in scenarios with many STAs assoc. to a single AP
● DSC, BSS coloring and Opportunistic SR seem very promising approaches to reduce the exposed node problem in high-density WLANs, and hence, improve the area throughput
– First results show performance gains far below (my) expectations. Where is the problem?
– DSC may not be included in 11ax, but maybe in pre-11ax solutions
50
About this Webinar
High-density WLANs
Dynamic Channel Bonding
MU transmissions
Dynamic Sensitivity
Control & BSS coloring
Introduction
Final remarks
51
Upcoming 802.11 amendments
52
11ad/11ay
11ax
5G
● IEEE 802.11aq: Pre-association Service Discovery
– Providing more information about the services offered by that WLAN
● IEEE 802.11ai: Fast and lightweight hand-off
– Hand-off delay < 100 ms
– Reduction of control & management frames
● IEEE 802.11ad/ay: Short-range, very high-throughput communications at 60 GHz (mmW)
– Highly directional links
– Multi Gbits / second
Multiple radio interfaces● Software Defined Radios● 5G, 11ax, 11ad/ay, others● Multi-interface networking (MPTCP)● Dynamic network selection based on QoE requirements
53
11ad/11ay
11ax
5G
LTE
Coexistence between Random-access and Scheduled-access technologies in ISM bands
Is time for WLANs to move to a (more) scheduled operation?
54
Comments
● OFDMA, and UL-MU-MIMO are the most disruptive changes in 11ax from previous high-throughput amendments
– From single, or few packets transmitted at each channel access, to many
● High-density WLANs require some new considerations for coexistence, but not everything is negative, and there is room for innovation
● The use of wider channels improves the spectrum utilization and allows for higher transmission rates in low-density WLAN scenarios
– In high-density WLAN scenarios, the extra channel contention is harmful
● MU transmissions allow for reducing overheads, though new ones are required: Is there any optimal tradeoff?
● 802.11ax may only outperform 802.11ac in scenarios with many STAs assoc. to a single AP
● DSC, BSS coloring and Opportunistic SR seem very promising approaches to reduce the exposed node problem in high-density WLANs, and hence, improve the area throughput
– First results show performance gains far below (my) expectations. Where is the problem?
– DSC may not be included in 11ax, but maybe in pre-11ax solutions
Future WLAN scenarios will integrate 11ax & 11ad/ay, and many other 802.11 features
55
Will IEEE 802.11ax improve the user experience in dense WLAN deployments?
56
Our research
● Understanding the dynamics and fundamental performance limits of adaptive systems
– an ultimate goal is to understand and characterize formally the implications and effects of including self-adaptive capabilities in the performance of wireless networks
● In order to operate as close as possible to optimal conditions in highly-dynamic scenarios, decisions about next actions/configurations can be made at the network edge or in the cloud
– Combining efficiently "short-term" local decisions based on partial information with much more accurate but "mid/long term" decisions based on a global view of the network is still an unsolved challenge
– Machine Learning and Information Extraction techniques from collected datasets
● Strategies, policies and protocols for maximizing the use of the spectrum resources in all dimensions (space, frequency and time)
– interference cancellation/full-duplex communications, NOMA, dynamic spectrum access, transmit power control, sensitivity adaptation and directional transmissions