802.11ec: Collision Avoidance without Control Messagesem9/publications/802_11ec-slides.pdf · 802.11ec: Collision Avoidance without Control Messages Eugenio Magistretti, Omer Gurewitz,

802.11ec: Collision Avoidance withoutControl Messages

Eugenio Magistretti, Omer Gurewitz,and Edward Knightly

Rice UniversityBen Gurion University

802.11ec stands for 802.11 with Encoded Control and does not represent an IEEE standard

Ideal Wi-Fi

tota

lth

roug

hput

CLI

QU

E

100%

Low Power & Obstructions

100%

HT

Collisions

tota

lth

roug

hput

CLI

QU

E

100%

HT

Collisions

tota

lth

roug

hput

CLI

QU

ETwo hidden terminals70% throughput loss

Ex.


100%

tota

lth

roug

hput

CLI

QU

E

HT


RTS/CTS

RTS

/CTSCollisio

ns

100%

Overhead

RTS

/CTS

tota

lth

roug

hput

CLI

QU

E

HT

Collision Penalty Overhead Penalty

Collisio

ns

100%

Overhead

RTS

/CTSCollisio

ns

tota

lth

roug

hput

CLI

QU

E

HT

RTS/CTS/ACK overhead~ max size packet

@ 54 Mbps

Ex.

Collision Penalty Overhead Penalty

Objective

Fundamentally Fundamentally re-think the way controlre-think the way controlinformation is conveyedinformation is conveyed in order to in order toguarantee guarantee low overheadlow overhead and and robustnessrobustness

Control Messages CSS

ACK

DATARTS

CTSLong Duration

Messages

- Preamble- Header- Base Rate Data


• Correlatable Symbol Sequences (CSS) are pre-defined pseudo-noise (PN) bit sequences detected via cross-correlation

ACK

RTS

CTS

Messages

Correlatable SymbolSequences


Long Duration

DATA

DATA


ACK

RTS

CTS

Messages



Long Duration

ShorterMore Robust

✔✔

- Preamble- Header

DATA

DATA

• Correlatable Symbol Sequences (CSS)

• Control Information via CSS

• 11ec Protocol Illustration

• Experimental Results

Outline

Why Correlatable Symbol Sequences?

CSS’s key idea: Signal detection is more robust than decoding



• Sender and receiver agree upon a set of pre-defined CSS

• The receiver can detect a specific CSS u(k) bycorrelating it with the incoming samples y(k)



• Sender and receiver agree upon a set of pre-defined CSS


)()()(1

0

* !+=! "#

kykuCL

!

C(") # L 0)( !"C

CSS NON MATCHINGCSS MATCHING



• Sender and receiver agree upon a set of pre-defined CSS– CSS length L determines the max size of the CSS set


)()()(1

0

* !+=! "#

kykuCL

!

C(") # L 0)( !"C

CSS NON MATCHINGCSS MATCHING

CSS Intuition

CSS u(k)

CSS+INTERFERENCE as received by R

Information K

• Node S wants to convey control information (“Information K”) to node Rvia a pre-defined CSS u(k)

CSS Intuition

CSS u(k)

CSS+INTERFERENCE as received by R

Noise Signal

• Node S wants to convey control information (“Information K”) to node R via a pre-defined CSS u(k)

CSS Intuition

MATCHING at R

Information KReceived

• The correlator spikes when thepre-defined CSS is received

||


CSS Intuition

MATCHING at R

Information KReceived

NON MATCHING(at any other node)

||


CSS Intuition

MATCHING at R NON MATCHING(at any other node)

||


Information KReceived ???||

CSS Summary

• AdvantagesLow Overhead

– No preamble– No encoding

High Robustness– To low SINR– To collisions

• Correlatable Symbol Sequences (CSS) are pre-definedPN bit sequences detected via cross-correlation

• CSS do not need any preamble/header

Length L limits number of pre-defined CSS





Outline


ACK

DATARTS

CTS

Messages


Long Duration

ShorterMore Robust

✔✔

Limited Number

DATA

Control Information via CSS

• 802.11 control message structure

Preamble Fixed ControlField Type DurationAddresses


• Define a small dictionary of CSS to representthe message information content


Control Information via CSS - Dictionary



– Type Limited Set


Map to different CSS(e.g., RTS CTS ACK)





– Address



One CSS per nodeLarge Set butLocally Limited





– Address

– Duration Large Set



New CSS types + TimingCode (e.g., channel free)

Large Set butLocally Limited


One CSS per node


• CSS Information Scope Control


Control Information via CSS - Scope



– Public

Preamble Fixed ControlField

To be received by all nodes(e.g., Channel Reservation/Release)

Type DurationAddresses

All nodes detect the CSS(i.e., possess the correlator)




– Public

– Unique Feature: Private

Preamble Fixed ControlField

To be received by all nodes(e.g., Channel Reservation/Release)

To be received by theother endpoint (e.g., ACK,source/destination address)

Type DurationAddresses

All nodes detect the CSS(i.e., possess the correlator)

Selected nodes detect the CSS






Outline

• 11ec follows the fundamental concepts of 802.11

11ec Protocol Illustration

neighbor

sender

receiver

hidden terminal HT


s

r

neighbor


hiddenterminal


Private CSS conveying thereceiver address

Ic(r) InitiationIc(r)

s

r


neighbor

hiddenterminal


Private CSS conveying thereceiver address

Ic(r) InitiationIc(r)

s

r


Every Receiver isassociated with an Ic(r)

neighbor

hiddenterminal


Public CSS

Ic(r) Initiation

Rc Channel Reservation

Ic(r)

s

r

Rc


Rc is a unique CSS

neighbor

hiddenterminal

BUSY CHANNEL

TIMEOUT


Ic(r)

s

r

Rc

DATA

Ic(r) Initiation


BUSY CHANNEL


DATA contains 802.11preamble P

Data Reception

TIMEOUT

neighbor

hiddenterminal

P


Ic(r)

s

r

Rc

DATA

Ic(r) Initiation



P

Neighbor nodes detect Pand NAVs

Data Reception

BUSYneighbor

hiddenterminal

BUSY CHANNEL

TIMEOUT


Ic(r)

s

r

Rc

DATA

Ic(r) Initiation



P

Neighbor nodes detect Pand NAVs

Data Reception

BUSY

802.11ec can co-exist with 802.11

neighbor

hiddenterminal

BUSY CHANNEL


Private CSS conveying thesender address

Ac(s)

Ic(r)

s

r

Rc

DATA

Ic(r) Initiation


Ac(s) Acknowledgement


P

Data Reception

BUSYneighbor

hiddenterminal

BUSY CHANNEL

TIMEOUT


Private CSS conveying thesender address

Ac(s)

Ic(r)

s

r

Rc

DATA

Ic(r) Initiation




Every Sender isassociated with an Ac(s)

P

Data Reception

BUSYneighbor

hiddenterminal

BUSY CHANNEL

TIMEOUT

Ac(s)


Ic(r)

s

r


Ic(r) Initiation

Rc Channel ReservationRc

DATA


Fc Channel ReleasePublic CSS

Fc

BUSY CHANNEL

P

BUSY

Data Reception

Fc is a unique CSS

neighbor

hiddenterminal

Experimental Results Roadmap

• WARP

- CSS Robustness

• Channel Measurement Emulation

- Tput/Fairness/Airtime Utilization

- Fundamental and Small Topologies

• Large Network Simulations

- 20-node Random Topology

CSS Implementation

1. Repeatable and controllable experiments

Sender Receiver Interferer

LogicalConnection

PhysicalCabling

CSS Performance Metric

CSS Performance Metric

• Threshold SelectionProbability of false positives ~10-8

PerformanceProbability of false negatives

CSS Robustness

• CSS are correlatable even at low SINR

Robustness

L=63

L=127

L=255

L=511

[dB]

Robustness and Overhead Gains

Duration: Message Control vsCoded Control

11b/g 11a 11ec~1000 µs ~184 µs ~41 µs

10 dB more robust 78% less overhead

Robustness

RTS Decoding

(BPSK 1/2)

Ic(r) Detection

(127-symbol)

[dB]

CSS Implementation

1. Repeatable and controllable experiments

Sender Receiver Interferer

LogicalConnection

PhysicalCabling

2. Emulation based on Realistic Channel Measurements

Hidden Terminals with Heterogeneous Rates

54 Mbps 6 Mbps


54 Mbps 6 Mbps

• Strong link RTS captures over weak link’s

• 802.11ec data collisions are rare

THROUGHPUT

802.11ec

RTS/CTS

802.11

THROUGHPUT

802.11ec

RTS/CTS

802.11


54 Mbps 6 Mbps

• Strong link RTS captures over weak link’s

• 802.11ec data collisions are rare

Jain Index 0.98 0.52 0.78fairest

FAIRNESS

THROUGHPUT

802.11ec

RTS/CTS

802.11

AIRTIME UTILIZATION

RTS/

CTS

802.

11802.

11ec


54 Mbps 6 Mbps

Tota

l Airt

ime

Util

izat

ion

• Differently from 802.11, 802.11ecdoes not penalize weak and lowdata-rate links

802.11ec increases airtime utilization

• 802.11ec increases fairness AND channel utilization

AIRTIME UTILIZATION

802.

11ec

RTS/

CTS

802.

11


54 Mbps 6 Mbps

Tota

l Airt

ime

Util

izat

ion

THROUGHPUT

802.11ec

RTS/CTS

802.11

Jain Index 0.98 0.52 0.78

20-Node Random Topologies

• Simulations

Frac

tion

of L

inks

Throughput [Kbps]

802.11ec

RTS/CTS802.11

• CDF of throughput distribution


• Simulations

802.11ec vsRTS/CTS

0.2 7x

0.4 4x

0.6 1.9x

GAINS

• By reducing collisions, 802.11ec improves the weak links

Throughput [Kbps]

802.11ec

RTS/CTS802.11

• CDF of throughput distribution

Frac

tion

of L

inks


• Simulations

802.11ec vsRTS/CTS

0.2 7x

0.4 4x

0.6 1.9x

GAINS

• By reducing collisions, 802.11ec improves the weak links

Throughput [Mbps]

Frac

tion

of L

inks• CDF of throughput distribution

Throughput [Kbps]

802.11ec

RTS/CTS802.11

Frac

tion

of L

inks

Summary and Conclusions

• Objective: Fundamentally Objective: Fundamentally re-thinkre-think the way controlthe way controlinformation is conveyedinformation is conveyed in order to guarantee in order to guarantee lowlowoverheadoverhead and and robustnessrobustness

•• CSSCSS’’s s have short duration and improve robustnesshave short duration and improve robustness

•• 802.11 802.11ecec uses uses CSSCSS’’s s to convey control informationto convey control information

—— Small CSS Dictionary Small CSS Dictionary

—— Scope Control Scope Control

•• 802.11802.11ecec improves fairness while also increasing channel improves fairness while also increasing channel utilizationutilization

—— Ex. 3x fairness, 1.5x airtime utilization, up to 12x throughputEx. 3x fairness, 1.5x airtime utilization, up to 12x throughput

802.11ec: Collision Avoidance withoutControl Messages

Eugenio Magistretti, Omer Gurewitz,and Edward Knightly

Rice UniversityBen Gurion University

Documents

802.11ec: Collision Avoidance without Control Messagesem9/publications/802_11ec-slides.pdf · 802.11ec: Collision Avoidance without Control Messages Eugenio Magistretti, Omer Gurewitz,