Synchronous Two-Phase Rate and Power Control in 802.11 WLANs - - PowerPoint PPT Presentation

synchronous two phase rate and power control in 802 11
SMART_READER_LITE
LIVE PREVIEW

Synchronous Two-Phase Rate and Power Control in 802.11 WLANs - - PowerPoint PPT Presentation

Synchronous Two-Phase Rate and Power Control in 802.11 WLANs Kishore Ramachandran, Ravi Kokku, Honghai Zhang, and Marco Gruteser WINLAB, Rutgers University and NEC Laboratories America Towards All-Wireless Enterprises Last-mile network access


slide-1
SLIDE 1

Synchronous Two-Phase Rate and Power Control in 802.11 WLANs

Kishore Ramachandran, Ravi Kokku, Honghai Zhang, and Marco Gruteser WINLAB, Rutgers University and NEC Laboratories America

slide-2
SLIDE 2

Towards All-Wireless Enterprises

  • Last-mile network access predominantly 802.11

High bandwidth, low/no cost

  • Voice and data converging onto mobile devices
  • 802.11 success => changing requirements

100 million WiFi handsets over the next 5 years. Source: IDC, Frost & Sullivan, Infonetics

Greater density Higher mobility Increased battery life

slide-3
SLIDE 3

Why Power and Rate Control?

Integral Components in Overall Solution

Adaptive transmit power control (TPC) can:

  • Improve spatial reuse and network capacity
  • Reduce energy consumption for mobile devices
  • Compensate for link changes due to mobility

Hard problem in WLANs due to distributed operation and unlicensed spectrum

slide-4
SLIDE 4

Symphony: A new approach to transmit power control (TPC) in 802.11 WLANs

  • Dense 802.11 Enterprise WLANs
  • Goals: increase network capacity and improve battery life

~50 ft. ~20 ft.

slide-5
SLIDE 5

Related Work

Dom ain Solution Granularity Realization Deployability Energy Capacity Rate Channel Access Asym m . Hidden Nodes W LANs [Sheth02] [Qiao03] [Akella05] [Chevillat05] Per-link Per-link Per-link Per-link √ X √ X √ √ √ √ √ √ X √ √ X √ √ √ √ √ √ X X X X X √ X X [Mhatre07] Per-cell √ √ X √ X √ X Ad-hoc Nets [Monks01] [Jung02] [Muqattash03] [Muqattash05] [Sheth05] [Kim06] [Shah07] Per-link Per-link Per-link Per-link Per-link Per-link Per-link X X X X X X X X √ X X X X X X √ √ √ √ √ √ √ X √ √ √ √ √ X X X X X √ X X X X X √ X √ √ √ √ √ √ X √ [Narayanaswa my02] Per-network √ √ X √ X √ √

  • No solution is both comprehensive and realizable
slide-6
SLIDE 6

Outline

  • Introduction
  • Challenges to TPC in WLANs
  • Symphony Design
  • Experimental Results
  • Summary
slide-7
SLIDE 7

Challenge #1: Receiver-side Interference and Asymmetric Channel Access

  • Receiver-side interference: Incorrect power reduction results

in increased packet error rate

B AP1 A

Receiver-side interference

C B AP1 AP2 A

Asym m etric channel access

  • Asymmetric channel access: Incorrect power reduction results

in increased channel access time

AP2

slide-8
SLIDE 8

Challenge #2: Interaction with Rate Adaptation

  • Measuring SINR at fine time scales --- infeasible with mobility
  • Use delivery ratio over a window of packets [Wong06]

Selecting rate and power non-trivial

  • Incorrect power reduction results in reduced rate

Power

(minRate, minPower)

Rate

maxPower maxRate

  • ve

+ve

slide-9
SLIDE 9

Challenge #3: Mobility

  • With mobility, challenges #1 and #2 can occur intermittently

Observations:

  • Performance at maximum power --- key reference point
  • Mobility => Periodic reference measurement needed
  • Some form of co-ordination between transmitters --- needed
slide-10
SLIDE 10

Outline

  • Introduction
  • Challenges to TPC in WLANs
  • Symphony Design
  • Experimental Results
  • Summary
slide-11
SLIDE 11

Symphony Design: Two-phase Execution

  • Goal: Link’s performance should be at least as good as in the

baseline maximum power network

REFERENCE Track performance at max. power OPERATIONAL Reduce power if possible

X1 ms have passed X2 ms have passed

slide-12
SLIDE 12

Symphony Design: Synchronous Operation

  • All transmitters cycle through the two phases in synchrony

Phase REF OPT At TX1 Time X1 X1 X1 X1 X2 X2 X2 Time REF OPT Phase At TX2 X1 X1 X1 X1 X2 X2 X2

slide-13
SLIDE 13

Outline

  • Introduction
  • Challenges to TPC in WLANs
  • Symphony Design
  • Experimental Results
  • Summary
slide-14
SLIDE 14

Symphony Evaluation

  • Does it deal well with the challenges?
  • Does Symphony increase spatial reuse?
  • Does it reduce energy consumption?
  • Is it incrementally deployable (with non-compliant nodes)?
slide-15
SLIDE 15

Platform: Indoor Office Testbed and ORBIT

  • Hardware: Dell laptops or ORBIT nodes (with Atheros NICs)
  • Software: Linux with MadWifi driver (v0.9.3.x)
  • Synchronization: NTP between APs, broadcast msg. for clients
  • Traffic: Mix of emulated VoIP, UDP @ 200pps, and TCP

Static Nodes Problem scenarios Mobile clients

slide-16
SLIDE 16

Results – Synopsis

Challenges Goals Channel Access Asymmetry Hidden Terminals Spatial Reuse Battery life Improvement 300% throughput improvement Detect and correct in 1 sec. Up to 50% improvement Up to 46% improvement

From a deployment perspective:

  • Symphony opportunistically reduces the transmit power by

6dB even in the presence of non-compliant nodes

slide-17
SLIDE 17

Summary

Symphony is

  • Effective in addressing TPC challenges
  • Easy to realize in a WLAN
  • Readily deployable (even with non-compliant nodes)
slide-18
SLIDE 18

Thanks!

For more information: http://www.winlab.rutgers.edu/~kishore/symphony-tech.pdf

slide-19
SLIDE 19

Extra Slides

slide-20
SLIDE 20

Traditional approach in WLANs: Static

  • r coarsely dynamic per-cell TPC
  • Strategy 1: Fixed TPC

For all clients, use maximum power (18-20dBm)

  • Strategy 2: Per-cell TPC [Mhatre07]

For all clients in a cell, use power level for weakest client Adjust power at coarse time intervals (10s of seconds) AP1A interferes with AP2B

A B AP1 AP2 C D

slide-21
SLIDE 21

Symphony: Dynamic per-link TPC

  • Choose power on a per-client (per-link) basis
  • Adjust power at fine-grained time intervals
  • Goal: Use “as much power as needed” and adapt to mobility

A B AP1 AP2 C D

AP1A and AP2B can “talk” simultaneously

slide-22
SLIDE 22

Symphony: Implementation

Implemented in MadWifi v0.9.3.1

Network Layer Device driver Interface card

  • Rcv. pkts

Receive path Transmit path SET Power Rate RTS Symphony Rate RX-side Asymm. access

  • Snd. pkts

tx_pkt() tx_pkt_complete() Power adaptation R O R O R O

slide-23
SLIDE 23

Symphony: Enables up to 18dB power reduction

  • Blah blah
slide-24
SLIDE 24

Symphony: Negligible R-score degradation

  • Average R-score difference: 2
  • Worst case R-score difference: 3.4
slide-25
SLIDE 25

Symphony: Up to 46% more battery life

Setup Symphony ORBIT Indoor Office Up to 46% (relative to max. power) Up to 33% (relative to max. power)

Active Mode Battery life I m provem ent

slide-26
SLIDE 26

Symphony: 30-50% throughput improvement

Tw o links Three links

slide-27
SLIDE 27

Results – Summary

  • Improves throughput of asymmetry-affected links by 300%
  • Is responsive to receiver side interference even at short

timescales of 1 second.

  • For mobile VOIP clients, opportunistically reduces transmit

power by up to 97%, with negligible voice quality degradation

  • Increases network throughput by up to 50%
  • Increases battery life by up to 46% (relative to max. power)
slide-28
SLIDE 28

Evaluation Details

  • Hardware: Dell laptops or ORBIT nodes (with Atheros NICs)
  • Software: Linux with MadWifi driver (v0.9.3.x)
  • Synchronization: NTP between APs, broadcast msg. for clients
  • Traffic: Mix of emulated VoIP, UDP @ 200pps, and TCP

Traditional Symphony Power: 18dBm fixed Rate: Variable (RRAA+) Power: 3dB step size Rate: Variable (RRAA+)

slide-29
SLIDE 29

Future Work