MadMAC: Building a Reconfigurable Radio Testbed using Commodity 802'11 Hardware

1
MadMAC Building a Reconfigurable Radio Testbed
using Commodity 802.11 Hardware

• A. Sharma et. al.

2
Agenda

• Introduction
• Related Work
• MadMAC Overview
• Implementation
• Evaluation
• Remarks

3
Introduction

• Chosen because of similarity with SoftQ
• Actual implementation based on MadWifi
• Service routine is similar.
• Both demonstrate the feasibility of an idea.

4
Introduction

• Motivation
• Adaptive radios need an adaptive MAC.
• Two common options for MAC protocols
• Reservation based are best for stable,
  predictable traffic.
• Random channel access best for variable traffic.
• Goals
• Design kernel-mode driver on top of MadWifi.
• Implement TDMA MAC protocol using MadMAC.
• Evaluate performance on 2-node test bed.

5
Agenda

• Introduction
• Related Work
• MadMAC Overview
• Implementation
• Evaluation
• Remarks

6
Related Work

• Two categories
• Build an overlay on top of existing CSMA
  hardware
• Overlay MAC Layer, Layer 2.5 softMAC
• Leverage configurable Atheros chips.
• SoftMAC, MultiMAC, MadMAC

7
Agenda

• Introduction
• Related Work
• MadMAC Overview
• Implementation
• Evaluation
• Remarks

8
MadMAC Overview

• MadMAC executes as a kernel-mode device driver.
• Requires specific MadWifi Configuration
• Monitor Mode.
• Set contention window to 1.
• Prevent MAC sequence numbers.

9
MadMAC Overview
10
Agenda

• Introduction
• Related Work
• MadMAC Overview
• Implementation
• Evaluation
• Remarks

11
Implementation

• Interface with Network Layer
• Add/Removes MAC specific header/checksum
• Adds packets to transmit buffer

12
Implementation

• Packet Transmission/Reception
• Transmits according predefined TDMA schedule.
• Slot structure maintained using guard times
• Multiple packets sent in each transmission slot.

13
(No Transcript)
14
Implementation

• Synchronization
• Based on Timing-sync Protocol for Sensor
  Networks.
• Activated during system initialization but not
  implemented over wireless link, only the ethernet
  link.
• Should be run every 100s because of measured 10
  ms clock drift.

15
Agenda

• Introduction
• Related Work
• MadMAC Overview
• Implementation
• Evaluation
• Remarks

16
Evaluation

• Test Bed Setup
• Two wireless nodes
• Alternating transmission slots
• Two Experiments
• 64 byte ping packets sent every second to measure
  RTT delay and processing overhead.
• Use IPerf to measure throughput (1470 byte
  packets).

17
Evaluation
18
Evaluation
19
Evaluation
20
Agenda

• Introduction
• Related Work
• MadMAC Overview
• Implementation
• Evaluation
• Remarks

21
Remarks

• Summary
• Reconfigurable MAC, MadMAC, architecture
  described.
• TDMA protocol implemented using MadMAC.
• Using 2-node test bed, it was shown that MadMAC
  has low processing overhead and is 20 better
  than standard CSMA.

22
Remarks

• Positives
• Practical TDMA implementation.
• Layered architecture.
• Negative
• TDMA protocol doesnt consider application
  requirements.
• Doesnt mention how to re-use architecture for
  non-TDMA protocol

23
Questions?