Enabling Dynamic Adaptivity in WLAN MAC Protocol

1
Enabling Dynamic Adaptivity inWLAN MAC Protocol

• Naomi Ramos
• Debashis Panigrahi
• Dr. Sujit Dey
• ECE Department
• University of California, San Diego
• http//esdat.ucsd.edu

2
Outline

• Presentation Outline
• Introduction Motivation
• Topics to be covered
• Considering channel, application, node
  requirements
• Determining adaptable parameters
• Dynamic adaptations
• Simulation results
• Conclusions
• Future work

3
Introduction and Research Plan

• Overall Objectives
• To enable adaptability within multiple protocol
  stacks to adjust to current conditions,
  application, and node requirements
• Overall Research Project Plan
• Identify parameters in selected MAC protocol
  tasks that are adaptable.
• Investigate the impact of these parameters on
  bandwidth, QoS, latency and energy consumption.
• Develop techniques to dynamically select protocol
  parameters based on channel and application
  requirements.
• Current research stage
• Identified adaptable parameters and developed
  techniques to enable dynamic adaptivity in IEEE
  802.11b MAC layer

4
Motivation for Dynamic MAC Layer

• Traditional functionality of a MAC layer
• Definition of a control scheme to access the
  shared link
• Ensure fairness, efficient use of channel
• Limited static configurations to satisfy
  application requirements and to adapt to dynamic
  channel conditions
• Potential impact of a dynamic MAC layer
• Satisfy latency constraints
• Provide service differentiation
• Conserve node energy

MAC Layer
Physical Channel
Using existing MAC Layer parameters mechanisms,
can we provide dynamic adaptability to address
application, network, node requirements?
5
MAC Layer Adaptation Framework

• Monitor current conditions requirements
• Provide dynamic adaptability
• Adaptation Block configures the MAC layer to
  adjust to these conditions

NODE Device Monitor
APP Application Module
PHY/MAC Channel Estimator
Run-time MAC Adaptation Layer
MAC Layer
6
Current Conditions Requirements

• PHY/MAC What information can be used to
  determine the current state of the network?
• Measurements Signal Strength, BER
• History Past Retransmissions, Idle Channel Time,
  Delay
• Predictive Virtual MAC
• APP How to relay application needs to MAC
  Adaptation Layer?
• Pre-defined classes provide information about
  needed priority, latency tolerance, traffic
  characteristics, etc. i.e. WCDMA classes ATM
• NODE What type of node constraints can be
  provided to the MAC layer?
• Static Computational Resources, Memory
• Dynamic Battery Capacity

1 Barry, M., Campbell, A.T and A. Veres ,
"Distributed Control Algorithms for Service
Differentiation in Wireless Packet Networks",
Proc. IEEE INFOCOM'2001, Anchorage, Alaska, April
2001.
7
MAC Adaptation Block

• Objective Given current conditions,
  application, node requirements, determine the
  appropriate adaptations in the MAC layer.
• Method of Adaptations
• Identify parameters in IEEE 802.11 MAC protocol
  tasks that are adaptable.

RTS
PAYLOAD
HDR
Src
SIFS
SIFS
SIFS
DIFS
BACKOFF
ACK
CTS
Dst

• Investigate the impact of these parameters on
  latency and energy consumption.
• Develop runtime adaptation policies to
  dynamically select protocol parameters.

8
MAC Layer Parameter Descriptions

• Fragmentation Threshold
• Packets exceeding this threshold are divided into
  smaller packets
• Useful in bad Channel conditions, but has
  associated overhead in good channel conditions

Bad Channel Condition
Good Channel Condition

• RTS/CTS Threshold
• Packets exceeding this threshold have the RTS/CTS
  mechanism enabled.
• Mechanism to combat Hidden Node Problem
• Can be used as a reservation mechanism
• Useful only when hidden node scenarios
• are more probable, otherwise energy/latency
  overhead

B
A
C
9
MAC Layer Parameters Descriptions contd.

• Contention Window
• Generated a random slot with in Contention Window
  to contend for the medium
• A node with lower slot number gets to have access
  to the medium gt lower contention window leads to
  low latency
• Depends on other nodes of the network gt range of
  contention window is important rather than
  absolute value
• Other Parameters
• Re-transmission Limit
• Back-off Window
• Transmit Power
• Power-saving Mechanisms

10
OPNET Simulation Setup

• Energy Model
• Introduced into OPNET simulation model
• Transmit 280mA Receive 200mA
• Idle 0 Doze 0 Lucent PC Card
• Transmitter Power
• Access Point 60mW
• Mobile 33mW
• Network
• Infrastructure Mode of IEEE 802.11
• No of nodes 1 32
• Simulated Time 20-30 min.
• Traffic Settings
• Traffic1 FTP, 100 Put, 8K File size,
  Inter-request Gap poisson(2)
• Traffic2 FTP, 50 Put, 8K File size,
  Inter-request Gap poisson(2)
• Traffic3 FTP, 100 Put, 4K File size,
  Inter-request Gap poisson(10)
• Traffic4 FTP, 50 Put, 4K File size,
  Inter-request Gap poisson(10)
• Channel Model
• Path Loss
• Co-channel Interference

11
Impact of Fragmentation Threshold
1 node Traffic1 Traffic2

• Fragmentation leads to energy savings in bad
  channel conditions
• Under good channel conditions, fragmentation is
  not useful because
• of additional overhead (header of fragments)

12
Impact of Fragmentation Threshold

• Effect on Re-Transmission/Latency
• Re-transmission percentage increases with
  increase in fragmentation threshold
• Latency
• BER 0 0.112-0.114s
• BER 10-4 4 900s
• Fragmentation reduces latency in bad channel
  conditions

13
Impact of Contention Window Parameter
Traffic1

• Contention Window Parameter leads to significant
  reduction in latency
• in a congested network (20)
• Effect of contention window is low for a network
  with few nodes

14
Impact of RTS/CTS Threshold
One node Traffic3 No fragmentation
When no hidden node problem exists, the overhead
of RTS/CTS is approximately 10.
15
Runtime Adaptation Policies Fragmentation

• Existing schemes to modify fragmentation
  threshold
• S1 Statically decides the fragmentation
  threshold at the configuration time (supported in
  existing cards)
• S2 Pessimistically decides the fragmentation
  threshold based on past BER of downlink
• Proposed Schemes
• Dynamically Variable Open Loop DVOL Decides a
  range of fragmentation threshold to use based on
  BER information
• Dynamically Variable Closed Loop DVCL Decides
  the initial fragmentation threshold based on
  current BER and adapts the fragmentation
  threshold using past re-transmission counts

16
Runtime Adaptation Policies Fragmentation
17
Runtime Adaptation Policies Fragmentation
Dynamic Schemes fair better in terms of Energy
Consumption (around 17) and Goodput (around
18) The closed loop adaptation scheme is
better than other open loop schemes
18
Additional Runtime Adaptation Policies

• RTS/CTS Policy
• Disable RTS/CTS after successful transmission of
  a fixed number of packets
• Enable RTS/CTS mechanism only when retransmission
  occurs
• Use signal strength and the number of users as a
  guide.
• Contention Window Policy
• Depending on the applications latency
  requirements, adjust the contention window to a
  lower value to ensure prioritized access to
  medium
• Scale countdown number based on the past waiting
  period instead of one slot each time

19
Conclusions Future Work

• Evaluated effects of MAC Layer Parameters through
  simulation based environments
• Proposed run-time adaptation policies based on
  the above study for each of the parameters
• Proposed a framework to enable MAC Layer
  Adaptations
• Investigating dependencies between different
  run-time adaptation policies
• Develop an integrated policy consisting of the
  parameter specific policies
• Implement MAC adaptation policy , as well as
  channel estimator, application module and node
  monitor
• Incorporate realistic user behavior models into
  future simulations
• Extend MAC level adaptability to other access
  technologies