Ingegneria dell'Informazione

1
Department of Information EngineeringUniversity
of Padova, ITALY
Special Interest Group on NEtworking
Telecommunications
A Soft QoS scheduling algorithm for Bluetooth
piconets
Andrea Zanella, Daniele Miorandi, Silvano
Pupolin, Cristian Andreola
andrea.zanella, daniele.miorandi,
silvano.pupolin, freccia_at_dei.unipd.it
WPMC 2003, 21-22 October 2003
2
Outline of the contents

• Motivations Purposes
• Bluetooth Basic
• Hard-QoS Soft-QoS
• Soft-QoS for Bluetooth SFPQ
• Results and Demostration
• Concluding Remarks

3
What and Why

• Motivations Purposes

4
Motivations

• Demand for QoS support over portable electronic
  devices is increasing
• audio/video streaming
• interactive games
• multimedia
• Unfortunately, Bluetooth does not provide native
  QoS support

5
Aim of the study

• Adding Soft-QoS support to BT piconets
• Definition of Soft-QoS parameters
• Design of Soft-QoS scheduling algorithm
• Analisys of the proposed algorithm

6
What the standard says

• Bluetooth basic

7
Bluetooth piconet

• Two up to eight Bluetooth units sharing the same
  channel form a piconet
• In each piconet, a unit acts as master, the
  others act as slaves
• Channel access is based on a centralized polling
  scheme
• Full-duplex is supported by Time-division-duplex
  (TDD), with time slots of T0.625 ms

8
Multi-slot packets

• Data packets can be
• 1, 3, or 5 slot long
• Unprotected or 2/3 FEC protected
• Unprotected packet formats (DH)
• higher data capacity
• more subject to errors
• Protected packet formats (DM)
• medium data capacity
• higher protection against errors

9
Introduction to QoS issues

• QoS in Bluetooth networks

10
Resource Allocation

• Different types of applications
• Web Browsing
• Medium-to-high data rate
• Streaming audio
• Low delay and jitter
• High data rate
• Voice
• Low delay and jitter
• Low data rate

11
Hard Soft QoS

• Hard QoS
• Widely used in wired networks
• Integrated Services flow based (RSVP)
• Differentiated Services class based
• Soft QoS
• Suitable for wireless networks
• Applications may work even if, for short periods
  of time, QoS requirements are not satisfied
• Deal with limited bandwith and radio channel

12
FPQ algorithm

• Aim providing Hard QoS support by means of a
  Fair and efficient Polling scheme
• QoS parameters required for each link
• Expected data rate
• Maximum acceptable delay
• Adjust priorities of the slaves on the basis of
• Slaves queue length estimation
• Traffic parameters
• QoS parameters

FPQ a fair and efficient polling algorithm
with QoS support for Bluetooth piconet, INFOCOM03
13
FPQ scheme

• Purpose
• Determine the most efficient polling sequence
  fulfilling QoS requirements
• Slave Analyzer determines
• Pdata probability of having queued packets
• NSLP interval of time since last POLL/NULL
  sequence
• Selection Algorithm
• Determine priority of each data flow
• Select the master/slave link with highest
  priority
• Limits
• Inefficient service differentiation under high
  traffic loads

14
Soft QoS support

• Soft-FPQ algorithm for Bluetooth piconets

15
Soft-FPQ algorithm

• Aim Providing Soft QoS by means of dynamic
  estimation of flows satisfaction
• Definition of a new Soft QoS parameter Target
  Satisfaction
• Priorities are adjusted according to QoS
  parameters and the estimated satisfaction margin
  for each slave
• Low traffic high satisfaction for all flows
• High traffic distribute resources to fulfill
  exactly QoS request

16
Soft QoS parameters

• Average packet inter-arrival time ITL
• Average packet length PL
• Maximum sustainable packet delay MD
• Target Satisfaction index
• Percentage of packets that are expected to
  satisfy the QoS constrains

17
Dynamic Satisfaction estimation
Empty queue
Arrival rate
Arrival time
Probability
Estimated Satisfaction
18
Example of dynamic estimation
19
Istantaneous Satisfaction

• Estimated Satisfaction is updated anytime an AP
  packet is received
• To cope with long silence periods of slaves, we
  introduce the Istantaneous Satisfaction that is
  updated slot by slot according to function ??
• Istantaneous satisfaction is reset at the first
  AP arrival

Maximum Delay
Number of points
20
Satisfaction Margins
Satisfaction Margin
Actual Satisfaction
Target Satisfaction
Normalized Satisfaction Margin
21
Priority Evaluation

• Traffic Demand
• QoS Request

Fairness
Normalized Satisfaction Index
Constants
22
NS2 Simulation

• Simulation of QoS Bluetooth Piconet

23
Simulation Scenario

• Piconet with 7 slaves
• Only upstream traffic
• One application per slave
• One application x1 (Hard QoS)
• One application x0.9
• One streaming video application x0.9
• 4 Best Effort applications x0.2
• Simulation dynamic
• Slaves with high x are active for all the
  simulation time
• Best Effort transmissions start sequentially
  seconds apart
• When all the applications are active the system
  gets congested

24
Satisfaction perceived (1/3)
Target Satisfaction x1
Heavy Load
25
Satisfaction perceived (2/3)
Target Satisfaction x0.9
Heavy Load
26
Satisfaction perceived (3/3)
Target Satisfaction x0.2
27
Delay Distribution

• Video Streaming Delay Distribution
• Low traffic

• Video Streaming Delay Distribution
• High traffic

28
Video Streaming Demo

• Demo Structure
• RTP Server send packets of video stream
• RTP Client receive packets and display video
• Delay introduce pre-computed packet delays

• Scenario
• Upstream traffic only
• One application per slave
• One application streaming video, x0.9
• Two Best Effort application, x0.2

FPQ
SFPQ
29
Conclusions Future work

• Support of Soft QoS in Bluetooth
• Better service differentiation
• Efficient resource distribution
• Better support to real time and audio/video
  streaming applications
• Better behavior of the piconet under high traffic
  conditions

• Next steps
• Improve algorithms setting and introduce dynamic
  parameters tuning
• Extension of the algorithm to Scatternet
  structures
• Development of low complexity Satisfaction
  Estimation algorithms

30
Thats all!

• Thanks for
• your attention!