Upload
shana-walsh
View
217
Download
0
Tags:
Embed Size (px)
Citation preview
7 May 2008 1
Performance Comparison of Scheduling Algorithms for Multimedia Traffic
over High-Rate WPANsA note on the use of these ppt slides:
We’re making these slides freely available to all, hoping they might be of use for researchers and/or students. They’re in PowerPoint form so you can add, modify, and
delete slides (including this one) and slide content to suit your needs. In return for use, we only ask the following:
If you use these slides (e.g., in a class, presentations, talks and so on) in substantially unaltered form, that you mention their source.
If you post any slides in substantially unaltered form on a www site, that you note that they are adapted from (or perhaps identical to) our slides, and put a link to the authors
webpage:
www.dei.unipd.it/~zanella
Thanks and enjoy!
A note on the use of these ppt slides:We’re making these slides freely available to all, hoping they might be of use for
researchers and/or students. They’re in PowerPoint form so you can add, modify, and delete slides (including this one) and slide content to suit your needs. In return for use,
we only ask the following:If you use these slides (e.g., in a class, presentations, talks and so on) in substantially
unaltered form, that you mention their source.If you post any slides in substantially unaltered form on a www site, that you note that they are adapted from (or perhaps identical to) our slides, and put a link to the authors
webpage:
www.dei.unipd.it/~zanella
Thanks and enjoy!
7 May 2008 2
Performance Comparison of Scheduling Algorithms for Multimedia Traffic
over High-Rate WPANs
Fabio Lorquando, Andrea Zanella
Department of Information Engineering, University of Padova
presented by Nicola Laurenti
7 May 2008 3
Outline
• Introduction
• Traffic types and models
• Polling procedure
• Scheduling algorithms
• Simulation results
• Conclusions
7 May 2008 4
Introduction
WPAN technologies:
• mobile connectivity among different devices• constant diffusion and adoption increase
The 802.15.3 advance:
• high data rates• strong QoS (Quality of Service) awareness• flexible medium access
Providing intensive multimedia services:
• capable network framework• consciousness of delay bounds• efficient resource management
Careful scheduling policies and design
7 May 2008 5
Objectives and Approach
Paper approach:
• focus on classical and well known algorithmswell established, tested in a great range of scenarios, robust and readily portable in the novel platform
• knowledge of data requirements, not of data typeswe won’t focus on narrowed and specialized solutions, considering as much heterogeneous traffic environments as possible
• fully standard-compliant implementationsnot any improvement/change over the IEEE 802.15.3 protocol will be proposed, entirely relying on standards
Paper objectives:
• investigate IEEE 802.15.3 potentialities in multimedia services• pursue high QoS values in multipurpose WPANs
7 May 2008 6
WPAN IEEE 802.15.3 Basics
Overview:
• ad-hoc network (piconet)
• high bit rate (11 – 55 Mbps)
• centralized by means of a PicoNet Coordinator, PNC
• completely organized in superframe units
• advanced power management
Superframe role:
• piconet management• medium access coordination• medium access differentiation (CSMA/CA, TDMA, ALOHA)
7 May 2008 7
All exchanged data is transported using conventional frames
Through a specific CTRq (Channel Time Request) command frame, any DEV can negotiate a CTAP connection with the PNC. Relevant request parameters are, for example:
• minimal and desired resources per superframe
• traffic type (synchronous or not) and priority
• destination DEVs
MAC Management in IEEE 802.15.3
Beacon – signaling and information(PNC → ALL)
CAP – contention period(ALL ↔ ALL)
CTAP – channel reservation(DEVx → DEVy,z,…)
7 May 2008 8
Soft and Hard Real-Time Processes
Hard real-time processes must obey to hard deadlines. Packets delivered:
• before: represent a success, without discrimination
• after: are unusable and the stream experiences errors
Soft real-time processes set 2 two deadlines: soft and hard. Between them a region extends in which user experience degrades progressively, but data is still meaningful.
7 May 2008 9
MPEG-4MPEG-4:
• VBR (extremely high variance)
• Frame based
VoIPVoIP:
• CBR on/off (VAD)
• Frame based
GamingGaming::
• VBR
• Random arrivals
MPEG-4 Trace
Gaming Trace
HARD DEADLINES
Multimedia Traffic Models
SOFT DEADLINES
VoIP Model
7 May 2008 10
Proposed Polling Procedure
Proposed polling procedure:
• contention free (MCTAs)• managed by the PNC• flexible and efficient• built on 802.15.3 standard
Information exchanged at polling time:
• Stream ID
• Channel Time Request (CTRq)transmission time requested by in-queue data
• Waited Timetime already spent in the DEV queue by data
Stream set-up: upon connection, the DEV transmits its stream soft/hard deadline requirements
7 May 2008 11
GPS:
• fair scheduling: proportionality between requests and allocations
• low complexity
Employed Scheduling Algorithms
Different resource management policies share the same principle: the PNC gathers traffic info from its associated DEVs and executes a scheduling algorithm
We compare 2 main approaches: fair (GPS) and priority based (EDF type)
7 May 2008 12
Employed Scheduling Algorithms
EDF: data closer to theirs deadlines have precedence in channel admittance
Two valuable EDF variations are also considered:
EDF-DS: packets that even so won’t be able to meet their deadlines, are discarded (no channel resources)
EDF-SH: before discarding packets, the scheduler tries to defer access to other streams belonging to a soft real-time process
7 May 2008 13
Heterogeneous traffic scenario: 5 traffic profiles and JFRs measured per traffic class
1
2
3
4
5
0
10
20
30
40
MPEG-4VoIPGAME
Schedulers Results
Traffic Profiles
JFR
[%]
Str
ea
ms/
Pe
ers
GPS
EDF
EDF-DS
MPEG-4
VoIP
GAME
Simulation Results (1/3)
• Remarkable gains with EDF-based schedulers
• Streams with larger frames (as in MPEG-4) greatly benefit of deadline-aware algorithms
• Discard policy may prove valuable particularly in heavily loaded conditions
• VoIP traffic, due to low bit rate and packet size, is not an issue in a CTAP environment
7 May 2008 14
Heterogeneous traffic scenario: mean delay versus increasing MPEG-4 traffic
0 5 10 15 20 25 300
0.5
1
0 5 10 15 20 25 300
0.5
1
0 5 10 15 20 25 300
0.5
1
MPEG-4VoIPGame
Scheduler: GPS
Scheduler: EDF-SH
Scheduler: EDF-DS
JFR
[%]
JFR
[%]
JFR
[%]
JFR
[%]
MPEG-4 Streams
Simulation Results (2/3)
• GPS performs in a proportional fashion
• EFD solutions exhibit a threshold behavior, with excellent QoS until resources are depleted
• With EDF, QoS degradation happens per class of traffic
• EDF-SH solution does not outperform the simpler -DS one.
7 May 2008 15
Heterogeneous traffic scenario: JFR versus increasing MPEG-4 traffic
Simulation Results (3/3)
• Delays are obtained only for successful packets, GPS results are then faulty
• Not even for average delay EDF-SH sets results apart from -DS
0 5 10 15 20 25 300
0.025
0.05
0 5 10 15 20 25 300
0.025
0.05
0 5 10 15 20 25 300
0.025
0.05
MPEG-4VoIPGame
Scheduler: EDF-SH
Scheduler: EDF-DS
Scheduler: GPS
Me
an
De
lay
[s]
Me
an
De
lay
[s]
Me
an
De
lay
[s]
MPEG-4 Streams
7 May 2008 16
Conclusions
In this work we have:
• compared some well known scheduling algorithms
• proposed a standard polling procedure
• simulated piconets with heterogeneous traffic contents
The analysis has shown that:
• as expected, EDF-based schedulers show better performance than GPS-based. This difference is much larger in heterogeneous traffic scenarios than in homogeneous.
• the performance gain obtainable by discarding policies can be relevant
• EDF-SH does not seem to be decisive against the simpler EDF-DS variant
EDF schedulers need to access cross-layer information (i.e. soft/hard stream deadlines). Optimization of polling and signaling might improve results significantly.