Page 1
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Adaptive Scheduling of Data Pathsusing Uppaal Tiga
Israa AlAttili Fred Houben Georgeta IgnaSteffen Michels Feng Zhu Frits Vaandrager
Radboud University Nijmegen
November 3, 2009
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 1/29
Page 2
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Outline
1 Introduction
2 Oce Case Study
3 Uppaal Tiga
4 Model
5 Results
6 Conclusions & Future Work
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 2/29
Page 3
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Outline
1 IntroductionScheduling ProblemsUncertaintyGoal of this Research
2 Oce Case Study
3 Uppaal Tiga
4 Model
5 Results
6 Conclusions & Future WorkAlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 3/29
Page 4
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Scheduling Problems
allocation of resources to activities over time
in order to achieve some goals
many different domains
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 4/29
Page 5
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Uncertainty
existing literature
a function of known, perfect inputs
scheduling processes in practice: driven by uncertainty
machine breakdownunexpected arrival of new jobsmodification of existing jobsuncertainty of task durations...
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 5/29
Page 6
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Goal of this Research
industrial case study: Oce printer/copier
address problem of uncertain job arrival times
use Uppaal Tiga
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 6/29
Page 7
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Outline
1 Introduction
2 Oce Case StudyOverviewData-PathsSchedule
3 Uppaal Tiga
4 Model
5 Results
6 Conclusions & Future WorkAlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 7/29
Page 8
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Oce Copy/Printer Overview
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 8/29
Page 9
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Oce Copy/Printer Data-Paths
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 9/29
Page 10
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Schedule of print/copy jobs
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 10/29
Page 11
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Outline
1 Introduction
2 Oce Case Study
3 Uppaal TigaController vs EnvironmentPrinter vs User
4 Model
5 Results
6 Conclusions & Future Work
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 11/29
Page 12
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Controller vs Environment
control: A[] a0.critical <= 11
not satisfied
control: A[] a0.critical <= 12
satisfied
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 12/29
Page 13
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Printer vs User
Printer
process jobsmeet timing constraints
User
add jobsmoment is unforeseeable
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 13/29
Page 14
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Outline
1 Introduction
2 Oce Case Study
3 Uppaal Tiga
4 ModelOverviewResourcesCopy JobsPrint JobWinning Condition
5 Results
6 Conclusions & Future Work
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 14/29
Page 15
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Uppaal Model Overview
based on model by Igna et al. [FORMATS’08]
network of timed game automata
each use case & resource described by automaton
restriction to simple scenario
continuous stream of copiesuncontrollable print job
observer for finished copies
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 15/29
Page 16
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Resource Template
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 16/29
Page 17
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Copy Jobs
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 17/29
Page 18
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Observer
control: A[](DC_OBSERVER.INIT imply DC_OBSERVER.x <= FIRST_DC_TIME)&&(!DC_OBSERVER.INIT imply DC_OBSERVER.x <= DC_TIME)
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 18/29
Page 19
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Print Job
control: A[](!DP0.INIT imply DP0.timeSinceArrival <= DP_TIME)
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 19/29
Page 20
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Winning Condition
control: A[](DC_OBSERVER.INIT imply DC_OBSERVER.x <= FIRST_DC_TIME)&&(!DC_OBSERVER.INIT imply DC_OBSERVER.x <= DC_TIME)&&(!DP0.INIT imply DP0.timeSinceArrival <= DP_TIME)
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 20/29
Page 21
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Outline
1 Introduction
2 Oce Case Study
3 Uppaal Tiga
4 Model
5 ResultsOptimal StrategiesExtracting StrategiesComparison With Fixed Strategies
6 Conclusions & Future WorkAlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 21/29
Page 22
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Optimal Strategies
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 22/29
Page 23
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Extracting Strategies
Tiga can generate strategies as set of rules
not usable for real printer controller
very large sizeinclude parts of model not existing in real worlddoes not abstract from number of jobs
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 23/29
Page 24
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Comparison With Fixed Strategies
in practice: strategy should not be too complex
comparing optimal strategies with simple ones can be helpful
can also give hints how to improve existing strategies
we built three simple strategies into our model
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 24/29
Page 25
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Comparison With Fixed Strategies
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 25/29
Page 26
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Outline
1 Introduction
2 Oce Case Study
3 Uppaal Tiga
4 Model
5 Results
6 Conclusions & Future WorkConclusionsFuture WorkQuestions
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 26/29
Page 27
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Conclusions
application of timed automata to industrial scheduling problem
limited to simplified model/scenario
Tiga close to the point being helpful for actual design
indication of how close implemented rules are from optimumfinding bottlenecksmay help to find out and test better rules
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 27/29
Page 28
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Future Work
improve efficiency of algorithm used by Tiga
more realistic modelmore complex scenarios
reduce size of generated strategy
abstract also from number of jobs
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 28/29
Page 29
Introduction Oce Case Study Uppaal Tiga Model Results Conclusions & Future Work
Questions?
AlAttili, Houben, Igna, Michels, Zhu, Vaandrager Adaptive Scheduling of Data Paths Using Uppaal Tiga 29/29