Workflow resource allocation through auctions
This presentation is the property of its rightful owner.
Sponsored Links
1 / 43

Workflow Resource Allocation through Auctions PowerPoint PPT Presentation


  • 70 Views
  • Uploaded on
  • Presentation posted in: General

Workflow Resource Allocation through Auctions. Universitat de Girona. Albert Plà , Beatriz López, Javier Murillo. eXiT. Line up. Introduction WMS The Auction Experimentation & Results Conclusions. Introduction Workflow Management System Workflow Resource Allocation: The Auction

Download Presentation

Workflow Resource Allocation through Auctions

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Presentation Transcript


Workflow resource allocation through auctions

Workflow Resource Allocation through Auctions

Universitat de Girona

AILOG 2011 - Barcelona Universitat de Girona

Albert Plà, Beatriz López, Javier Murillo

eXiT


Line up

Line up

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • Introduction

  • Workflow Management System

  • Workflow Resource Allocation: The Auction

    • Workflow Agents: The Auctioneers

    • Resource Agents: The Bidders

  • Experimentation and Results

  • Conclusions and Future Work

AILOG 2011 - Barcelona Universitat de Girona


Introduction

Introduction

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

AILOG 2011 - Barcelona Universitat de Girona


Introduction1

Introduction

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

OrganizationInteractions

Schedules & Deadlines

DecentralizedProductionChain

Business Processes

CostReduction

Resources

Resources

AILOG 2011 - Barcelona Universitat de Girona


Introduction2

Introduction

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • Decentralization of the production processes.

    • Distribute production plants

    • Service outsourcing

    • Third party elements

    • Some tasks can not be Scheduled (e.g. faults)

    • Unknown service providers schedule

  • Customer oriented production methodologies

    • Lean Manufacturing

    • Avoid unnecessary work

    • Interaction between customer & provider

    • Production under demand

    • Production flexibility

    • Resource allocation in real time

AILOG 2011 - Barcelona Universitat de Girona


Introduction3

Introduction

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

Dynamism

Decentralization

Third Party Oustourcing

Contingency Robustness

Customer Orientation

Providers Privacy

Process Planing:

+ Uncertainity

+ Complexity

AILOG 2011 - Barcelona Universitat de Girona


Introduction4

Introduction

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

Dynamism

Decentralization

Third Party Oustourcing

Contingency Robustness

Customer Orientation

Providers Privacy

Process Planing:

+ Uncertainity

+ Complexity

AILOG 2011 - Barcelona Universitat de Girona


Introduction5

Introduction

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

Unknown # of online workflows

AILOG 2011 - Barcelona Universitat de Girona

Dynamism

Decentralization

Third Party Oustourcing

Contingency Robustness

Customer Orientation

Providers Privacy

Process Planing:

+ Uncertainity

+ Complexity


Example

Example:

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

Internal technicians

FaultReparation

Outsource technicians

?

Company 1

Comp. 2

Comp. n

Hospital devicemaintenanceservice

AILOG 2011 - Barcelona Universitat de Girona


Our proposal

Ourproposal:

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • Resource allocation using:

  • Workflow Management System (WMS)

  • Multi-Ageng Systems (MAS) technolgoy

  • Auctions

AILOG 2011 - Barcelona Universitat de Girona


Our proposal1

Ourproposal:

Introduction

Background?

WMS

TheAuction

Experimentation & Results

Conclusions

Workflow Management System

Auction

Organization

ResourceProviders

AILOG 2011 - Barcelona Universitat de Girona


Workflow management system

Workflow Management System

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

AILOG 2011 - Barcelona Universitat de Girona


Workflow management system1

Workflow Management System

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

Type: Technician A

ResourcesCapacity: 2

Currentlyusedresources: 1

Pla A. et al. Petri Net BasedAgentsforCoordinatingResources. Rome 2011

  • Workflows modeled using Resource-aware Petri Nets (RAPN)

    • Resource requirements included in the model

AILOG 2011 - Barcelona Universitat de Girona


Workflow management system2

Workflow Management System

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

WF Agent A

ResourceAgent 1

WF Agent B

ResourceAgent 2

ResourceAgent n-1

ResourceAgent n

WF Agent C

WF Agent D

Pla A. et al. Petri Net BasedAgentsforCoordinatingResources. Rome 2011

  • Monitoring using agents [1]

    • 1 Workflow agent for each workflow typology

    • 1 Resource agent for each resource

AILOG 2011 - Barcelona Universitat de Girona


Workflow management system3

Workflow Management System

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

WorkflowAgent A

ResourceType A

ResourceAgent 1

ResourceAgent 2

Pla A. et al. Petri Net BasedAgentsforCoordinatingResources. Rome 2011

  • A workflow agent monitors and manages all the instances of a workflow type

    • Call for auctions when they need resources

AILOG 2011 - Barcelona Universitat de Girona


Workflow management system4

Workflow Management System

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

ResoruceAgent A

ResourceType A

ResourceAgent 1

ResourceAgent 2

Pla A. et al. Petri Net BasedAgentsforCoordinatingResources. Rome 2011

  • A workflow agent monitors and manages all the instances of a workflow type [1]

    • Call for auctions when they need resources

AILOG 2011 - Barcelona Universitat de Girona


Workflow management system5

Workflow Management System

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • A workflow agent monitors and manages all the instances of a workflow type [1]

    • Call for auctions when they need resources

ResoruceAgent A

ResourceType A

ResourceAgent 1

ResourceAgent 2

AILOG 2011 - Barcelona Universitat de Girona


Workflow management system6

Workflow Management System

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

ResoruceAgent A

ResourceType A

AUCTION!

ResourceAgent 1

ResourceAgent 2

  • A workflow agent monitors and manages all the instances of a workflow type [1]

    • Call for auctions when they need resources

AILOG 2011 - Barcelona Universitat de Girona


Workflow resource allocation the auction

Workflowresourceallocation: TheAuction

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

AILOG 2011 - Barcelona Universitat de Girona


Workflow agent the auctioneer

Workflow Agent: TheAuctioneer

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • Differentattributtestobefullfilled

    • Time Restritctions (Starting & Ending Time)

    • ResourceCategory (E.g.TehcnicianLicense)

  • Minimizecosts

AILOG 2011 - Barcelona Universitat de Girona


Workflow agent the auctioneer1

Workflow Agent: TheAuctioneer

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • Reverse SealedAuction

    • First Price

  • PenaltyMechanisms:

    • Fraudulentbids

    • Delays

    • AgreementRepudiation

  • AttributePromotion:

    • Differentstrategies

    • WinningDeterminationProblem (WDP)

AILOG 2011 - Barcelona Universitat de Girona


Auctioneers strategies

AuctioneersStrategies

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • Balanced Strategy:

    • WDP: Lowest Bid Price

    • Balanced market price

    • Providers equilibrium

    • Decrease costs for workflow agent

    • Suitable for internal & external providers

    • + Delays

AILOG 2011 - Barcelona Universitat de Girona


Auctioneers strategies1

AuctioneersStrategies

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • DelayPreventionStrategy

    • WDP: Takesintoaccounttaskending time

    • Shortensworkflowtimings

    • Reduces number of delays

    • Indicatedfordealingwithinternalproviders

    • Arisescost

    • Fasterproviders can increasetheirprices

AILOG 2011 - Barcelona Universitat de Girona


Resource agents the bidders

ResourceAgents: TheBidders

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • Characterizedby:

    • Theircategories (1 or more)

    • List of scheduledtasks (private agenda)

    • Estimation of times neededtoperformtasksbasedonpastexperience (mean, stddeviation, etc)

AILOG 2011 - Barcelona Universitat de Girona


Resoruce agents the bidders

ResoruceAgents: Thebidders

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • TheBid:

    • Agentsevaluatetheir agenda

      • Find gaps for new tasks

      • Entertoauctiosn

    • Chooseprice (Agentcharacter):

      • Arisethepricewhenwinning

      • Decreasewhenlosing

      • Reinforcementlearning

    • Evaluatebenefits of winningtheauction

      • They can cancel scheduledactivitysifthebenefit of realizing a new taskcompensatesthecancellationpenalty.

      • Riskybids

AILOG 2011 - Barcelona Universitat de Girona


Experimentation and results

Experimentationand Results

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

AILOG 2011 - Barcelona Universitat de Girona


Experimental setup

Experimental Setup

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • WorkflowSimulation

    • 200 t.u.simulations

    • Workflowprobabilityp = 0.2

  • Evaluation:

    • Economiccost

    • Delaysproduced

    • Workload balance

  • Strategies:

    • BalancedStrategy (BS)

    • DelayPreventionStrategy (DPS)

    • Firstavailableresource of a sortedlist (FRQ)

AILOG 2011 - Barcelona Universitat de Girona


Experimental setup1

Experimental Setup

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • Scenario 1:

    • 4 ResourceAgents

    • Eachresources can perform, at least, 3 differenttasks

    • Competitionbetweensomeresources

    • Workflow instances > resources

  • Scenario 2:

    • 4 ResourceAgents

    • Eachresource can performallthetasks

    • Highcompetitionbetweenallresources

    • Workflow instances >> resources

AILOG 2011 - Barcelona Universitat de Girona


Results

Results

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

Scenario 1Scenario 2

AILOG 2011 - Barcelona Universitat de Girona


Results delays

Results: Delays

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

Scenario 1Scenario 2

AILOG 2011 - Barcelona Universitat de Girona


Results costs

Results: Costs

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

Workflow agentsbudged

AILOG 2011 - Barcelona Universitat de Girona


Results workload balance

Results: Workload Balance

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

Balanced

Strategy

FIFO

Resource

Queue

Scenario 2

AILOG 2011 - Barcelona Universitat de Girona


Conclusions and future work

Conclusions and FutureWork

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

AILOG 2011 - Barcelona Universitat de Girona


Conclusions

Conclusions

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • Howtoallocateinternal and foreignresourcesunderdemand in a decentralizedproductionprocess?

    • Workflow Management System

      • Business Process

      • Resources

    • MultiAgentSystems

      • Useful in distributedenvironments

    • Auctions

      • Minimizecosts

      • Reduce production times

      • Balance resourceoccupation

AILOG 2011 - Barcelona Universitat de Girona


Conclusions1

Conclusions

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • 2 Kind of agents:

    • ResourceAgents

      • Bidders

    • Workflow Agents

      • Auctioneers

      • Differentstrategies

        • Resourcecosts (BalancedStrategy)

        • Workflow timings (DelayPreventionStrategy)

  • Improve simple schedulesmethods

AILOG 2011 - Barcelona Universitat de Girona


Further work

Furtherwork

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

  • Vickeryauctioningmechanism

    • Incentive compatibility

    • Fightcheatingagents

  • New auctioningstrategies

    • Multi-attribute

  • Improve MAS capabilities

    • Trust

    • Historical data

AILOG 2011 - Barcelona Universitat de Girona


Workflow resource allocation through auctions

Introduction

WMS

TheAuction

Experimentation & Results

Conclusions

[email protected]

AILOG 2011 - Barcelona Universitat de Girona


Workflow resource allocation through auctions1

Workflow Resource Allocation through Auctions

Universitat de Girona

AILOG 2011 - Barcelona Universitat de Girona

Albert Plà, Beatriz López, Javier Murillo

eXiT


Workflow resource allocation through auctions

AILOG 2011 - Barcelona Universitat de Girona


Workflow resource allocation through auctions

AILOG 2011 - Barcelona Universitat de Girona


Workflow resource allocation through auctions

WMS Architecture

AILOG 2011 - Barcelona Universitat de Girona


Workflow resource allocation through auctions

  • Multiple Reactive Maintenance Intervention (MRMI)

Assigning to in-house technicians

Maintenance action generated in the system

Ranking the action (assigning a level)

Locating the source of the action

Classifying the action

Sequential Solution

Technician 1

reparation

Concurrent Solution

Technician 2

reparation

Technician 1

reparation

Technician 2

reparation

Device returned to the medical center

Assigning to an external service provider

Device Check Out

Resoruces:

TechnicianType A

TechnicianType B

AILOG 2011 - Barcelona Universitat de Girona


Workflow resource allocation through auctions

T4

P4

T3

P1

P2

P3

T5

P5

T1

T2

AILOG 2011 - Barcelona Universitat de Girona


  • Login