SlideShare uma empresa Scribd logo
1 de 32
Double Auction-based Scheduling of
Scientific Applications in Distributed
   Grid and Cloud Environments




                       Francesc Lordan Gomis
                       Metodologia de Recerca en la Informàtica
                       25/06/2012
Outline

Introduction
Auction Theory
Model and strategies
Results
Conclusions




          Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   1
Introduction
Grid and Cloud Computing allow user to execute applications on a third-party
 infrastructure by paying for the access to the remote resources




                Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   2
Introduction
Users are represented by Schedulers that picks which resource a job will be run

The providers are represented by Resource Managers that try to maximize the
 incomes of the provider by optimizing the resource usage




                                           Resource                                             Resource
                                           Manager                                              Manager




       Scheduler
                                             Resource                                         Resource
                                             Manager                                          Manager




                   Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   3
Introduction
When many users try to obtain the same resources, they compete for them.

Since access to these resources is paid, it can be seen as a market governed by
  the supply-demand law

Economy-based negotiation techniques are attractive from bussiness point of view

              Scheduler                                                                                 Resource
                                                                                                        Manager
              Scheduler
              Scheduler
                                              Market                                                    Resource




                                                      ?
                                                                                                        Manager
              Scheduler
              Scheduler                                                                                 Resource
                                                                                                        Manager
              Scheduler
              Scheduler                                                                                 Resource
                                                                                                        Manager
              Scheduler

                Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   4
Auction Theory
Auctions are an economic mechanism used for allocating a set of resources
 among a group of bidders.

Models:

 Single-sided auctions: the seller submits a resource and the bidders submit their
   offers to buy (bid)

 - English auction: buyers constantly increase their offers until the highest one wins

 - Dutch auction: the seller decreases the price of the resource until a buyer decides
   to pay the proposed price

 - Sealed auctions: buyers give their bids on an envelope
     First-price: the highest bid wins
     Vickrey second-price: the second highest bid wins




                Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   5
Auction Theory
Auctions are an economic mechanism used for allocating a set of resources
 among a group of bidders.

Models:

 Double-sided auctions: all the participant submit their offers to sell (asks) and
   buy (bids)

 - Call Auctions: during a timeframe all the participants submit their offers and
   finally they try to match

 - Continuous Double Auction (CDA): offers are continously submitted until a
   match between the lowest ask and the highest bid is achieved or the auction
   times out.

 Combinatorial auctions: a single bid for multiple resource (NP-complete)




                Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   6
Model and strategies




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   7
Model and strategies
The limitation on resources converts the schedulers into competitors with
 their own negotiation strategy.




Any strategy is based on 4 factors
 Resource selection
 Resource valuation
 Auction participation
 Bidding




               Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   8
Model and strategies

Resource Selection

Before bidding, the scheduler decides for which resources it will participate
 in an auction.

When a task is sumitted the scheduler looks for a pending auction. If there is
 no such auction the scheduler postpones the bidding until there is one.


The selection is done with a lottery where probabilities take into account:
Availability of the resource
Execution time
Budget cost




               Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   9
Model and strategies
Resource Valuation

User Valuation
     2 steps:
     - Update the value of the resource depending on the final price of its
        last auction and a smoothing factor α.
     - Increases the value depending on the number of lost auctions and a
        revaluation factor β.


Provider Valuation
      Depending on the success ratio on the last auctions for the resource
  the price is increased or decreased by a resource revaluation factor γ




               Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   10
Model and strategies

Auction participation

Scheduler
     participates when there is at least one task waiting for that resource.
         Tasks are kept in a priority queue sorted by the bottom levels

Resource Manager
    The time between two auctions of the same resource are created
        depend on the result for the last auctions.

     If it was won, it wait until the resource has been used and adds a
 delay
     If it was lost, the delay is incremented exponentially with a delaying
           factor δ




               Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   11
Model and strategies

Bidding
Zero Intelligence: participants submit their offers periodically

Scheduler
 tries to reduce the value of the resource by decreasing the distance
    between its valuation and the highest bid

Resource Manager
 tries to increase the value of the resource by reducing the distance
    between its valuation and the lowest ask




                Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   12
Model and strategies
Modifications to the base strategy:

Self-limitation
 Reduce the price by reducing the competence
 limit the number of auctions where the scheduler participates in based on:
    The average number of retries
    A resource limitation factor ζ




Pricing Agressiveness
  Change the revaluation factor of a single scheduler
  Value of the resource are increased faster or slower




               Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   13
Results
Resource Valuation




            Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   14
Results
Resource Valuation




            Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   15
Results
Self-limitation




                  Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   16
Results
Pricing aggressiveness




             Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   1
                                                                                                                    17
Conclusions
Creation of an open market of resources implemented as a CDA instance

Identification the general behaviour patters

Evaluation of strategies:
 Self-limitation:
   - has marginal improvements on specific configuration
   - NOT RECOMENDED
 Concessive strategy:
   - very good for the common-drift case
 Aggressive-strategy:
   - reduces execution time with low impact on the budget on the single-user
      deviation case
   - higher budget and similar execution time on the common-drift case
   - RISKY




               Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   18
Questions




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment   19
Resource Selection
                                    availability



                                    Expected execution time and cost at any resource



                                 Expected execution time

                          Resource value




0 <= rsfmin <=rsfmax<=1                                               Resource selection ratio




       Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Resource Valuation
Scheduler                                  Smoothing factor


                                           Last final price                                 Last valuation
            Linear


            Exponential                                      Resource revaluation factor

                                                                                # lost auctions for the resource

Resource Manager
                                                              Average success ratio




                                              Resource revaluation factor


                Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Auction Participation


Delay factor
                                                                   # lost auctions for the resource




 Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Self-limitation


    Resource limitation factor




                                 Average retry number




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Workflows




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Pricing aggressiveness




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Pricing aggressiveness




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Pricing aggressiveness




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Pricing aggressiveness




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Pricing aggressiveness




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Pricing aggressiveness




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
Pricing aggressiveness




Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment

Mais conteúdo relacionado

Destaque

Cloud Computing Principles and Paradigms: 11 t-systems cloud-based solutions ...
Cloud Computing Principles and Paradigms: 11 t-systems cloud-based solutions ...Cloud Computing Principles and Paradigms: 11 t-systems cloud-based solutions ...
Cloud Computing Principles and Paradigms: 11 t-systems cloud-based solutions ...
Majid Hajibaba
 
8 secure distributed data storage in cloud computing
8 secure distributed data storage in cloud computing8 secure distributed data storage in cloud computing
8 secure distributed data storage in cloud computing
Majid Hajibaba
 
Cloud Computing Principles and Paradigms: 10 comet cloud-an autonomic cloud e...
Cloud Computing Principles and Paradigms: 10 comet cloud-an autonomic cloud e...Cloud Computing Principles and Paradigms: 10 comet cloud-an autonomic cloud e...
Cloud Computing Principles and Paradigms: 10 comet cloud-an autonomic cloud e...
Majid Hajibaba
 
Fashion of 1960s 70s 80s 90s
Fashion of 1960s 70s 80s 90sFashion of 1960s 70s 80s 90s
Fashion of 1960s 70s 80s 90s
Erkan Kahraman
 

Destaque (7)

Cloud Computing Principles and Paradigms: 11 t-systems cloud-based solutions ...
Cloud Computing Principles and Paradigms: 11 t-systems cloud-based solutions ...Cloud Computing Principles and Paradigms: 11 t-systems cloud-based solutions ...
Cloud Computing Principles and Paradigms: 11 t-systems cloud-based solutions ...
 
8 secure distributed data storage in cloud computing
8 secure distributed data storage in cloud computing8 secure distributed data storage in cloud computing
8 secure distributed data storage in cloud computing
 
Cloud Computing Principles and Paradigms: 10 comet cloud-an autonomic cloud e...
Cloud Computing Principles and Paradigms: 10 comet cloud-an autonomic cloud e...Cloud Computing Principles and Paradigms: 10 comet cloud-an autonomic cloud e...
Cloud Computing Principles and Paradigms: 10 comet cloud-an autonomic cloud e...
 
Fashion of 1960s 70s 80s 90s
Fashion of 1960s 70s 80s 90sFashion of 1960s 70s 80s 90s
Fashion of 1960s 70s 80s 90s
 
Intro to High Performance Computing in the AWS Cloud
Intro to High Performance Computing in the AWS CloudIntro to High Performance Computing in the AWS Cloud
Intro to High Performance Computing in the AWS Cloud
 
Cloud computing Basics
Cloud computing BasicsCloud computing Basics
Cloud computing Basics
 
Introduction of Cloud computing
Introduction of Cloud computingIntroduction of Cloud computing
Introduction of Cloud computing
 

Semelhante a CDA for Grid and Cloud

OPTIMIZED RESOURCE PROVISIONING METHOD FOR COMPUTATIONAL GRID
OPTIMIZED RESOURCE PROVISIONING METHOD FOR COMPUTATIONAL GRID OPTIMIZED RESOURCE PROVISIONING METHOD FOR COMPUTATIONAL GRID
OPTIMIZED RESOURCE PROVISIONING METHOD FOR COMPUTATIONAL GRID
ijgca
 
Optimized Resource Provisioning Method for Computational Grid
Optimized Resource Provisioning Method for Computational GridOptimized Resource Provisioning Method for Computational Grid
Optimized Resource Provisioning Method for Computational Grid
ijgca
 
r.elativity Capabilities Summary
r.elativity Capabilities Summaryr.elativity Capabilities Summary
r.elativity Capabilities Summary
bronzerocket
 
Accumulo Summit 2016: Accumulo Indexing Strategies for Searching Semantic Net...
Accumulo Summit 2016: Accumulo Indexing Strategies for Searching Semantic Net...Accumulo Summit 2016: Accumulo Indexing Strategies for Searching Semantic Net...
Accumulo Summit 2016: Accumulo Indexing Strategies for Searching Semantic Net...
Accumulo Summit
 
Iss 6
Iss 6Iss 6
Iss 6
ijgca
 
Event driven optimizer an approach to optimize the events that transition fr...
Event driven optimizer  an approach to optimize the events that transition fr...Event driven optimizer  an approach to optimize the events that transition fr...
Event driven optimizer an approach to optimize the events that transition fr...
prjpublications
 

Semelhante a CDA for Grid and Cloud (20)

IRJET-Framework for Dynamic Resource Allocation and Efficient Scheduling Stra...
IRJET-Framework for Dynamic Resource Allocation and Efficient Scheduling Stra...IRJET-Framework for Dynamic Resource Allocation and Efficient Scheduling Stra...
IRJET-Framework for Dynamic Resource Allocation and Efficient Scheduling Stra...
 
Allocation Strategies of Virtual Resources in Cloud-Computing Networks
Allocation Strategies of Virtual Resources in Cloud-Computing NetworksAllocation Strategies of Virtual Resources in Cloud-Computing Networks
Allocation Strategies of Virtual Resources in Cloud-Computing Networks
 
OPTIMIZED RESOURCE PROVISIONING METHOD FOR COMPUTATIONAL GRID
OPTIMIZED RESOURCE PROVISIONING METHOD FOR COMPUTATIONAL GRID OPTIMIZED RESOURCE PROVISIONING METHOD FOR COMPUTATIONAL GRID
OPTIMIZED RESOURCE PROVISIONING METHOD FOR COMPUTATIONAL GRID
 
Optimized Resource Provisioning Method for Computational Grid
Optimized Resource Provisioning Method for Computational GridOptimized Resource Provisioning Method for Computational Grid
Optimized Resource Provisioning Method for Computational Grid
 
CloudLighting - A Brief Overview
CloudLighting - A Brief OverviewCloudLighting - A Brief Overview
CloudLighting - A Brief Overview
 
Overview of CloudLightning
Overview of CloudLightningOverview of CloudLightning
Overview of CloudLightning
 
IRJET- Improving Data Availability by using VPC Strategy in Cloud Environ...
IRJET-  	  Improving Data Availability by using VPC Strategy in Cloud Environ...IRJET-  	  Improving Data Availability by using VPC Strategy in Cloud Environ...
IRJET- Improving Data Availability by using VPC Strategy in Cloud Environ...
 
Efficient Resource Sharing In Cloud Using Neural Network
Efficient Resource Sharing In Cloud Using Neural NetworkEfficient Resource Sharing In Cloud Using Neural Network
Efficient Resource Sharing In Cloud Using Neural Network
 
A Review on Resource Discovery Strategies in Grid Computing
A Review on Resource Discovery Strategies in Grid ComputingA Review on Resource Discovery Strategies in Grid Computing
A Review on Resource Discovery Strategies in Grid Computing
 
B017240812
B017240812B017240812
B017240812
 
r.elativity Capabilities Summary
r.elativity Capabilities Summaryr.elativity Capabilities Summary
r.elativity Capabilities Summary
 
De Nigris
De Nigris De Nigris
De Nigris
 
HNSciCloud Overview
HNSciCloud Overview HNSciCloud Overview
HNSciCloud Overview
 
Camera ready-nash equilibrium-ngct2015-format
Camera ready-nash equilibrium-ngct2015-formatCamera ready-nash equilibrium-ngct2015-format
Camera ready-nash equilibrium-ngct2015-format
 
Accumulo Summit 2016: Accumulo Indexing Strategies for Searching Semantic Net...
Accumulo Summit 2016: Accumulo Indexing Strategies for Searching Semantic Net...Accumulo Summit 2016: Accumulo Indexing Strategies for Searching Semantic Net...
Accumulo Summit 2016: Accumulo Indexing Strategies for Searching Semantic Net...
 
Session19 Globus
Session19 GlobusSession19 Globus
Session19 Globus
 
Iss 6
Iss 6Iss 6
Iss 6
 
Optimization of resource allocation in computational grids
Optimization of resource allocation in computational gridsOptimization of resource allocation in computational grids
Optimization of resource allocation in computational grids
 
HNSciCloud: Project Results and lessons learned
HNSciCloud: Project Results and lessons learnedHNSciCloud: Project Results and lessons learned
HNSciCloud: Project Results and lessons learned
 
Event driven optimizer an approach to optimize the events that transition fr...
Event driven optimizer  an approach to optimize the events that transition fr...Event driven optimizer  an approach to optimize the events that transition fr...
Event driven optimizer an approach to optimize the events that transition fr...
 

Último

Call Girls In DLf Gurgaon ➥99902@11544 ( Best price)100% Genuine Escort In 24...
Call Girls In DLf Gurgaon ➥99902@11544 ( Best price)100% Genuine Escort In 24...Call Girls In DLf Gurgaon ➥99902@11544 ( Best price)100% Genuine Escort In 24...
Call Girls In DLf Gurgaon ➥99902@11544 ( Best price)100% Genuine Escort In 24...
lizamodels9
 
Insurers' journeys to build a mastery in the IoT usage
Insurers' journeys to build a mastery in the IoT usageInsurers' journeys to build a mastery in the IoT usage
Insurers' journeys to build a mastery in the IoT usage
Matteo Carbone
 
Call Girls Electronic City Just Call 👗 7737669865 👗 Top Class Call Girl Servi...
Call Girls Electronic City Just Call 👗 7737669865 👗 Top Class Call Girl Servi...Call Girls Electronic City Just Call 👗 7737669865 👗 Top Class Call Girl Servi...
Call Girls Electronic City Just Call 👗 7737669865 👗 Top Class Call Girl Servi...
amitlee9823
 
Call Girls Navi Mumbai Just Call 9907093804 Top Class Call Girl Service Avail...
Call Girls Navi Mumbai Just Call 9907093804 Top Class Call Girl Service Avail...Call Girls Navi Mumbai Just Call 9907093804 Top Class Call Girl Service Avail...
Call Girls Navi Mumbai Just Call 9907093804 Top Class Call Girl Service Avail...
Dipal Arora
 
Call Girls Jp Nagar Just Call 👗 7737669865 👗 Top Class Call Girl Service Bang...
Call Girls Jp Nagar Just Call 👗 7737669865 👗 Top Class Call Girl Service Bang...Call Girls Jp Nagar Just Call 👗 7737669865 👗 Top Class Call Girl Service Bang...
Call Girls Jp Nagar Just Call 👗 7737669865 👗 Top Class Call Girl Service Bang...
amitlee9823
 
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
Sheetaleventcompany
 

Último (20)

Call Girls In DLf Gurgaon ➥99902@11544 ( Best price)100% Genuine Escort In 24...
Call Girls In DLf Gurgaon ➥99902@11544 ( Best price)100% Genuine Escort In 24...Call Girls In DLf Gurgaon ➥99902@11544 ( Best price)100% Genuine Escort In 24...
Call Girls In DLf Gurgaon ➥99902@11544 ( Best price)100% Genuine Escort In 24...
 
Insurers' journeys to build a mastery in the IoT usage
Insurers' journeys to build a mastery in the IoT usageInsurers' journeys to build a mastery in the IoT usage
Insurers' journeys to build a mastery in the IoT usage
 
Forklift Operations: Safety through Cartoons
Forklift Operations: Safety through CartoonsForklift Operations: Safety through Cartoons
Forklift Operations: Safety through Cartoons
 
RSA Conference Exhibitor List 2024 - Exhibitors Data
RSA Conference Exhibitor List 2024 - Exhibitors DataRSA Conference Exhibitor List 2024 - Exhibitors Data
RSA Conference Exhibitor List 2024 - Exhibitors Data
 
Call Girls Electronic City Just Call 👗 7737669865 👗 Top Class Call Girl Servi...
Call Girls Electronic City Just Call 👗 7737669865 👗 Top Class Call Girl Servi...Call Girls Electronic City Just Call 👗 7737669865 👗 Top Class Call Girl Servi...
Call Girls Electronic City Just Call 👗 7737669865 👗 Top Class Call Girl Servi...
 
Call Girls Ludhiana Just Call 98765-12871 Top Class Call Girl Service Available
Call Girls Ludhiana Just Call 98765-12871 Top Class Call Girl Service AvailableCall Girls Ludhiana Just Call 98765-12871 Top Class Call Girl Service Available
Call Girls Ludhiana Just Call 98765-12871 Top Class Call Girl Service Available
 
Business Model Canvas (BMC)- A new venture concept
Business Model Canvas (BMC)-  A new venture conceptBusiness Model Canvas (BMC)-  A new venture concept
Business Model Canvas (BMC)- A new venture concept
 
Call Girls Navi Mumbai Just Call 9907093804 Top Class Call Girl Service Avail...
Call Girls Navi Mumbai Just Call 9907093804 Top Class Call Girl Service Avail...Call Girls Navi Mumbai Just Call 9907093804 Top Class Call Girl Service Avail...
Call Girls Navi Mumbai Just Call 9907093804 Top Class Call Girl Service Avail...
 
MONA 98765-12871 CALL GIRLS IN LUDHIANA LUDHIANA CALL GIRL
MONA 98765-12871 CALL GIRLS IN LUDHIANA LUDHIANA CALL GIRLMONA 98765-12871 CALL GIRLS IN LUDHIANA LUDHIANA CALL GIRL
MONA 98765-12871 CALL GIRLS IN LUDHIANA LUDHIANA CALL GIRL
 
Uneak White's Personal Brand Exploration Presentation
Uneak White's Personal Brand Exploration PresentationUneak White's Personal Brand Exploration Presentation
Uneak White's Personal Brand Exploration Presentation
 
Mysore Call Girls 8617370543 WhatsApp Number 24x7 Best Services
Mysore Call Girls 8617370543 WhatsApp Number 24x7 Best ServicesMysore Call Girls 8617370543 WhatsApp Number 24x7 Best Services
Mysore Call Girls 8617370543 WhatsApp Number 24x7 Best Services
 
John Halpern sued for sexual assault.pdf
John Halpern sued for sexual assault.pdfJohn Halpern sued for sexual assault.pdf
John Halpern sued for sexual assault.pdf
 
Value Proposition canvas- Customer needs and pains
Value Proposition canvas- Customer needs and painsValue Proposition canvas- Customer needs and pains
Value Proposition canvas- Customer needs and pains
 
Call Girls Service In Old Town Dubai ((0551707352)) Old Town Dubai Call Girl ...
Call Girls Service In Old Town Dubai ((0551707352)) Old Town Dubai Call Girl ...Call Girls Service In Old Town Dubai ((0551707352)) Old Town Dubai Call Girl ...
Call Girls Service In Old Town Dubai ((0551707352)) Old Town Dubai Call Girl ...
 
Falcon's Invoice Discounting: Your Path to Prosperity
Falcon's Invoice Discounting: Your Path to ProsperityFalcon's Invoice Discounting: Your Path to Prosperity
Falcon's Invoice Discounting: Your Path to Prosperity
 
The Path to Product Excellence: Avoiding Common Pitfalls and Enhancing Commun...
The Path to Product Excellence: Avoiding Common Pitfalls and Enhancing Commun...The Path to Product Excellence: Avoiding Common Pitfalls and Enhancing Commun...
The Path to Product Excellence: Avoiding Common Pitfalls and Enhancing Commun...
 
B.COM Unit – 4 ( CORPORATE SOCIAL RESPONSIBILITY ( CSR ).pptx
B.COM Unit – 4 ( CORPORATE SOCIAL RESPONSIBILITY ( CSR ).pptxB.COM Unit – 4 ( CORPORATE SOCIAL RESPONSIBILITY ( CSR ).pptx
B.COM Unit – 4 ( CORPORATE SOCIAL RESPONSIBILITY ( CSR ).pptx
 
Cracking the Cultural Competence Code.pptx
Cracking the Cultural Competence Code.pptxCracking the Cultural Competence Code.pptx
Cracking the Cultural Competence Code.pptx
 
Call Girls Jp Nagar Just Call 👗 7737669865 👗 Top Class Call Girl Service Bang...
Call Girls Jp Nagar Just Call 👗 7737669865 👗 Top Class Call Girl Service Bang...Call Girls Jp Nagar Just Call 👗 7737669865 👗 Top Class Call Girl Service Bang...
Call Girls Jp Nagar Just Call 👗 7737669865 👗 Top Class Call Girl Service Bang...
 
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
 

CDA for Grid and Cloud

  • 1. Double Auction-based Scheduling of Scientific Applications in Distributed Grid and Cloud Environments Francesc Lordan Gomis Metodologia de Recerca en la Informàtica 25/06/2012
  • 2. Outline Introduction Auction Theory Model and strategies Results Conclusions Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 1
  • 3. Introduction Grid and Cloud Computing allow user to execute applications on a third-party infrastructure by paying for the access to the remote resources Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 2
  • 4. Introduction Users are represented by Schedulers that picks which resource a job will be run The providers are represented by Resource Managers that try to maximize the incomes of the provider by optimizing the resource usage Resource Resource Manager Manager Scheduler Resource Resource Manager Manager Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 3
  • 5. Introduction When many users try to obtain the same resources, they compete for them. Since access to these resources is paid, it can be seen as a market governed by the supply-demand law Economy-based negotiation techniques are attractive from bussiness point of view Scheduler Resource Manager Scheduler Scheduler Market Resource ? Manager Scheduler Scheduler Resource Manager Scheduler Scheduler Resource Manager Scheduler Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 4
  • 6. Auction Theory Auctions are an economic mechanism used for allocating a set of resources among a group of bidders. Models: Single-sided auctions: the seller submits a resource and the bidders submit their offers to buy (bid) - English auction: buyers constantly increase their offers until the highest one wins - Dutch auction: the seller decreases the price of the resource until a buyer decides to pay the proposed price - Sealed auctions: buyers give their bids on an envelope First-price: the highest bid wins Vickrey second-price: the second highest bid wins Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 5
  • 7. Auction Theory Auctions are an economic mechanism used for allocating a set of resources among a group of bidders. Models: Double-sided auctions: all the participant submit their offers to sell (asks) and buy (bids) - Call Auctions: during a timeframe all the participants submit their offers and finally they try to match - Continuous Double Auction (CDA): offers are continously submitted until a match between the lowest ask and the highest bid is achieved or the auction times out. Combinatorial auctions: a single bid for multiple resource (NP-complete) Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 6
  • 8. Model and strategies Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 7
  • 9. Model and strategies The limitation on resources converts the schedulers into competitors with their own negotiation strategy. Any strategy is based on 4 factors Resource selection Resource valuation Auction participation Bidding Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 8
  • 10. Model and strategies Resource Selection Before bidding, the scheduler decides for which resources it will participate in an auction. When a task is sumitted the scheduler looks for a pending auction. If there is no such auction the scheduler postpones the bidding until there is one. The selection is done with a lottery where probabilities take into account: Availability of the resource Execution time Budget cost Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 9
  • 11. Model and strategies Resource Valuation User Valuation 2 steps: - Update the value of the resource depending on the final price of its last auction and a smoothing factor α. - Increases the value depending on the number of lost auctions and a revaluation factor β. Provider Valuation Depending on the success ratio on the last auctions for the resource the price is increased or decreased by a resource revaluation factor γ Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 10
  • 12. Model and strategies Auction participation Scheduler participates when there is at least one task waiting for that resource. Tasks are kept in a priority queue sorted by the bottom levels Resource Manager The time between two auctions of the same resource are created depend on the result for the last auctions. If it was won, it wait until the resource has been used and adds a delay If it was lost, the delay is incremented exponentially with a delaying factor δ Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 11
  • 13. Model and strategies Bidding Zero Intelligence: participants submit their offers periodically Scheduler tries to reduce the value of the resource by decreasing the distance between its valuation and the highest bid Resource Manager tries to increase the value of the resource by reducing the distance between its valuation and the lowest ask Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 12
  • 14. Model and strategies Modifications to the base strategy: Self-limitation Reduce the price by reducing the competence limit the number of auctions where the scheduler participates in based on: The average number of retries A resource limitation factor ζ Pricing Agressiveness Change the revaluation factor of a single scheduler Value of the resource are increased faster or slower Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 13
  • 15. Results Resource Valuation Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 14
  • 16. Results Resource Valuation Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 15
  • 17. Results Self-limitation Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 16
  • 18. Results Pricing aggressiveness Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 1 17
  • 19. Conclusions Creation of an open market of resources implemented as a CDA instance Identification the general behaviour patters Evaluation of strategies: Self-limitation: - has marginal improvements on specific configuration - NOT RECOMENDED Concessive strategy: - very good for the common-drift case Aggressive-strategy: - reduces execution time with low impact on the budget on the single-user deviation case - higher budget and similar execution time on the common-drift case - RISKY Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 18
  • 20. Questions Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment 19
  • 21. Resource Selection availability Expected execution time and cost at any resource Expected execution time Resource value 0 <= rsfmin <=rsfmax<=1 Resource selection ratio Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 22. Resource Valuation Scheduler Smoothing factor Last final price Last valuation Linear Exponential Resource revaluation factor # lost auctions for the resource Resource Manager Average success ratio Resource revaluation factor Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 23. Auction Participation Delay factor # lost auctions for the resource Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 24. Self-limitation Resource limitation factor Average retry number Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 25. Workflows Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 26. Pricing aggressiveness Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 27. Pricing aggressiveness Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 28. Pricing aggressiveness Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 29. Pricing aggressiveness Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 30. Pricing aggressiveness Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 31. Pricing aggressiveness Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment
  • 32. Pricing aggressiveness Double Auction-based Scheduling of Scientific applications in Distributed Grid and Cloud Environment