Lecture 30 cloud mktplace
- 1. 1 Prof. Soumya K Ghosh Department of Computer Science and Engineering IIT KHARAGPUR 9/20/2017 Cloud Computing : Broker for Cloud Marketplace
- 2. INTRODUCTION 9/20/2017 2 • Rapid growth of available cloud services • Huge number of providers with varying QoS • Different types of customer use cases – each with different requirements
- 3. INTRODUCTION 9/20/2017 3 • Rapid growth of available cloud services • Huge number of providers with varying QoS • Different types of customer use cases – each with different requirements • Need for a “middle man” (Intelligent Broker!) to • Suggest the best cloud provider to the customer • Safeguard the interests of the customer
- 4. MOTIVATION • Flexible selection of cloud provider • Trustworthiness of provider • Monitoring of services • Avoiding vendor lock-in 9/20/2017 4
- 5. OBJECTIVES • Selection of the most suitable provider satisfying customer's QoS requirements • Calculation of the degree of SLA satisfaction and trustworthiness of a provider • Decision making system for dynamic service migration based on experienced QoS 9/20/2017 5
- 6. Different Approaches • CloudCmp: a tool that compares cloud providers in order to measure the QoS they offer and helps users to select a cloud. • Fuzzy provider selection mechanism. • Framework with a measure of satisfaction with a provider for keeping in mind the fuzzy nature of the user requirements. • Provider selection framework which takes into account the trustworthiness and competence of a provider. 9/20/2017 6
- 7. CUSTOMER QoS PARAMETERS 9/20/2017 7 • More QoS parameter can be added easily. Infrastructure-as-a-Service Software-as-a-Service
- 8. PROVIDER • Promised QoS values : • Trust values : Note: They have been kept independent as they pertain to different parameters 9/20/2017 8
- 10. PROVIDER SELECTION 9/20/2017 10 • Selection of provider is done using a fuzzy inference engine • Input : QoS offered by a provider and its trustworthiness • Output : Suitability of the provider for the customer • Customer request is dispatched to provider with maximum suitability • Membership functions are built using the user requirements
- 12. PROVIDER SELECTION – INPUT MEMBERSHIP FUNCTION 9/20/2017 12
- 13. PROVIDER SELECTION – INPUT MEMBERSHIP FUNCTION 9/20/2017 13
- 14. PROVIDER SELECTION – OUTPUT MEMBERSHIP FUNCTION 9/20/2017 14
- 15. MONITORING MODULE 9/20/2017 15 From provider/3rd party monitoring tool From Repository Performance for SIi in current monitoring period
- 16. MIGRATION DECIDER 9/20/2017 16 • Makes use of a fuzzy inference engine • Input : • Output : Degree of SLA Satisfaction for • If Degree of SLA Satisfaction < threshold, migrate
- 17. MIGRATION DECIDER – OUTPUT MEMBERSHIP FUNCTION 9/20/2017 17
- 18. MIGRATION MODULE - SELECTION OF TARGET PROVIDER 9/20/2017 18 • Similar to provider selection • Selection done using a fuzzy inference engine
- 19. Case study on IaaS Marketplace 9/20/2017 19 • 10 providers with varying offered QoS • 500 requests for VMs • Year long simulation • Few providers exhibit performance degradation. Degraded QoS parameters follow a Gaussian distribution • Comparison made with conventional (minimum cost) crisp broker
- 23. Case study on SaaS Marketplace 9/20/2017 23 • 10 providers with varying offered QoS • 500 service requests • Year long simulation • Few providers exhibit performance degradation. Degraded QoS parameters follow a Gaussian distribution • Comparison made with conventional (minimum cost) crisp broker
- 27. Future Scope 9/20/2017 27 • Specification of flexibility in QoS requirements • Comparison against existing approaches on production workload • Service classes for customers
- 28. 289/20/2017