SlideShare a Scribd company logo

Ismis2014 dbaas expert

•
•
Rim Moussa
Rim Moussa

The document presents DBaaS-Expert, a recommender system for selecting the appropriate cloud database from various database-as-a-service (DBaaS) providers. It uses a multi-criteria decision making approach, specifically the Analytic Hierarchy Process (AHP), to score and rank DBaaS offerings based on criteria such as performance, availability, security, and cost. The framework includes a DBaaS ontology for describing offerings and the AHP methodology to obtain criteria weights from user preferences and assess each DBaaS based on the weighted criteria. The goal is to maximize quality and capacity while minimizing cost. Future work includes evaluating the framework and incorporating user feedback into the DBaaS ranking.

1 of 16
Download to read offline
DBaaS-Expert: A Recommender for the Selection of the Right Cloud Database presented by Alfredo Cuzzocrea 21st Intl. Symposium on Methodologies for Intelligent Systems Roskilde, Denmark Soror Sahri*, Rim Moussa‡, Darrel D.E. Long†, Salima Benbernou* * {soror.sahri,salima.benbernou}@parisdescartes.fr, Univ. Paris Descartes, France ‡ rim.moussa@esti.rnu.tn, LaTICE Univ. of Carthage, Tunisia † darrel@cs.ucsc.edu, Storage Systems Research Center, Univ. of California , USA 26th, June 2014
26th, June 2014 ISMIS’2014@Roskilde 3 Context DBaaS Providers (non exhautive list of providers ) Cloud Rationale Which DBaaS to choose?
26th, June 2014 ISMIS’2014@Roskilde 4 ●DBaaS ? »An on-demand, secure, and scalable self-service database platform that automates provisioning and administration of databases (Forrester 2012) »DBaaS Examples: Amazon Relational Database Service (Amazon RDS), Microsoft SQL Azure, Heroku PostgreSQL as a service, Amazon Dynamo DB, Google BigQuery, ... ●Why DBaaS ? »Might be a suitable solution for companies for which in-house database solutions are cost-prohibitive »Cost relates to technology licenses' purchase, administration expertise, hardware purchase, hardware maintenance, ... »Cloud advantages: »Cost management: all services are provided on a pay-per-use basis, »Improved quality of service (assuming that providers hire experts) »Resources elasticity: auto-provisioning for scalability through fast new nodes' deployment and fast release of non-needed nodes DBaaS
26th, June 2014 ISMIS’2014@Roskilde 5 DBaaS adoption Source: Next-Generation Operational Databases: 2012-2016 https://451research.com/report­long?icid=2852 Conducted by 451Research, 2013
26th, June 2014 ISMIS’2014@Roskilde 6 ●Motivation: »Increasing number of DBaaSs offerings with different services and cost plans, »It's hard to compare offers, »through check of an advertisement-oriented documentation »through a common benchmark, Indeed, offers target different data management requirements (OLAP, OLTP, document-oriented, ...) ●DBaaSs' offerings ranking problem is a typical Muti-Criteria Decision Making (MCDM) problem. Indeed, given »M DBaaS offerings: DBaaS1 , DBaaS2 , ..., DBaaSM »e.g.: Amazon RDS, Google BigQuery, ... »N Decision criteria: C1 , C2 , ...,CN »e.g.: Performance, High-availability, Security, Elasticity, ... »Our goal is to assess DBaaSs' offerings in terms of the set of criteria with two objectives: »Maximize Quality and Capacity of Service »Minimize Cost Problem Statement
26th, June 2014 7 DBaaS-Expert Framework ISMIS’2014@Roskilde
26th, June 2014 ISMIS’2014@Roskilde 8 DBaaS Ontology The concepts of DBaaS ontology are divided into four categories: ● Basic concepts:Basic concepts: DBaaS offer, Cloud Service Provider, Workload Type, Storage Model, Data Model, Consistency Model, System Constraints, Resource, Trial Version ● Quality of service concepts:Quality of service concepts: SLA, Client Support ● Capacity of service concepts:Capacity of service concepts: High Availability, Security, Elasticity, Scalability, Interoperability ● Cost of service concepts:Cost of service concepts: Cost Model
26th, June 2014 ISMIS’2014@Roskilde 9 DBaaS Ontology
26th, June 2014 ISMIS’2014@Roskilde 10 DBaaS Ontology Windows Azure SQL Database ontologyWindows Azure SQL Database ontology
26th, June 2014 ISMIS’2014@Roskilde 11 AHP for DBaaS Ranking ●Analytic Hierarchy Process (AHP): »Developed by Thomas L. Saaty in 1970, »Structured technique for complex decisions making, based on mathematics and user-preferences, »Well-known and extensively used in problems of priority setting, university faculty members selection (university of Pennsylvania), quality of software systems quantification (MicroSoft). ●The outline of the solution of using AHP for DBaaS ranking is summarized below, »Devise the AHP Tree ● Decision goal at the root of the AHP Tree ● Hierarchy of criteria at internal nodes ● Alternatives (DBaaS offerings) at leafs »Compute Criteria Weights according to user preferences »For each criterion, assess DBaaS offers »Compute the score of each DBaaS offer
26th, June 2014 ISMIS’2014@Roskilde 12 AHP for DBaaSs' Ranking ---AHP Tree Decision Goal Hierarchy of Criteria Criterion DBaaS Offers
26th, June 2014 ISMIS’2014@Roskilde 13 »Using Pairwise Comparisons for criteria belonging to same hierarchy and same level. The global weight of a criterion is the product of the weights of its parent-criteria. »The process includes the check of the consistency of weights, and obliges the user to update initial weights in case of inconsistency. ●Example: AHP for DBaaSs' Ranking ---Criteria Weighting Resulting Weights Initial Matrix All criteria belong to the same hierarchy, The user considers that, ● Criterion Ci is half important than Cj (inversely Cj is twice more important than Ci), ● Criterion Ci is 3 times more important than Ck, Cj is the most important Criterion (55.8%) followed by Ci (32%) then Ck (12.2%) Iterative Eigenvector calculus through successive matrix squaring and normalization Stop when  is negligible which denotes that eigenvector is the same than last iteration
26th, June 2014 ISMIS’2014@Roskilde 14 »For each criterion, an assessment matrix is proposed for comparing DBaaSs. It is also based on pairwise comparisons. »The Decision matrix allows the calculus of the score of each DBaaS. AHP for DBaaSs' Ranking ---Scoring DBaaSs
26th, June 2014 ISMIS’2014@Roskilde 15 Related Work ●Comparison through Benchmarking: »OLTP-bench for the cloud: by Curino et al., proposal of new requirements and metrics for running an OLTP workload in the cloud, 2012 »Numerous benchmarks exist for different needs: Terasoft for sort of 1TB, ... ●Comparison using Recommenders, »CloudRecommender for infrastructure services (IaaS) selection: by Zhang et al. 2012 »Cloud Genius framework, also for IaaS services selection: by Menzel et al. 2012 »SMI Cloud: for measuring quality of Cloud Service Providers based on QoS attributes, by Gark et al. 2011 »Different Services selection from multiple cloud providers: by Quinton et al. 2013
26th, June 2014 16 Conclusion & Future Work ISMIS’2014@Roskilde ● Contributions:Contributions: proposing DBaaS-Expert framework, which allows a user to choose the most suitable DBaaS. ● a DBaaS ontology that aims at description of DBaaS offerings. ● application of AHP to DBaaSs' scoring in terms of a hierarchy of criteria. ● Future work:Future work: ● Evaluation of DBaaS-Expert ● take into account experiences and feedbacks in the ranking of DBaaS offerings.
Thank you for Your Attention Q & A DBaaS-Expert: A Recommender for the Selection of the Right Cloud Database ISMIS'2014@Roskilde 26th June, 2014

More Related Content

Ismis2014 dbaas expert

  • 1. DBaaS-Expert: A Recommender for the Selection of the Right Cloud Database presented by Alfredo Cuzzocrea 21st Intl. Symposium on Methodologies for Intelligent Systems Roskilde, Denmark Soror Sahri*, Rim Moussa‡, Darrel D.E. Long†, Salima Benbernou* * {soror.sahri,salima.benbernou}@parisdescartes.fr, Univ. Paris Descartes, France ‡ rim.moussa@esti.rnu.tn, LaTICE Univ. of Carthage, Tunisia † darrel@cs.ucsc.edu, Storage Systems Research Center, Univ. of California , USA 26th, June 2014
  • 2. 26th, June 2014 ISMIS’2014@Roskilde 3 Context DBaaS Providers (non exhautive list of providers ) Cloud Rationale Which DBaaS to choose?
  • 3. 26th, June 2014 ISMIS’2014@Roskilde 4 ●DBaaS ? »An on-demand, secure, and scalable self-service database platform that automates provisioning and administration of databases (Forrester 2012) »DBaaS Examples: Amazon Relational Database Service (Amazon RDS), Microsoft SQL Azure, Heroku PostgreSQL as a service, Amazon Dynamo DB, Google BigQuery, ... ●Why DBaaS ? »Might be a suitable solution for companies for which in-house database solutions are cost-prohibitive »Cost relates to technology licenses' purchase, administration expertise, hardware purchase, hardware maintenance, ... »Cloud advantages: »Cost management: all services are provided on a pay-per-use basis, »Improved quality of service (assuming that providers hire experts) »Resources elasticity: auto-provisioning for scalability through fast new nodes' deployment and fast release of non-needed nodes DBaaS
  • 4. 26th, June 2014 ISMIS’2014@Roskilde 5 DBaaS adoption Source: Next-Generation Operational Databases: 2012-2016 https://451research.com/report­long?icid=2852 Conducted by 451Research, 2013
  • 5. 26th, June 2014 ISMIS’2014@Roskilde 6 ●Motivation: »Increasing number of DBaaSs offerings with different services and cost plans, »It's hard to compare offers, »through check of an advertisement-oriented documentation »through a common benchmark, Indeed, offers target different data management requirements (OLAP, OLTP, document-oriented, ...) ●DBaaSs' offerings ranking problem is a typical Muti-Criteria Decision Making (MCDM) problem. Indeed, given »M DBaaS offerings: DBaaS1 , DBaaS2 , ..., DBaaSM »e.g.: Amazon RDS, Google BigQuery, ... »N Decision criteria: C1 , C2 , ...,CN »e.g.: Performance, High-availability, Security, Elasticity, ... »Our goal is to assess DBaaSs' offerings in terms of the set of criteria with two objectives: »Maximize Quality and Capacity of Service »Minimize Cost Problem Statement
  • 6. 26th, June 2014 7 DBaaS-Expert Framework ISMIS’2014@Roskilde
  • 7. 26th, June 2014 ISMIS’2014@Roskilde 8 DBaaS Ontology The concepts of DBaaS ontology are divided into four categories: ● Basic concepts:Basic concepts: DBaaS offer, Cloud Service Provider, Workload Type, Storage Model, Data Model, Consistency Model, System Constraints, Resource, Trial Version ● Quality of service concepts:Quality of service concepts: SLA, Client Support ● Capacity of service concepts:Capacity of service concepts: High Availability, Security, Elasticity, Scalability, Interoperability ● Cost of service concepts:Cost of service concepts: Cost Model
  • 8. 26th, June 2014 ISMIS’2014@Roskilde 9 DBaaS Ontology
  • 9. 26th, June 2014 ISMIS’2014@Roskilde 10 DBaaS Ontology Windows Azure SQL Database ontologyWindows Azure SQL Database ontology
  • 10. 26th, June 2014 ISMIS’2014@Roskilde 11 AHP for DBaaS Ranking ●Analytic Hierarchy Process (AHP): »Developed by Thomas L. Saaty in 1970, »Structured technique for complex decisions making, based on mathematics and user-preferences, »Well-known and extensively used in problems of priority setting, university faculty members selection (university of Pennsylvania), quality of software systems quantification (MicroSoft). ●The outline of the solution of using AHP for DBaaS ranking is summarized below, »Devise the AHP Tree ● Decision goal at the root of the AHP Tree ● Hierarchy of criteria at internal nodes ● Alternatives (DBaaS offerings) at leafs »Compute Criteria Weights according to user preferences »For each criterion, assess DBaaS offers »Compute the score of each DBaaS offer
  • 11. 26th, June 2014 ISMIS’2014@Roskilde 12 AHP for DBaaSs' Ranking ---AHP Tree Decision Goal Hierarchy of Criteria Criterion DBaaS Offers
  • 12. 26th, June 2014 ISMIS’2014@Roskilde 13 »Using Pairwise Comparisons for criteria belonging to same hierarchy and same level. The global weight of a criterion is the product of the weights of its parent-criteria. »The process includes the check of the consistency of weights, and obliges the user to update initial weights in case of inconsistency. ●Example: AHP for DBaaSs' Ranking ---Criteria Weighting Resulting Weights Initial Matrix All criteria belong to the same hierarchy, The user considers that, ● Criterion Ci is half important than Cj (inversely Cj is twice more important than Ci), ● Criterion Ci is 3 times more important than Ck, Cj is the most important Criterion (55.8%) followed by Ci (32%) then Ck (12.2%) Iterative Eigenvector calculus through successive matrix squaring and normalization Stop when  is negligible which denotes that eigenvector is the same than last iteration
  • 13. 26th, June 2014 ISMIS’2014@Roskilde 14 »For each criterion, an assessment matrix is proposed for comparing DBaaSs. It is also based on pairwise comparisons. »The Decision matrix allows the calculus of the score of each DBaaS. AHP for DBaaSs' Ranking ---Scoring DBaaSs
  • 14. 26th, June 2014 ISMIS’2014@Roskilde 15 Related Work ●Comparison through Benchmarking: »OLTP-bench for the cloud: by Curino et al., proposal of new requirements and metrics for running an OLTP workload in the cloud, 2012 »Numerous benchmarks exist for different needs: Terasoft for sort of 1TB, ... ●Comparison using Recommenders, »CloudRecommender for infrastructure services (IaaS) selection: by Zhang et al. 2012 »Cloud Genius framework, also for IaaS services selection: by Menzel et al. 2012 »SMI Cloud: for measuring quality of Cloud Service Providers based on QoS attributes, by Gark et al. 2011 »Different Services selection from multiple cloud providers: by Quinton et al. 2013
  • 15. 26th, June 2014 16 Conclusion & Future Work ISMIS’2014@Roskilde ● Contributions:Contributions: proposing DBaaS-Expert framework, which allows a user to choose the most suitable DBaaS. ● a DBaaS ontology that aims at description of DBaaS offerings. ● application of AHP to DBaaSs' scoring in terms of a hierarchy of criteria. ● Future work:Future work: ● Evaluation of DBaaS-Expert ● take into account experiences and feedbacks in the ranking of DBaaS offerings.
  • 16. Thank you for Your Attention Q & A DBaaS-Expert: A Recommender for the Selection of the Right Cloud Database ISMIS'2014@Roskilde 26th June, 2014