Presentation is loading. Please wait.

Presentation is loading. Please wait.

A Collaborative Quality Ranking Framework for Cloud Components

Similar presentations


Presentation on theme: "A Collaborative Quality Ranking Framework for Cloud Components"— Presentation transcript:

1 A Collaborative Quality Ranking Framework for Cloud Components
Zibin Zheng and Michael R. Lyu Department of Computer Science & Engineering The Chinese University of Hong Kong Hong Kong, China DSN 2010, Chicago, Illinois, USA, June 28 – July 01, 2010

2 Introduction Cloud application is composed by a number of distributed components. Some cloud components are reusable by other cloud applications. Influenced by the communication links, the component quality ranking cannot be reused directly to another user. Personalized component quality ranking is required.

3 Motivating Examples Personalized component quality ranking:
Select the optimal cloud component from a set of functionally equivalent component candidates. For the components within a cloud application, the application designer wants to identify the well performing components from user perspective.

4 Kendall Rank Correlation Coefficient
Kendall Rank Correlation Coefficient (KRCC) : subset of cloud components commonly invoked by user u and user v. : QoS value of component i observed by user u. Indicator function

5 User Preference User preference on two cloud components which have been invoked previously: User preference on a pair of components that has not been both invoked by the current user u:

6 Problem Given a preference function, we want to choose a ranking of components that agrees with the preferences as much as possible. Object function: Target: to produce a ranking that maximizes the objective function.

7 CloudRank Algorithm

8 Experiments WS-DREAM QoS dataset Evaluation matrix
100 service components and 150 different users. 1.5 million invocations. Evaluation matrix NDCG: Normalized Discounted Cumulative Gain Larger value stands for better ranking performance.

9 Conclusion and Future Work
Identify the need for component ranking in Cloud Computing Propose a personalized component quality ranking approach for cloud applications Conduct experiments using real-world QoS data Future work Investigate more QoS properties Perform further experiments in larger scale.

10 Thank you! WS-DREAM QoS Dataset:


Download ppt "A Collaborative Quality Ranking Framework for Cloud Components"

Similar presentations


Ads by Google