Presentation is loading. Please wait.

Presentation is loading. Please wait.

A Framework for Composing Pervasive Applications Oleg Davidyuk, Ivan Sanchez, Jon Imanol Duran and Jukka Riekki Advances in Methods of Information and.

Similar presentations


Presentation on theme: "A Framework for Composing Pervasive Applications Oleg Davidyuk, Ivan Sanchez, Jon Imanol Duran and Jukka Riekki Advances in Methods of Information and."— Presentation transcript:

1 A Framework for Composing Pervasive Applications Oleg Davidyuk, Ivan Sanchez, Jon Imanol Duran and Jukka Riekki Advances in Methods of Information and Communication Technology (AMICT'08) Workshop University of Oulu, Finland

2 2 laptop plasma display speaker system projector web/database servers mobile phone The concept Application Resources User

3 3 Potential Application Scenarios Virtual Devices (or Resource Sharing) Load Distribution (Grids and Web services) Multimodal User Interfaces (speech, video kinetic, tactile) speechtexttouchvideo gesturesaudio From: http://press.web.cern.ch/press/PressReleases/Releases2003/Images/

4 4 Ubiquitous middleware Conceptual Architecture S1 S2 S3 S4 R2 R1 R3 Service discovery Resource Management Application Assembly Context providers Composed pervasive applications

5 5 Differences from the related work Related work – Selecting resources according to a goal (COCOA) – Applications with proprietary architecture (Gaia) – Semantically independent systems (AURA, COCOA) Our approach – Applications adapt their architecture – Support for applications regardless of their specific properties

6 6 Application and Platform Graphs Application Model Platform Model 30 nodes

7 7 Proposed Solution Conclusions from the previous work: – The search problem is uncorrelated – Larger graphs  higher failure ratios The algorithms: – Evolutionary (EA) and Genetic (GA) allocation algorithms – Hybrid problem handling (both CSP and OP) – Novel solution validation schema The objective function values Index Increasing

8 8 Comparison of the Algorithms Evolutionary Algorithm – Simple implementation – Relies on random mutation operator Genetic Algorithm – Complex (population handling, sorting, etc) – Uses guided genetic operators

9 9 Analysis Performance Quality Failure Ratio Most Stable The fastest Difference 5~12% Logarithmic scale

10 10 Practical Contribution Installation: Displays and media servers 3 application components RFid tag Remote UI

11 11 The scalability test 93% 91% 84% 79% Application setup time Number of resources in the enviroment Average time, ms

12 12 User experiments Resources: 8 displays, 3 media servers 10 users, STO’s students and research personnel 100% of them wanted an additional control over the algorithm’s choices: –70% wanted to confirm choices manually –30% wanted to receive additional notifications Users indicated usefulness of the approach, especially in public places –Many options to choose (many available resources) –Unfamiliar environment 80% were satisfied with the algorithm’s choices (20% expected different results) The usability was rated very high (9,5 out of 10 points)

13 13 Future work Increase the algorithm’s performance Implement the next application scenario –Modify user interfaces and provide functionality required by the users Study human-related aspects of application composition

14 14


Download ppt "A Framework for Composing Pervasive Applications Oleg Davidyuk, Ivan Sanchez, Jon Imanol Duran and Jukka Riekki Advances in Methods of Information and."

Similar presentations


Ads by Google