SLIDE 1
2
Presentation Overview
- 1. Objective for thesis
- 2. Convergence
- 3. S2Store Simulations
- 4. Evaluation
- 5. Questions and Answers
Oberseminar Convergence Mechanisms for a Smart Space App Store - - PowerPoint PPT Presentation
Oberseminar Convergence Mechanisms for a Smart Space App Store - - PowerPoint PPT Presentation
Lehrstuhl fr Netzarchitekturen und Netzdienste Institut fr Informatik Technische Universitt Mnchen Oberseminar Convergence Mechanisms for a Smart Space App Store Bibek Shrestha bibek.shrestha@tum.de Under supervision of Marc-Oliver
SLIDE 2
SLIDE 3
For Smart Space AppStore study how using Crowdsourcing Mechanisms bring Convergence
3
Objective
Services
SLIDE 4
For Smart Space AppStore study how using Crowdsourcing Mechanisms bring Convergence
S2Store
3
Objective
Services
SLIDE 5
For Smart Space AppStore study how using Crowdsourcing Mechanisms bring Convergence
S2Store
3
Objective
Services
SLIDE 6
For Smart Space AppStore study how using Crowdsourcing Mechanisms bring Convergence
S2Store
3
Objective
Services
SLIDE 7
For Smart Space AppStore study how using Crowdsourcing Mechanisms bring Convergence
S2Store
3
Objective
Services
Large and global user contributions cause: a) duplication b) good, average and low quality
SLIDE 8
For Smart Space AppStore study how using Crowdsourcing Mechanisms bring Convergence
S2Store
3
Objective
Services
Large and global user contributions cause: a) duplication b) good, average and low quality
SLIDE 9
4
S2Store Entities
a) Services b) Context Models c) Access Groups
a) Service Lamp Device Driver (Java) b) Context Model (XML)
SLIDE 10
4
S2Store Entities
a) Services b) Context Models c) Access Groups
a) Service Lamp Device Driver (Java) b) Context Model (XML) (c)
SLIDE 11
5
S2Store activities
Developer and User interaction with the S2Store and its Entities
SLIDE 12
6
S2Store Simulation
Initialize Ecosystem Developer agents build service, context model and access groups Update S2Store User agents browse and download services User provides feedback of the services System calculates rankings for all entities Increase agent population Exit loop for N timesteps
Inspired by AppEco Simulation from [LP12]
SLIDE 13
7
Service Convergence with Reputation System
1) Explicit Reputation Systems [FG10] 2) Implicit Reputation Systems [GM10] * Error reports * User action - install, update, uninstall
a) Services b) Context Models c) Access Groups
SLIDE 14
7
Service Convergence with Reputation System
1) Explicit Reputation Systems [FG10] 2) Implicit Reputation Systems [GM10] * Error reports * User action - install, update, uninstall
a) Services b) Context Models c) Access Groups
Pick random service Explicit and Implicit Reputation Feedback Observe Convergence N times Service Simulation
SLIDE 15
8
Context Model Convergence with Graph Simulation
Context Model (XML)
a) Services b) Context Models c) Access Groups
SLIDE 16
8
Context Model Convergence with Graph Simulation
Context Model (XML) Create context model Create or choose node dependencies Update S2Store Context Model Repository Update node ranking
N times
Context Model Simulation
a) Services b) Context Models c) Access Groups
SLIDE 17
9
Context Model Convergence with Graph Simulation
Node Ranking Algorithms
- 1. PageRank (Eigenvector Centrality)
- 2. In-Degree Centrality
Graph Properties for Convergence
- 1. Small World and Scale Free [LW04]
- 2. Disassortative [LW04]
- 3. Hierarchical distribution [Hal03]
a) Services b) Context Models c) Access Groups
SLIDE 18
10
Access Groups Convergence
Access Groups
Create context model Create or choose existing Access Groups Update S2Store
N times
Access Groups Simulation
a) Services b) Context Models c) Access Groups
Calculate ranking of Access Groups
SLIDE 19
11
Evaluation: Input
SLIDE 20
12
Evaluation: Expected Output
Node Distribution Node Importance
Beginning of Simulation End of Simulation 3/4th of the Simulation 2/4th of the Simulation Legend
SLIDE 21
13
Questions?
SLIDE 22
14