Refining Reliability Estimation of Mobile Software Systems

1
Refining Reliability Estimation of Mobile
Software Systems
The International Workshop on Software
Architectures and Mobility, ICSE-SAM 2008,
Leipzig, Germany
Sam Malek Department of Computer Science George
Mason University
Roshanak Roshandel Department of Computer
Science Software Engineering Seattle University
2
Motivation

• Programming-in-the-small-and-many (Prism)
• Challenges
• The role of software architecture in building
  dependable systems
• Structure
• Behavior
• Style
• Tools

3
Architectural Reliability Modeling

• Focus
• Predictive model for software quality
• Quantitative analysis of architectural design
  decisions
• Our past work
• Reliability prediction of software components
• Reliability prediction of software systems
• Challenges
• Uncertainties
• Information sources
• What about connectors?
• Connectors are not specifically addressed in
  reliability approaches

4
Mobility A Different Set of Challenges

• Dynamic Structure
• Dynamic Interactions and Interaction Protocols
• Network Connectivity Issues

5
A Proposal

• We have a static reliability prediction of the
  system
• Components reliabilities
• A model of components interactions in one
  specific structural/behavioral setting
• A prediction of the systems reliability in this
  setting
• We want
• Runtime monitoring of the systems interactions
  at runtime and/or data obtained from prototype
  simulation information source
• Use an architectural reliability modeling
  approach to refine the above reliability
  prediction

6
Component Reliability Prediction
L. Cheung, R. Roshandel, N. Medvidovic, L.
Golubchik, Early Prediction of Software Component
Reliability, ICSE 2008
7
Static System Reliability Prediction
R. Roshandel, N. Medvidovic, L. Golubchik, A
Bayesian Model for Predicting Reliability of
Software Systems at the Architectural Level, QoSA
2007.
8
The New Piece

• Runtime Monitoring of prototype and/or system
• Explicitly identifying connectors and
  interactions among components and connectors
• Gather data
• Refine the reliability model and parameterize it
  for connectors
• Apply coupling, cohesion, clustering, centrality
  and other graph-based algorithms to model
  connections
• Apply an architectural reliability methodology to
  obtain a more accurate reliability analysis

9
Reflection and Monitoring Prism-MW
10
Putting it all together
11
Questions?