DSN 2004 INSTITUTO DE COMPUTAO INTERNATIONAL CONFERENCE ON - - PowerPoint PPT Presentation

dsn 2004
SMART_READER_LITE
LIVE PREVIEW

DSN 2004 INSTITUTO DE COMPUTAO INTERNATIONAL CONFERENCE ON - - PowerPoint PPT Presentation

DSN 2004 INSTITUTO DE COMPUTAO INTERNATIONAL CONFERENCE ON DEPENDABLE SYSTEMS Centro Superior de Educao Tecnolgica AND NETWORK JUNE/2004 Architecture-based Strategy INSTITUTO DE INSTITUTO DE COMPUTAO COMPUTAO for


slide-1
SLIDE 1

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

DSN 2004 INTERNATIONAL CONFERENCE ON DEPENDABLE SYSTEMS AND NETWORK

JUNE/2004

slide-2
SLIDE 2

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

Architecture-based Strategy for Interface Fault Injection

Authors: Eliane Martins Regina Lúcia de Oliveira Moraes State University of Campinas - UNICAMP Brazil

INSTITUTO DE COMPUTAÇÃO

slide-3
SLIDE 3

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

Contents

  • Fault Injection
  • Architectural View
  • The Strategy
  • Future Works
  • Jaca
  • Motivation
slide-4
SLIDE 4

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

Motivation

  • Systems as a combination of several

components

  • Helps to attend the increasing pressures to

reduce time and money

  • A good solution for system’s architecture

increases the system’s quality

slide-5
SLIDE 5

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

Motivation

  • Malfunctioning in the interaction among

components can compromise the overall system’s quality

  • Each component’s implementation needs to

behave in accordance with its specification

  • Fault Injection can be a valuable approach to

component-based system’s testing

slide-6
SLIDE 6

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

Fault Injection

  • Fault Injection

– faults (or errors) are deliberately introduced into a system – useful to validate error-handling mechanisms – useful to assess system behavior when its components fails

  • We use sw-implemented fault injection

–faults are introduced during runtime – faults represent failure modes of components (internal or external to the system)

slide-7
SLIDE 7

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

Jaca

  • Software Fault Injection Tool
  • Written in Java language
  • Uses computational reflection implemented by

Javassist

  • Does not need the source code of the system

under test

  • Can inject high level faults in an object-oriented

system written in Java language

slide-8
SLIDE 8

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

Problems to Solve

  • Which components to inject?
  • Where to inject the faults?
  • Which error models to use?
  • When to inject the faults?
  • How to inject them?
slide-9
SLIDE 9

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

Architectural View

IExample3 IExample4 IExample4N IExample3N IExample2 IExample1 Interface IExample1N Example5 Example1 Example2

Connector1 Connector2

Example3 Example4 Component Connector

slide-10
SLIDE 10

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

The Strategy

A Strategy based on Risk

  • Where to inject

1. Prioritize Components 2. Select Components 3. Select Operations 4. Generate Injection Points

Aimed to answer the following questions:

  • Which components to inject

New Component Changed Component Upstream Dependency Downstream Dependency Critical Popular Strategic Third Party Distributed Not Understandable

slide-11
SLIDE 11

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

The Strategy

  • Which error models to use
  • When to inject the faults

Other questions:

5. Define Error Model

  • Ballista’s Robustness

Testing

  • Integration Faults

6. Decide Temporal Characterization

  • Randomly selected

among permanent, transient and intermittent faults

  • How to inject

Jaca Tool

slide-12
SLIDE 12

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

Future Works

Open Problems

  • How many criteria must be satisfied?
  • How should these criteria be weighted?
  • How should successors and predecessors be

determined?

  • What error model should be selected?
  • How can good controlability and observability be

achieved?

  • How should each factor be quantified?
slide-13
SLIDE 13

INSTITUTO DE COMPUTAÇÃO

Centro Superior de Educação Tecnológica

Thank You for Coming!

Questions and Suggestions eliane@ic.unicamp.br regina@ceset.unicamp.br