Architectural Support for Model-driven Performance Prediction of Distributed Real-time Embedded Systems of Systems

Architectural Support for Model-driven Performance Prediction of Distributed Real-time Embedded Systems of Systems

Vanea CHIPRIANOVKatrina FALKNER

Claudia SZABOGavin PUDDY

Agenda

University of Adelaide 2

Challenges in predicting the performance of Distributed Real-time Embedded (DRE) Systems of Systems (SoS)

Requirements for an architectural framework

Architectural framework

Conclusions and perspectives

Challenges in predicting the performance of DRE SoS

University of Adelaide 3

SoS Operational independence

Different levels of performance for component systems

Managerial independence

Different organisation agendas

Evolutionary behaviour

No complete specification at design time

Emergent behaviour

How does it impact performance ? Where are the solutions needed ?

Geographic distribution

Numerous and long paths of interaction

Challenges in predicting the performance of DRE SoS

University of Adelaide 4

DRE Hard constraints on non-functional properties to meet the

requirements of space, weight, power, etc.

Usually fielded, extensively tested systems => important assests to be reused

=> Explore the expected performance through the investigation of several alternatives to system architecture

Requirements for an architectural framework

University of Adelaide 5

1) Loose coupling

2) Interoperability of composing systems

3) Interaction specification

4) Time and data distribution

5) Adaptability

6) Sustainable evolution

7) User interaction

Architectural framework

University of Adelaide 6

Event Driven Architecture (EDA) Event generator

Event sink

Event processing engine

Event channel

Event triggering, notification propagation & publishing

Architectural framework

University of Adelaide 7

Architectural framework

University of Adelaide 8

Software Architecture for Performance Prediction of DRE Standalone Systems

System Execution Modelling (SEM)

Measurement-based performance prediction Based on models of resource consumption Enable prediction of performance through execution of

representative source code of behaviour and workload models

Architectural framework

University of Adelaide 9

Software Architecture for Performance Prediction of DRE Standalone Systems

Falkner, K., Chiprianov, V., Falkner, N., Szabo, C., Hill, J., Puddy, G., Fraser, D., Johnston, A., Rieckmann, M., Wallis, A.: Model-drivenperformance prediction of distributed real-time embedded defence systems. In: The 18th Intl Conf. On Engineering of Complex ComputerSystems, Singapore (2013) 155–158

Architectural framework

University of Adelaide 10

Scenario Domain Specific Modelling Language (DSML)

Describe factors that influence the behaviour of the SoS

(Describe dynamics of interactions between composing systems)

Contains concepts of Event Driven Architecture

=> generic w.r.t. composing systems models

Model Driven Engineering (MDE) and code generation for adaptability

Architectural framework

University of Adelaide 11

Scenario DSML

Conformity of the Architectural framework to Requirements

University of Adelaide 12

Conclusions and Perspectives

University of Adelaide 13

Conclusions Identified and analysed challenges to modelling and

predicting performance of DRE SoS

Proposed architectural framework based on Model Driven Engineering, DSML, code generation

Event Driven Architecture Scenario DSML

Perspectives Integrate other non-functional properties (e.g. security,

fault tolerance)