ArchitectureCentric Software Generation: an Experimental Study on Distributed Systems

1
Architecture-Centric Software Generation an
Experimental Study on Distributed Systems

• Chung-Horng Lung
• Kamalachelva Selvarajah
• Dept. of Systems Computer Eng.
• Carleton University
• Ottawa, Canada

Balasangar Balasubramaniam Poopalasinkam
Elankeswaran Umatharan Gopalasundaram Nortel
Networks Ottawa, Cananda
2
Agenda

• Motivations
• Development Process
• Generative Framework
• Use of the Framework
• Related Work
• Conclusions

3
Motivations

• Rapid development high quality (including
  performance) are crucial to enterprises.
• Software development time often runs over
  schedule.
• Distributed systems are complex and involve
  various areas.
• Software performance engineering (SPE) is
  challenging at the architecture level.
• Architects/designers may want to focus on the
  application.
• Design patterns in this area are well understood.
• Need a generative approach to support rapid
  development based on architectural
  alternatives/patterns.

4
Architecture-Centric Software Generation
Development Process

• Define the scope and conduct a variability
  analysis at the architecture level.
• Conduct design recovery of existing robust
  software systems.
• Reengineer existing systems using patterns.
• Conduct evaluation of architecture alternatives.
• Construct reusable components and the framework.

5
Variability Basic Networked and Concurrent
Architectural Patterns
select()
select()
network
network
(c) Leader/Followes
Single Thread
(b)Half-Sync/Half-Async
6
Design Recovery and Reengineering of Existing
Systems

• Study existing systems, including one developed
  by industry, to identify concrete similarities
  and differences.
• Reengineer existing systems with design patterns
  at different levels
• Structure ST-gt HS/HA, ST -gt LFs
• Others Monitor Object, Scoped Locking idiom,
  Reactor, Connector, and Acceptor

7
Architecture Evaluation

• Evaluate each reengineered system with focus on
  performance, including QoS.
• Performance issue is challenging
• Opposite results for different kernel versions
• Different conclusions from real measurements and
  performance modeling
• A working prototype for concept demonstration and
  providing concrete data could help better
  understand the system.

8
Construction of Reusable Components and the
Framework

• Develop reusable components based on design
  patterns and variability analysis at various
  levels in Java.
• Incorporate architectural alternatives into the
  framework.
• Develop a GUI to interface with the framework.

9
Structure and Reusable Components of the
Generative Framework
10
Usage GUI for Selecting an Architectural
Alternative
11
Related Work

• Domain analysis/engineering
• ACE framework
• Component-based software engineering (CBSE)
• CoSMIC
• MicroQoSCorba
• Ours architecture-centric has built-in synthesis

12
Conclusions

• Few tools are available to them for conducting
  front-end analysis.
• Evaluations of architectures are often performed
  on high-level descriptions concrete or realistic
  data may not be available.
• Generative framework allows rapid generation of
  working systems based on different architectures.
• Designers can focus on the application by adding
  specific services.
• Collect realistic performance data which can be
  used for architecture evaluations and further
  performance modeling and analysis.