Paper information
Title
Formalizing the Asynchronous Evolution of Architecture Patterns
Formalizing the Asynchronous Evolution of Architecture Patterns
Published in
In proc. of Workshop on Self-Organizing Architectures (SOAR09), held at the Working IEEE/IFIP Conference on Software Architecture (WICSA) and European Conference on Software Architecture (ECSA). Cambridge, UK, 14th September 2009. - 2009
In proc. of Workshop on Self-Organizing Architectures (SOAR09), held at the Working IEEE/IFIP Conference on Software Architecture (WICSA) and European Conference on Software Architecture (ECSA). Cambridge, UK, 14th September 2009. - 2009
Abstract
In order to reduce management and maintenance efforts, large software-intensive systems can be built of several instances of self-managed architectural elements. Each self-managed architectural instance has local autonomy to reconfigure its internal architecture, but only while satisfying an architecture pattern. However, this architecture pattern may need to be modified at runtime to introduce unanticipated changes. Due to the autonomous nature of each self-managed instance, runtime changes must be performed asynchronously: changes require a gradual propagation and execution, as soon as each instance is ready for evolution. This paper introduces the semantics of asynchronous evolution, which allows introducing unanticipated changes concurrently in a running system. That is, both types and instances evolve dynamically at different rates, but while preserving: (i) type-conformance of instances, and (ii) the order of different type evolutions. This is illustrated by describing how the architectural types of the PRISMA ADL (i.e. architecture patterns) are evolved asynchronously.
In order to reduce management and maintenance efforts, large software-intensive systems can be built of several instances of self-managed architectural elements. Each self-managed architectural instance has local autonomy to reconfigure its internal architecture, but only while satisfying an architecture pattern. However, this architecture pattern may need to be modified at runtime to introduce unanticipated changes. Due to the autonomous nature of each self-managed instance, runtime changes must be performed asynchronously: changes require a gradual propagation and execution, as soon as each instance is ready for evolution. This paper introduces the semantics of asynchronous evolution, which allows introducing unanticipated changes concurrently in a running system. That is, both types and instances evolve dynamically at different rates, but while preserving: (i) type-conformance of instances, and (ii) the order of different type evolutions. This is illustrated by describing how the architectural types of the PRISMA ADL (i.e. architecture patterns) are evolved asynchronously.
BibTeX
@misc{issi_web:id:329,
title = "Formalizing the Asynchronous Evolution of Architecture Patterns",
author = "Cristóbal Costa Soria and Reiko Heckel",
booktitle = "In proc. of Workshop on Self-Organizing Architectures (SOAR09), held at the Working IEEE/IFIP Conference on Software Architecture (WICSA) and European Conference on Software Architecture (ECSA). Cambridge, UK, 14th September 2009.",
year = "2009",
eprint = "http://issi.dsic.upv.es/publications/archives/",
url = "http://distrinet.cs.kuleuven.be/events/soar/2009/contents/papers/costa-soria-soar09.pdf",
abstract = "In order to reduce management and maintenance efforts, large software-intensive systems can be built of several instances of self-managed architectural elements. Each self-managed architectural instance has local autonomy to reconfigure its internal architecture, but only while satisfying an architecture pattern. However, this architecture pattern may need to be modified at runtime to introduce unanticipated changes. Due to the autonomous nature of each self-managed instance, runtime changes must be performed asynchronously: changes require a gradual propagation and execution, as soon as each instance is ready for evolution. This paper introduces the semantics of asynchronous evolution, which allows introducing unanticipated changes concurrently in a running system. That is, both types and instances evolve dynamically at different rates, but while preserving: (i) type-conformance of instances, and (ii) the order of different type evolutions. This is illustrated by describing how the architectural types of the PRISMA ADL (i.e. architecture patterns) are evolved asynchronously."
}