Applying Model Transformations to Optimizing Real-Time QoS Configurations in DRE Systems

  • Authors:

  • Amogh Kavimandan;Aniruddha Gokhale

  • Affiliations:

  • ISIS, Dept. of EECS, Vanderbilt University, Nashville, USA;ISIS, Dept. of EECS, Vanderbilt University, Nashville, USA

  • Venue:
  • QoSA '09 Proceedings of the 5th International Conference on the Quality of Software Architectures: Architectures for Adaptive Software Systems
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

The quality of a software architecture for component-based distributed systems is defined not just by its source code but also by other systemic artifacts, such as the assembly, deployment, and configuration of the application components and their component middleware. In the context of distributed, real-time, and embedded (DRE) component-based systems, bin packing algorithms and schedulability analysis have been used to make deployment and configuration decisions. However, these algorithms make only coarse-grained node assignments but do not indicate how components are allocated to different middleware containers on the node, which are known to impact runtime system performance and resource consumption. This paper presents a model transformation-based algorithm that combines user-specified quality of service (QoS) requirements with the node assignments to provide a finer level of granularity and precision in the deployment and configuration decisions. A beneficial side effect of our work lies in how these decisions can be leveraged by additional backend performance optimization techniques. We evaluate our approach and compare it against the existing state-of-the-art in the context of a representative DRE system.