Romain Rouvoy

Using constraint-based optimization and variability to support continuous self-adaptation

Self-adaptation is one of the upcoming paradigms that accurately tackles nowadays systems complexity. In this context, Dynamic Software Product Lines model the intrinsic variability of a family of systems, and dynamically support their reconfiguration according to updated context. However, when several configurations are available for the same context, making a decision about the right one is a hard challenge: further dimensions such as QoS are needed to enrich the decision making process. In this paper, we propose to combine variability with Constraint-Satisfaction Problem techniques to face this challenge. The approach is illustrated and validated with a context-driven system used to support the control of a home through mobile devices.

Hadoop-Benchmark: Rapid Prototyping and Evaluation of Self-Adaptive Behaviors in Hadoop Clusters (Artifact)

Arising with the popularity of Hadoop, optimizing Hadoop executions has grabbed lots of attention from research community. Many research contributions are proposed to elevate Hadoop performance, particularly in the domain of self-adaptive software systems. However, due to the complexity of Hadoop operation and the difficulty to reproduce experiments, the efforts of these Hadoop-related research are hard to be evaluated. To address this limitation, we propose a research acceleration platform for rapid prototyping and evaluation of self-adaptive behavior in Hadoop clusters. It provides an automated manner to quickly and easily provision reproducible Hadoop environments and execute acknowledged benchmarks. This platform is based on the state-of-the-art container technology that supports both distributed configurations as well as standalone single-host setups. We demonstrate the approach on a complete implementation of a concrete Hadoop self-adaptive case study.