Laurent Broto

Autonomic Management for Grid Applications

Distributed software environments are increasingly complex and difficult to manage, as they integrate various legacy software with specific management interfaces. Moreover, the fact that management tasks are performed by humans leads to many configuration errors and low reactivity. This is particularly true in medium or large-scale grid infrastructures. To address this issue, we developed Jade, a middleware for self-management of distributed software environments. In this paper, we report on our experiments in using Jade for the management of grid applications.

A High Level Approach for Generating Model's Graphical Editors

Domain Specific Languages (DSL) are increasingly used in software engineering and other domains. The result is an increasing need of appropriate DSLs tools, especially platform for building, editors and runtime associated with DSLs. Different experiences show that existent DSL tools are generally not user friendly enough, or simply unadapted for the generation of graphical DSL editors. In this paper we present a higher level and effortless framework for generating graphical DSL editors. This framework was designed and experienced in the context of an autonomic management system based on a component model.

Towards a Model Driven Autonomic Management System

Distributed software environments are increasingly complex and difficult to manage, as they integrate various legacy software with specific management interfaces. Moreover, the fact that management tasks are performed by humans leads to many configuration errors and low reactivity. This is particularly true in medium or large-scale distributed infrastructures. To address this issue, we explore the design and implementation of an autonomic management system. The main principle is to wrap legacy software pieces in components in order to administrate a software infrastructure as a component architecture. However, we observed that the interfaces of a component model are too low-level and difficult to use. Consequently, we explore the use of a model driven approach where several UML profiles are used to specify the different facets of an autonomic management policy.

Power-aware scheduler for virtualized systems

Nowadays, virtualization is present in almost all computing infrastructures. Thanks to server consolidation and VM migration, virtualization helps in power reduction. However, modern powerful computers with higher processor frequency, multiple cores and multiple CPUs constitute the main factor contributing to the continuously increase of energy consumption in numerous computing infrastructures. In this context, energy management takes a critical importance. A hardware technology, called Dynamic Voltage and Frequency Scaling (DVFS), serves to dynamically modify the processor frequency (according to the CPU needs) in order to achieve less energy consumption. However, lowering frequency also generates poor virtual machine (VM) performance.n In this paper, we propose a solution consisting of an extended VM scheduler and DVFS, and report some experiments based on this proposal. This enhanced scheduler, according to VM CPU load, dynamically scales processor frequency in order to save energy. The idea is to adapt the current VM scheduler to analyze CPU load, and modify the current processor frequency to the lowest possible, but still support the guaranteed VM performance. The algorithm is designed and simulated on a web server as the sample application and Xen as the virtualization platform. Test results and performance evaluations prove our design and implementation.

Energy-QoS tradeoffs in J2EE hosting centres

Nowadays, hosting centres are widely used to host various kinds of applications e.g., web servers or scientific applications. Resource management is a major challenge for most organisations that run these infrastructures. Many studies show that clusters are not used at their full capacity which represents a significant source of waste. Autonomic management systems have been introduced in order to dynamically adapt software infrastructures according to runtime conditions. They provide support to deploy, configure, monitor, and repair applications in such environments. In this paper, we report our experiments in using an autonomic management system to provide resource aware management for a clustered application. We consider a standard replicated server infrastructure in which we dynamically adapt the degree of replication in order to ensure a given QoS while minimising energy consumption.

Autonomic energy management in clusters

Nowadays, computer clusters are widely used in many areas (scientific computing, commercial web servers, databases, etc). Energy management in such infrastructures is becoming increasingly important as their energy consumption is continuing to raise.n In this paper, we report on our experiments in using an autonomic management system to provide energy aware management in a cluster. We improve energy management by dynamically resizing the active server set according to varying workload conditions. We illustrate our approach with two scenarios: the case of a dedicated cluster, and the case of a hosting center.

Metamodeling Autonomic System Management Policies - Ongoing Works

Autonomic computing is recognized as one of the most promising solution to address the increasingly complex task of distributed environments administration. In this context, many projects relied on software components and architectures to organize such an autonomic management software. However, we observed that the interfaces of a component model are too low-level, difficult to use and still error prone. Therefore, we introduced higher-level languages for the modeling of deployment and management policies. These domain specific languages enhance simplicity and consistency of the policies. Our current work is to formally describe the metamodels and the semantics associated with these languages.