Reflective autonomic systems management

The next frontier for systems and middleware research is probably the construction of self-managed distributed systems, including notably automating the management of system configurations, performance, failure handling, and security. In the ObjectWeb community, work at INRIA on Jade, for instance, provides an erly attempt at developing a component-based framework for self-managing systems.

Scope and Objectives

This working group (WG) is responsible for the development of key services required for the construction of adaptive autonomic systems. It focuses on providing instrumentation and (re)configuration services at the middleware level, as an integral part of the component architecture, allowing the support of autonomic standards with minimal changes to the business applications.

Monitoring provides support for observing the state of a component-based distributed heterogeneous system. This includes an information model for expressing monitoring data and services for data dissemination. It also includes middleware instrumentation services that support adaptable monitoring.

Assessment provides support for identifying system state and making decisions on which adaptation procedures should be employed. This includes local and distributed assessment tools that can validate the system health, detect problems, and predict faults. It also includes policies and rules to feed the assessment tools, knowledge information models, and services to create and manage the knowledge.

Adaptation allows for dynamic (re)configuration of individual components and orchestrated state changes in a component-based distributed heterogeneous system.

The goal of collaboration on this theme would be twofold:

  • a/ to develop a comprehensive system architecture for distributed systems self-management, together with the necessary software technology tols (e.g. in terms of ADL extensions, policy languages and suppoting frameworks).
  • b/ to develop control technology (algorithms, and tools - e.g. exploiting planification and statistical techniques) required for the construction of self-managing capabilities.On this theme, experience gained and approaches developed e.g. at the U. of Bejing around the management of PKUAS, and at ISCAS on QoS management in J2EE servers, should prove of particular interest.

Partners and contributions

INRIA: Jade
JADE focuses on autonomous distributed systems management. The main goal of the Jade project is to develop component-based software infrastructures for the control and administration of large scale, long-lived distributed systems such as clustered application servers or Grid infrastructures.

Contact: Jean-Bernard Stefani

ISCAS: QoS-Driven resources management
The ISCAS will provide a framework and a set of techniques to deal with the following issues

  • QoS requirements specification describes the applications’ expectation of quality of service that middleware would provide, which is used to direct the behaviour of autonomic management.
  • Monitoring provides supports for observing the states of system, the usage of resources, the load characterization and the provision of quality of service.
  • Evaluation analyzes the information from the monitors and gives some indications about how to take reconfiguration or other management measures to adapt the behaviour of system according the target of QoS. Evaluation should be a self-learning process, which can make the matching between the scenarios and the reconfiguration policy.
  • Adaptation is the process of adjusting QoS policy and reconfiguring systematic services based on QoS specification and runtime evaluation. In the procedure, provision of resources and trade-off among multiple QoS characteristics should be considered.
  • Resource management deals with the mapping between the requirements of QoS and the resources, allocation of resources dynamically and optimization of allocation among multiple requirements according to system status.
The most important contribution we expect is to provide QoS-Driven resources management, which can deal with the reservation and reassignment of the resources to guarantee the QoS specified.

Contact: WEI Jun

*Peking University:*ReflectAll

Contact: HUANG Gang

Workplan:

  • Compare ReflectAll and Fractal/Julia
  • PKUAS to support Fractal API
  • Extend Fractal with ReflectAll concepts