DBSitter-AS: um Framework Orientado a Agentes para Construção de Componentes de Gerenciamento Autônomo para SGBD

Abstract

Autonomic Computing is a research area that aims at developing software capable of self-configuring, self-optimization, self-protecting, self-healing, self- knowledge and the anticipation of needs, through the use of open patterns. In the specific case of Data Base Management Systems, these features can be summarized in self-management characteristics. One of the means of implementing self- management features is by developing Multiagent Systems (MAS) which monitor, prevent and solve faults. To the best of our knowledge, there are not many formal specifications of how to implement autonomic characteristics for DBMS, in particular if they are developed as MAS. In this light, the goal of this research work is to specify an architectural framework to guide the development of MAS components that can provide DBMS with self-management functionalities, thus helping the DBA in their daily activities. The DBSitter-AS framework was designed with the Tropos methodology. It specifies how to develop an agent society that is capable of performing coordinated actions of monitoring, fault prevention and resolution in commercial DBMS. The developed agent society is set up as a layer external to the DBMS. It counts with a persistence structure and allows for the recording of organizational rules and politics that must be followed by the agent society. The DBSitter-AS allows for the addition of new fault symptoms and fault-solving actions for different DBMS. In order to show how the specification can be adapted to different situations, we present an implementation example for a typical fault case.