A BDI agent-based approach for Cloud Application autonomic management

Abstract

Cloud endows application with the ability of controlling runtime environment dynamically. Comparing with traditional one, Cloud APPlication (CAPP) could be more dynamic and autonomic. However, enabling CAPP to handle problems such as performance optimization or resources utility efficiency improvement in complex cloud environment with existing technologies is difficult and problematic. To address this problem, we propose a BDI (Belief-Desire-Intention) model and multi-agent oriented approach to decouple autonomous management functionality with CAPP, and simplify the management policy configuration process. We use the BDI agent paradigm to specify application and environment fact as belief, the CAPP's SLOs (Service Level Objectives) and Cloud Environment (CE) resources utility improvement desire as goal, means to achieve goal as action plan. A BDI agent oriented architecture and a Bayesian Network based belief inference model for the process of reasoning in intelligent agent is proposed to tackle CAPP autonomic management problem in cloud environment also. A prototype of this design is applied to automatically manage several multi-tiered web applications running in cloud environment to demonstrate the effectiveness of our design.