Designing Multi-Layers Self-Adaptive Complex Applications

The impossibility of statically determining the behavior of complex systems that interact at runtime with heterogeneous devices and remote entities, may lead to unexpected system failures and performance degradation. Recently, self-adaptive applications have been recognized as viable solutions for dealing with systems where size and complexity increase beyond the ability of humans. However, self-adaptive solutions have been always studied in isolation, only involving single layers of the system (e.g. operative system, middleware, firmware, hardware). In this paper we discuss our novel idea of multi-layers deep adaptability of complex systems. We present an integrated approach for designing and coordinating applications with three-layers of self-adaptation. The solution proposed is based on specialized sense-plan-act control loops that interact each others to monitor both specific parts and also the global behavior of the system. These control loops sense unexpected behaviors that can compromise the system, then diagnose the system layer that needs an adaptation action, and finally act the proper adaptation strategy to overcome the revealed problems without user intervention. We validate the approach with a SOA-based application, the Virtual Tour Guide, designed with three-layers self-adaptive abilities in order to overcome functional/non-functional problems that can derive from the integration of heterogeneous and remote third-party service