Abstract

This paper describes a biologically-inspired architecture,called SymbioticSphere, which allows data centersto autonomously adapt to dynamic environmental changes.SymbioticSphere follows biological principles such as decentralization,evolution and symbiosis to design applicationservices and middleware platforms in a data center. Eachservice and platform is designed as a biological entity, andimplements biological behaviors such as energy exchange,migration, reproduction and death. Each service/platformalso possesses behavior policies, as genes, each of which defineswhen to and how to invoke a particular behavior. Thispaper presents a set of behaviors for services and platforms,and describes how services and platforms act and interactwith each other. Simulation results show that services andplatforms autonomously adapt to dynamic network conditions(e.g., user location, network traffic and resource availability)by evolving their behavior policies across generations.Simulation results also show that services and platformscoevolve to improve their adaptability by adjustingtheir behavior policies cooperatively.