Development and Evaluation of a hierarchical artificial hormone system for task allocation in large scaled distributed systems (HiKüHoS).

• Applicant: Professor Dr. Uwe Brinkschulte
• Subject Area: Computer Architecture, Embedded and Massively Parallel Systems
• Term: from 2012 to 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 224969246

The artificial hormone system (AHS) is a flexible, robust and real-time capable concept for the task allocation in a distributed system. It fulfills the properties of Self-Configuration, Self-Optimization and Self-Healing. Thereby it is capable of an autonomous operation and eases the configuration and maintenance of the system by its developer or user. The aim of the HiKüHos project is to realize the properties of the AHS in large scaled distributed systems (i.e. 10000 processing elements). In the first two project years we researched and developed the fundamental concepts and techniques of the HiKüHos project. Based on the findings and experiences of those first two years, we want to continue our research and bring the project to a successful end. First of all, the remaining work packages of the already initially requesetd but not yet granted third project year shall be completed (detailed comparative evaluation, usability in a real application). Furthermore, a new idea we developed in the first two project years is to expand the idea of hierarchy to a recursive multistage topology. Such topologies can be found for example in networked buildings (factory, smart home) up to smart cities. A recursive multistage artificial hormone system can be realized by extendingg the well-known AHS. Within a hierachy level, task distribuition will be handled by several horizontal AHS. To manage the task allocation between the adjacent levels, a vertical AHS will be introduced. This concept supports many different topologies and reduces the communication load. The applicant is convinced that continuing the research work in the HiKüHoS project will substantially contribute to research in the field of robust, real-time-capable task allocation with self-X properties.

• DFG Programme: Research Grants