Task-Allocation in a hierarchical network topology by means of an organic middleware

Abstract:
We developed the Hierarchical Artificial Hormone System (HAHS) to assign a large set of tasks into a large distributed system of processing elements. The tasks are allocated according to their suitability and the performance of the PEs. The HAHS is a decentralised, hierarchical system avoiding single points of failures. It provides the properties of self-configuration, self-optimisation and self-healing. After we already presented first implementations of the HAHS, we will present in this paper first experiments with the HAHS on a real setup, built from Raspberry PIs. With the experiments we evaluate two new aspects of the HAHS. First, a separation of communication ways using the builtin ethernet and wireless device of the Raspberry PIs. Additionally, we will present a method for autonomous clusterhead elections in the clusters of the HAHS. With this method we can add the properties of self-configuration, self-optimisation and self-healing to the clusterheads and avoid a single point of failure. In our former implementations, a failed clusterhead would lead to a cluster loss automatically, since the cluster would lose its connection to the inter-cluster regulation cycle and thus also the connection to all other clusters.