Data Stream Processing at Network Edges

This paper studies the problem of finding an assignment of data stream processing components onto servers under the objective to minimize both energy consumption and average delay experienced by end users within the system. The aforementioned problem is studied under the context of internet of things (IoT), where servers belonging to micro-datacenters are placed at network edges and thus close to data. We propose two algorithms to tackle the aforementioned problem taking into account limited server and network resources at micro-datacenters. The first algorithm is called Delay Aware Algorithm (DA) and results in a delay efficient assignment without taking into account energy consumption. The second algorithm is called Energy Efficient and Delay Aware Algorithm (EFDA) and results in a both energy and delay efficient assignment of application processing components onto servers within the system. We provide an experimental evaluation to show the behavior of the proposed algorithms against algorithms in the literature.