A methodology based on computational patterns for offloading of big data applications on cloud-edge platforms

Internet of Things (IoT) is becoming a widespread reality, as interconnected smart devices and sensors have overtaken the IT market and invaded every aspect of the human life. This kind of development, while already foreseen by IT experts, implies additional stress to already congested networks, and may require further investments in computational power when considering centralized and Cloud based solutions. That is why a common trend is to rely on local resources, provided by smart devices themselves or by aggregators, to deal with part of the required computations: this is the base concept behind Fog Computing, which is becoming increasingly adopted as a distributed calculation solution. In this paper a methodology, initially developed within the TOREADOR European project for the distribution of Big Data computations over Cloud platforms, will be described and applied to an algorithm for the prediction of energy consumption on the basis of data coming from home sensors, already employed within the CoSSMic European Project. The objective is to demonstrate that, by applying such a methodology, it is possible to improve the calculation performances and reduce communication with centralized resources.