Authors:
Egberto A. R. de Oliveira
1
;
Flavia C. Delicato
2
;
Atslands R. da Rocha
3
and
Marta Mattoso
1
Affiliations:
1
PESC/COPPE, Universidade Federal do Rio de Janeiro, Rio de Janeiro, RJ, Brazil
;
2
Instituto de Computação, Universidade Federal Fluminense, Niterói, RJ, Brazil
;
3
Universidade Federal do Ceará, Fortaleza, CE, Brazil
Keyword(s):
IoT, Internet of Things, Data Streams, Data Stream Processing, Edge Computing, Adaptive Sampling.
Abstract:
The Internet of things (IoT) has transformed the internet, enabling the communication between every kind of objects (things). The growing number of sensors and smart devices increased the possibilities of data generation and collection. This led to an explosion of data streams being produced which are challenging to be processed in real-time. Regarding the nature of the data, the huge volume, heterogeneity, continuity, disordering, noise and unpredictable rate are some challenging aspects to tackle. Regarding the data processing, the core activities from the data acquisition to the production of high-level knowledge also pose challenges related to limited computational and energy resources and high network latency. In this context, we propose a framework to support activities of a data stream processing workflow for IoT. It aims allowing real-time data processing with low power consumption. Edge computing is used to bring the data processing closer to the data sources and allow actio
ns to be triggered quickly. An adaptive sampling strategy combined with a data prediction model are adopted to reduce the network traffic, thus decreasing the power consumption of the network devices. Experiments show that the proposed framework is able to achieve up to 60.58% average energy consumption savings to sensor nodes and still meet a strict execution time threshold of 1s without compromising the accuracy of the output data on different scales of input streams.
(More)