STEAM++ Framework.pdf

STEAM++ is a lightweight and extensible framework for real-time data stream processing and decision-making in the network edge, targeting hardware-limited devices. It addresses challenges in IoT application development, such as lacking standardization, heterogeneity of IoT hardware and data formats, and complexity in implementing data analytic functions in the fog.

Key features of STEAM++ include:

Real-time data stream processing and decision-making
Targeting hardware-limited devices
Bringing data analytics techniques to the network edge using fog computing concept
Enabling design of rich solutions regarding data collection and analysis, event detection, and publication of results

STEAM++ architecture consists of:

Data production from generic raw data sources transmitted over sensor networks
Data processing by a gateway at the network edge
Analysis, Enrichment, and Evaluation processes executed on far-edge devices
Client applications responsible for data consumption and business rules processing

The framework has been tested in a real-case experiment in a semiconductor industry, processing an entire data flow and achieving:

Average RAM consumption of less than 500kb
Average CPU usage of 1.0%
Processing up to 239 data packets per second
Reducing output data size to 14% of input raw data size when notifying events

A micro-benchmark methodology is also proposed for assessing embedded IoT applications, monitoring CPU and memory usage, measuring processing time, and calculating performance metrics.