Abstract
This study explores the stream of process measurements after processing at the edge node in the chain “field-level equipment → edge node → SCADA → database”. The task addressed relates to the insufficient quantitative assessment of the impact of processing scenarios at the edge level on the stream intensity, time characteristics, and load of system nodes. This paper considers an edge-based approach to the stream processing of process data from industrial automation systems. 5 scenarios for processing the stream of process data at the edge node have been considered. A comparative analysis was performed by the number of records after processing, end-to-end latency, reduction ratio, jitter, and resource load. To that end, a bench was implemented that reproduces data transmission from field-level equipment through an edge node to the SCADA system and database. The basic mode without optimization, filtering by the change threshold, periodic selection, aggregation in a time window, event filtering, and a hybrid scheme were investigated. All edge processing scenarios reduced the stream intensity compared to the baseline. The largest reduction was provided by EV and DB: 0.976 and 0.962. For most scenarios, the end-to-end latency was 71.064–81.197 ms, for HYB – 112.457 ms. The lowest jitter was obtained for EV – 25.98 ms and DB – 38.521 ms. This is explained by the fact that event and threshold selection cut off background signal changes. Preprocessing reduces the network load without a noticeable increase in the load of the edge node. This makes it possible to consider processing at the edge node as a means of data reduction and formation of a controlled stream for the SCADA system and database. The practical significance implies using the results for comparative selection of processing scenarios for process data taking into account network traffic, latency stability, as well as edge node resources