论文库

原文摘要

The performance of real-time IoT communication systems depends not only on individual transport protocols, serialization formats, and data-processing stages, but also on how these components interact within the complete communication pipeline. Selecting an appropriate communication stack is therefore application-dependent and requires considering processing capabilities, network characteristics, communication patterns, and reliability requirements. This paper provides an experimental, evidence-based analysis of these interactions through a controlled benchmark of integrated communication pipelines deployed in the MASA living lab. We jointly evaluate local processing overhead, end-to-end latency, jitter, throughput, and delivery reliability across different combinations of transport, messaging, serialization, and compression mechanisms. Results show that component-level processing costs do not directly predict end-to-end performance. In particular, differences of only tens of microseconds in serialization or compression costs can coexist with several-millisecond differences in end-to-end latency, while middleware buffering and transport behavior substantially affect timing performance. The results demonstrate that communication-stack components should be evaluated as integrated pipelines and under representative operating loads rather than ranked independently.