encoding

This study focuses on classifying serialization formats used in inter-service communication (ISC) within distributed systems and exploring their historical development. We have examined key features of humanreadable formats such as XML, JSON, and YAML, binary formats like Protocol Buffers and Apache Avro, and columnar formats such as Apache Parquet and ORC, among others. Our results have indicated a significant shift toward binary formats optimized for speed and compactness in recent years.

This study provides an evaluation of the impact of various serialization formats on inter-service communication performance, with a focus on serialization speed, space efficiency, and latency in environments integrating middleware, which are characteristics of microservice architectures. Through an empirical analysis of a wide range of serialization formats and comparison to the traditional standards, it highlights that the compactness of serialized payloads is more critical in reducing end-to-end latency than the sheer speed of serialization itself.

In this research, an analysis of the qualitative characteristics of messages in the Telegram messenger was carried out, which are used as raw data for further analysis of textual content. A thorough review of the parameters of these messages, such as their format, size, presence of noise, and speed. The main goal of the article is to model the optimal approach to saving a large amount of data before the important stage of text analysis. During the research, a detailed analysis of literary sources devoted to this topic was carried out.

Distributed systems rely on efficient inter-service communication, heavily impacted by data transmission costs. This study investigates alternative serialization formats, like Avro and MessagePack, to reduce data size compared to the common JSON format. We utilize a custom model to comprehensively assess the space efficiency of serialization formats across various data types. Our findings demonstrate that adopting alternative formats achieves a median reduction in serialized data exceeding 30 %.