Perfomance evaluation of Java, JavaScript and PHP serialization libraries for XML, JSON and binary formats

Abstract

The aim of this paper is to compare the formats and libraries used for serialization and deserialization of data, typically with RESTful web services, in terms of the processing time and size of the output data. The formats tested include XML, JSON, MessagePack, Avro, Protocol Buffers, and native serialization of each of the tested programming languages. Serialization and deserialization is tested in PHP, Java and JavaScript using 49 different official and third party libraries. The benchmark is fully open-sourced and automated, thus easily repeatable and extensible. The testing environment is designed to be isolated from the rest of the operating system using Docker containers having zero performance penalty in contrast to virtualization. The results show huge differences in processing time among libraries. Considering the output data size, binary formats with predefined schema, such as Avro and Protocol Buffers, provide the best efficiency.