Scholarlink

The exponential growth of semi-structured data from IoT devices and modern applications has challenged traditional relational databases, driving interest in hybrid storage solutions that combine relational reliability with NoSQL flexibility. This study systematically evaluates the performance implications of JSON and JSONB storage formats in PostgreSQL compared to traditional relational storage, using the standardized TPC-H benchmark across datasets of 1 GB, 5 GB, and 10 GB. The evaluation encompasses import efficiency, storage requirements, query performance with and without indexing, and concurrent load resilience. Results demonstrate that while JSONB incurs 2.7 \(\times \) storage overhead and 2.7 \(\times \) longer import times, it significantly outperforms plain JSON in query execution, achieving 23 \(\times \) slower performance than relational storage without indexes, which improves to 6.6 \(\times \) slower with combined BTREE and GIN indexing. Under concurrent workloads, JSONB maintains stable throughput and low latency comparable to relational storage, while plain JSON exhibits rapid performance degradation. The findings indicate that JSONB represents a viable hybrid solution for organizations requiring schema flexibility without severely compromising analytical performance, positioning PostgreSQL as a competitive alternative to pure NoSQL systems in semi-structured data environments. The study concludes that storage format selection should align with specific workload characteristics, with JSONB offering an optimal balance for applications demanding both flexibility and performance.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

JSON Vs JSONB in PostgreSQL: Performance Implications for Querying and Data Management

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Maryam Abbasi,
  • José Silva,
  • Paulo Váz,
  • Filipe Cardoso,
  • Pedro Martins

摘要

The exponential growth of semi-structured data from IoT devices and modern applications has challenged traditional relational databases, driving interest in hybrid storage solutions that combine relational reliability with NoSQL flexibility. This study systematically evaluates the performance implications of JSON and JSONB storage formats in PostgreSQL compared to traditional relational storage, using the standardized TPC-H benchmark across datasets of 1 GB, 5 GB, and 10 GB. The evaluation encompasses import efficiency, storage requirements, query performance with and without indexing, and concurrent load resilience. Results demonstrate that while JSONB incurs 2.7 \(\times \) storage overhead and 2.7 \(\times \) longer import times, it significantly outperforms plain JSON in query execution, achieving 23 \(\times \) slower performance than relational storage without indexes, which improves to 6.6 \(\times \) slower with combined BTREE and GIN indexing. Under concurrent workloads, JSONB maintains stable throughput and low latency comparable to relational storage, while plain JSON exhibits rapid performance degradation. The findings indicate that JSONB represents a viable hybrid solution for organizations requiring schema flexibility without severely compromising analytical performance, positioning PostgreSQL as a competitive alternative to pure NoSQL systems in semi-structured data environments. The study concludes that storage format selection should align with specific workload characteristics, with JSONB offering an optimal balance for applications demanding both flexibility and performance.