Provincial Public Security Bureau: #1 in Comprehensive Database Benchmark Results
#1 in comprehensive database benchmark results!
Across large-scale batch loading, concurrent random reads, and record-by-record ingestion benchmarks, IvorySQL delivered the top overall performance score — now powering Oracle-replacement rollouts for China's Ministry of Public Security and criminal investigation departments in more than ten provinces.
Background
This customer's project was a domestic database replacement initiative for a criminal investigation (technical reconnaissance) unit within China's public security system. The core requirement was to verify whether IvorySQL could reliably handle high-concurrency writes and real-time queries against massive case data — covering joint case-handling, specialized data entry, and third-party data intake — in order to replace the existing Oracle database.
On-Site Test Environment
The benchmark used three database servers networked over dual Gigabit and 10-Gigabit switches to simulate the throughput and query load of real production scenarios:
- Database servers: Hygon C86 7285 CPU, 256GB RAM, 2.4TB SAS storage, Kylin V10 SP1 OS
- Test data: table
TB_EVT, loaded from 365 individual 4GB files, mapped to 365 table partitions
Test Scenarios
Test cases were designed around real criminal-investigation workloads:
- Batch loading scenario test
- DISTINCT computation test
- Concurrent random read scenario test
- Concurrent random read + sort scenario test
- Record-by-record loading scenario test
- Concurrent random read + record-by-record loading test
- Extreme-load stability test
Core Performance Metrics
| Core metric | Value | What it measures |
|---|---|---|
| Batch load throughput | 292MB/s | An initial measure of data entry and index-creation efficiency during batch loading — commonly used in joint case-handling, specialized data entry, and third-party data intake scenarios |
| Index processing rate | 1.4M rows/s | |
| Record-by-record load rate | 77K rows/s | An initial measure of overall efficiency processing large volumes of investigative data in real time |
| Random read response time | 3.55 sec | An initial measure of response latency for investigators' target queries |
| Random read + TOP-N response time | 13.8 sec | An initial measure of retrieval efficiency when investigators fetch the latest target data |
Customization & Compatibility Work
To meet the specific requirements of criminal-investigation workloads, the project team built substantial customizations and performance tuning on top of IvorySQL:
- Custom full-text search functionality
- Virtual partitioning support
- Compatibility with Oracle
SYSTEMpartitions - Compatibility with the Oracle dictionary tables used in the project
- Parallel index creation for partitioned tables
- Automatic replication of table structures and indexes
- A complete multi-model database plugin ecosystem and extensibility
- Optimized foreign-table functionality to better fit the project's real-world needs
- Continuous testing and parameter tuning to maximize server performance
- High-speed data file loading, several times faster than conventional ingestion
- Compatibility with Oracle's exchange-loading feature, extended with custom capabilities beyond Oracle's own implementation
- Data file backup and recovery, several times faster than conventional backup methods
- Automatic data file mounting for cross-database data synchronization, several times faster than conventional sync methods
Outcome
IvorySQL is now deeply integrated into public security criminal-investigation operations, providing a solid domestic database foundation for a "sense–analyze–decide–act" smart-policing architecture. The project has since supported domestic database replacement work for the Ministry of Public Security, the China Academy of Information and Communications Technology, and criminal-investigation departments across more than ten provinces, including Guangdong, Hebei, and Shanxi.