After Astra Group's expansion, the client had two different PostgreSQL database management systems: one was long-used and included in the Astra Linux operating system (it also came with a data protection system), and the other was a version by Tantor Labs.
The client wanted to know which of the builds was more efficient and which had more potential for use in new tasks. That's why we needed to test both DBMSs while emulating real load and its fluctuations.
The client approached ICL Services with the request, well aware of our competence in load testing, and was extremely satisfied with the result.
Key Challenges
- Compare the performance of two DBMSs on Astra Linux to select the optimal configuration.
- Identify 'bottlenecks' in OS and DBMS operation.
- Produce detailed summary reporting on client metrics.
Implemented our solution
The project was implemented in November–December 2022. ICL Services conducted the testing in its own test environment in the shortest possible time — the main work on the project was completed in two weeks.
The virtual machine tests were conducted in two stages (at the second stage, the DBMS builds were optimized to fit the configuration of the customer's Tantor platform) using such metrics as:
— average number of successful transactions per second;
— average number of transactions with errors for the whole measurement period;
— CPU load;
— RAM use;
— disk I/O operations.
We collected performance data using the hybrid transaction/analytical processing (HTAP) methodology, running the script in a loop five times with a 25–40 second break on each machine.
Products and technologies
- Astra Linux
- DBMS (database management systems) based on PostreSQLTantor
- PostreSQL Tantor-based DBMS administration and monitoring platform
- Swarm64 DA Benchmark Toolkit
NMon, pgmetrics, iostat
Results
- Following the tests, we compiled key metrics into a 120-page report, presenting and sending them to the client.
- Changing the configuration of the builds to fit client settings helped considerably boost the performance: for example, after optimization, the performance of the Tantor-based DBMS increased by 3.3 times in terms of the number of processed transactions.
- The tests helped the client identify a previously unnoticed bug in the taking of CPU Usage metrics in the Tantor administration and monitoring platform.
