Home > Servers > Rack and Tower Servers > AMD > White Papers > Workload-Based DDR5 Memory Guidance for Next-Generation PowerEdge Servers > Structured data management and analytics (medium)
We used MariaDB to look at performance in a structured data management and analytics (medium) type of workload. The MariaDB database server benchmark uses mysqlslap, a command-line benchmarking tool that comes bundled with MySQL, an open-source relational database management system. The MariaDB benchmark is designed to simulate concurrent client connections and run SQL queries against a MySQL server to measure its performance under a load similar to real-world usage.
The metric for the workload is the number of queries per second, which is a crucial performance indicator that shows the server's ability to handle concurrent requests efficiently. Concurrent requests are the number of simultaneous database operations or queries run by multiple clients or connections. The maximum number of concurrent requests offered in the benchmark is 8,196.
We tested the benchmark with 8,196 concurrent requests. The following table shows the results:
Table 2. Test results: Structured data management and analytics (medium) type of workload, 8,196 concurrent requests
MariaDB MySQL – 8,196 concurrent requests | |||||||
Server | CPU | Memory (scaling) | Configuration | BIOS setting | Power (W) | Utilization (%) | Performance metric |
PowerEdge R7625 | 2 x AMD EPYC 9654 | 24 x 16 GB | 12 DIMMs per CPU | Default | 685.21 | 1.4 | 442 |
PowerEdge R7615 | AMD EPYC 9654P | 12 X 16 GB | 12 DIMMs per CPU | Default | 476.28 | 3.22 | 581 |
We see better performance—38 percent—in PowerEdge R7615 and low CPU utilization in both 1-socket (R7615) and 2-socket (R7625) versions. Based on these insights, we moved to test performance only on the PowerEdge R7615 and changed to the 32-core CPU, AMD EPYC 9354P. To enhance the stress on the server, we adjusted the test-definition.xml file to double the number of concurrent users to 16,392.
The following figures show the performance, energy efficiency, and performance per dollar for MariaDB MySQL—concurrent requests: 16,384 on PowerEdge R7615 with 4th Gen AMD EPYC 9354P:
Figure 16. Performance trend for MariaDB MySQL—concurrent requests: 16,384 with different DIMM configurations and memory capacity for 4th Gen AMD EPYC 9354P processor-based PowerEdge R7615 with default BIOS settings
Figure 17. Performance per watt trend for MariaDB MySQL—concurrent requests: 16,384 with different DIMM configurations and memory capacity for 4th Gen AMD EPYC 9354P processor-based PowerEdge R7615 with default BIOS settings
Figure 18. Performance per dollar trend for MariaDB MySQL—concurrent requests: 16,384 with different DIMM configurations and memory capacity for 4th Gen AMD EPYC 9354P processor-based PowerEdge R7615 with default BIOS settings
The configuration of 8 DIMMs per CPU with 128 GB DIMMs had the best performance. The most energy efficient was the configuration of 6 DIMMs per socket with 32 GB DIMMs. The configuration of 8 DIMMs per socket with 32 GB DIMMs had the highest cost efficiency.
We recommend the PowerEdge R7615 featuring AMD EPYC 9354P in the 8-DIMMs-per-socket configuration. Based on test results, it provides the best performance per dollar and best performance in most DIMM sizes for structured data management and analytics (medium) workloads.