OneAPI Tools Enhance Performance Of TencentDB For MySQL

​What’s New: Tencent has significantly enhanced the performance of its database hosting service, TencentDB for MySQL. Based on the open source relational database management system MySQL and built on Intel® Xeon® processors, TencentDB for MySQL increased its performance by using the advanced Intel® oneAPI DPC++/C++ Compiler and Intel® VTune™ Profiler (part of the Intel® oneAPI Base Toolkit).

“Tencent’s results optimizing MySQL demonstrate the importance both of using up-to-date developer tools like the Intel oneAPI DPC++/C++ Compiler and the latest optimization techniques using the Intel VTune Profiler. The significant improvements in performance yield either faster time to results or more results for business-critical applications.”

Why It Matters: Distributed data storage serves a critical role across industries and use cases, including internet, finance and e-commerce. Solutions like TencentDB for MySQL provide developers with a service for distributed data storage that supports easy setup, operation and expansion of relational databases in the cloud. However, an ongoing rise in storage and data processing requirements and demands for higher performance makes it challenging for developers. Optimizing for higher MySQL performance can help organizations process transactions and data queries faster, helping to meet evolving business and customer needs.

How It Works: Intel is leading the open ecosystem by combining its unique strengths in hardware and software tools with a commitment to building strong, open software that accelerates compute and innovation. The Intel oneAPI DPC++/C++ compiler is used for parallel programming applications and delivers productivity and performance across CPUs and accelerators. Using the compiler, the teams drove link time optimization (LTO) and profile-guided optimization (PGO) methods that helped Tencent build a high-performance MySQL. With LTO, the compiler applied various forms of intraprocedural optimization (IPO) to the entire program, allowing for deeper analysis, more optimization and better program performance. PGO provides information to the compiler about areas of an application that are most frequently executed. Together, these techniques delivered up to an 85% performance improvement for TencentDB applications¹.

Intel® VTune™ Profiler was used to collect the performance of MySQL based on the default configuration and to identify and analyze hot spots in call stacks to find additional areas to maximize performance.

_{The Small Print:}

¹MySQL performance is affected by many factors, one of which is its configuration. Different configurations may have a different impact on performance. Tencent used a configuration file, and built MySQL, both with LTO and PGO, and evaluated the performance with Sysbench. The QPS performance of oltp_read_write.lua is shown in the above figure.

_{Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.}

_{Performance varies by use, configuration, and other factors. Learn more at www.Intel.com/PerformanceIndex​.}  

_{Your costs and results may vary.} 

_{Intel technologies may require enabled hardware, software, or service activation.}