2022-08-01 Reflection

Commonalities of slow SQL problems encountered:

1. These slow SQL problems encountered are not only unilateral, but also the designer of the module's architectural ability and lack of understanding of the basic knowledge of computer systems. While the system is well designed, the performance of the systemIt was also designed together.The performance problem didn't happen by accident
2. The memory management part is extremely bad, redundant memory copies, unlimited requests for memory, when to use buffers and their occupancy issues
3. The design of parallel execution failed to achieve its purpose, and the final execution of three slow SQLs only used one cpu core, which actually became serial execution

Thinking:

1. Performance analysis tools can actually do some more in-depth analysis, which is convenient for users to analyze slow SQL. Now most of them use per flame graph and explain query plan for macro analysis, use mysql workbench performance to view the statistics of mysql itself, and use the codeBuried point log coloring for detailed analysis.You can start to do some more detailed analysis reports similar to oracle awr
2. Details, details, or details, after understanding the big module interaction, the key lies in understanding the details.Any line of code requires reflection.
3. For a paper and optimization direction, the best way is to implement it in person and create it with the owner consciousness, so that it can be integrated with it.Just knowing, but not being able to do it, is not knowing.