What optimizations can achieve the best performance in stress testing?

This topic has been translated from a Chinese forum by GPT and might contain errors.

Original topic: 哪些优化能为压测拿到最好成绩

| username: 江湖故人

Not considering business usage, only considering the best performance results.

| username: 小龙虾爱大龙虾 | Original post link

Various caches, create a result set cache, and then during testing, only pass a few values so that everything is cached, and it will directly return the data to you.

| username: 春风十里 | Original post link

What benchmarking tool are you using? Sysbench, TPCC, or something else? I noticed that on the forum, the results vary with the same machine but different methods. For example, with the same mid-to-low hardware configuration, TPCC 1000 warehouse benchmarking results favor TiDB, while Sysbench with 10 million data points favors MySQL. Reference: Tidb vs Mysql 8 Low Configuration Host Performance Test Comparison
I think it still depends on the specific benchmarking tool, scenario, and data volume, and then we should study how to improve the benchmarking performance results.

| username: Jellybean | Original post link

Deploy the cluster according to the configuration parameters recommended for production environments on the official website. If possible, avoid mixed deployment and keep components independent.

Additionally, for some key performance tuning during stress testing, you can also refer to the recommendations on the official website.

| username: tidb菜鸟一只 | Original post link

If it’s a business stress test, just cache all the data to be stressed into Redis, and then stress test Redis. It will definitely be much more effective than stressing the database.

| username: dba远航 | Original post link

Are you doing this to prepare a presentation for others?

| username: system | Original post link

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.