Performance is not good how to do? I'll go through the list.Brendan Gregg is a senior performance architect for Netflix and a renowned performance tuning expert. The "Top of performance: insight into systems, enterprise and cloud Computing" is a book that can be said to be the culmination of the performance tuning landscape.
Not long ago, Brendan Gregg in SREcon16 shared the "performance checklists for SREs", is also a dry full. Let's see it together.
For a single instance, you can consider these 10 commands. 10 commands, one-minute analysis of Linux performance issues is detailed in this article.
What is Performance engineering? SRE (Site Reliability Engineer, website reliability engineer) Everyone should be no stranger. SRE when encountering performance problems, make an emergency response. But performance engineering is not the same as this.
The goal of performance engineering is to achieve the best possible cost-effective, continuous optimization, while optimization is endless. Fixing performance issues can take a few hours, days, weeks, or even months, depending on how long it takes. It takes time to read the documentation and source code to do various experiments. There is usually not a so-called "good state", good or bad, which may depend on experimentation or intuition.
This diagram is important to see what commands can be used to perform performance analysis in different parts of the system.
Then take a look at the emergency response treatment of SRE.
The goal is to solve the problem in a few minutes. Quick fix is kingly. You can scale up, roll back, or even turn traffic away. The problem must be solved under pressure, and it may take three o'clock in the morning to get up. In this case, the historical state is the "good state".
The Cloud analytics process for Netflix.
Read the original
What about poor Linux performance? I'll go through the list.