Everything runs on Linux, including databases and cloud services. And the performance metrics usually show %CPU, or load average, I/O wait, free memory,… which are easy to show but very difficult to interpret. We will run a workload, look at “top” together and see what it tells us… or not.
People go to cloud, look at CloudWatch or other performance metric charts, and scale up and scale out their instances (and their bill) without understanding if they are really in CPU, IO or memory pressure. In this full demo session we will look together at a workload running and see how the commonly used metrics are often not understood and sometimes meaningless. A good occasion to go back to the basics: process states, virtual memory and CPU threads