Current estimation suggests that 70% of all cloud services run on the Linux kernel. In this lecture, we will discuss performance optimization for Linux workloads in the Azure cloud. Simply adding compute, memory, or opting for larger VM sizes does not resolve performance issues. True optimization requires analyzing workload behavior — meaning we must closely examine how an application consumes CPU, responds to memory pressure, its I/O patterns, and observe network latency. To take a proper approach to optimization, we need to characterize the workload with appropriate metrics. Part of this data is provided directly by Microsoft Azure through platform metrics, operating system monitoring, and logs using tools such as Azure Monitor with Azure Managed Grafana. In addition to these solutions, alongside classic resource monitoring tools for CPU, memory, disks, and network connections, newer kernels come with enabled support for (e)BPF. BPF enables the use of powerful tools which provide deeper real-time insights into the running kernel and application code execution. During the presentation, we will go through practical examples and demonstrate how some of these tools can be used to get richer context on performance and in parallel used in troubleshooting sessions. Micro and macro performance analysis of applications running on the Linux kernel, combined with the tools available in the Azure cloud, helps us understand how the environment operates, ultimately aiding in both performance and cost optimization.