Fei has over 10 years of development experience at Tencent and Sogou. Through this account, he shares his deep insights on performance.
This article explains how to generate and interpret flame graphs for CPU performance profiling on Linux, covering the use of perf for sampling, the underlying kernel mechanisms, and the processing steps with Brendan Gregg's FlameGraph scripts to visualize hot functions.
This book explores fundamental computer science concepts including programming language execution, compiler operations, memory allocation mechanisms, and system architecture through accessible explanations and practical examples.
This article explains how to correctly obtain CPU utilization inside containers, compares host and container metrics, describes the use of lxcfs and cgroup files (including cgroup V1/V2) for accurate measurement, and clarifies why container statistics omit nice, irq, and softirq fields.
CentOS's end-of-life announcement has sparked significant discussion in China's tech community, with experts viewing it as a catalyst for domestic OS development despite short-term challenges.
This article explains how Linux cgroup’s CPU controller works, covering the creation of cgroups, the kernel structures involved, how CPU time limits are configured via cfs_period_us and cfs_quota_us, how processes are attached to cgroups, and the scheduling mechanisms that enforce bandwidth limits in containers.
This article explains how Linux calculates CPU utilization shown by the top command, detailing the role of /proc/stat, kernel sampling mechanisms, and the handling of user, nice, system, IRQ, soft‑IRQ, iowait and idle times, while providing code examples and practical insights.
This article explains how Linux allocates a process's stack virtual address space at load time, when and how physical pages are actually allocated on demand via page‑fault handling, and how the stack can automatically grow within configurable limits.
This article explains Linux load average, covering how it is calculated, its relationship with CPU usage, and how the kernel exposes load data to applications.
This article explains how Linux creates and allocates PID namespaces, how Docker containers obtain their own PID namespaces using CLONE_NEWPID, the internal kernel structures involved, and how processes view their PID inside a container.
This article explains the concept of CPU turbo boost, why single‑core performance metrics can be misleading, how different CPUs vary in base frequency and cache, and how turbo technology automatically raises clock speeds within safe limits to improve overall processing power.
