Master Linux performance from the kernel up. LINUX KERNEL PERFORMANCE ENGINEERING is a practical, production-focused guide to understanding, diagnosing, and optimizing Linux systems across CPU scheduling, memory, NUMA, storage, filesystems, networking, interrupts, and kernel observability. Instead of relying on generic tuning advice, this book teaches an evidence-driven workflow: Measure → Localize → Profile → Trace → Optimize → Validate You will learn how to investigate scheduler latency, context switching, CPU migrations, memory pressure, page faults, reclaim, MGLRU, DAMON, NUMA locality, page cache behavior, blk-mq, NVMe queueing, TCP performance, SoftIRQ saturation, XDP, AF_XDP, and more. Hands-on coverage includes perf, ftrace, trace-cmd, eBPF, bpftrace, BCC, bpftool, BTF, libbpf, CO-RE, PSI, fio, iperf3, sysstat, numactl, ethtool, and other essential Linux performance tools. You will also learn how to: Diagnose CPU, memory, storage, filesystem, and network bottlenecks Interpret p95, p99, and tail-latency problems Use PSI to identify real resource pressure Analyze on-CPU and off-CPU time Trace scheduler, lock, I/O, and TCP behavior Investigate IRQ and SoftIRQ imbalance Tune cgroup v2 CPU, memory, I/O, and cpuset controls Evaluate NUMA placement and memory locality Build repeatable performance benchmarks Apply kernel and sysctl changes safely Validate every optimization with before-and-after evidence The final capstone brings everything together in a complete zero-to-production performance engineering project where you diagnose a deliberately degraded Linux server, identify the true kernel-level bottlenecks, apply controlled optimizations, and prove measurable improvements in throughput and tail latency. Whether you are a Linux administrator, SRE, DevOps engineer, platform engineer, systems engineer, performance engineer, or advanced Linux user, this book gives you a practical framework for turning Linux performance problems into measurable, explainable, and repeatable engineering decisions.
