BPB Publications · 2026
A 424-page, methodology-first guide to the diagnosis and resolution of Linux performance problems across CPU, memory, storage, networking, containers and eBPF, written by an infrastructure engineer.
Distributed worldwide by BPB Publications and listed on Amazon in 16 countries.
Why this book
Incidents are frequently closed by restarting a service, with no explanation of the underlying cause, and the problem subsequently recurs. This book sets out a systematic alternative: a diagnostic method that produces conclusions which can be documented and defended.
Formal measurement methodology, queueing models and bottleneck theory are established first, so that each metric is understood before further data is collected. The tooling is then introduced within an existing analytical framework.
CPU scheduling and latency, memory and NUMA behaviour, the block layer and filesystems, and the network stack. Each receives dedicated treatment rather than a summary paragraph and a list of commands.
The case studies are drawn from production troubleshooting rather than constructed examples. They originate in high-severity enterprise incidents in which an incorrect diagnosis carried measurable cost.
Containers and virtualization performance, cgroup accounting, and a full chapter on eBPF observability alongside perf and ftrace, reflecting current practice.
The construction of load tests whose results remain valid under production conditions, and the identification of benchmarks that measure the wrong quantity.
The conversion of measured behaviour into headroom forecasts and scaling decisions that can be substantiated to stakeholders, rather than provisioning additional capacity speculatively.
Performance engineering is a specialised discipline rather than an afterthought. The objective of this book is to enable the reader to explain system behaviour precisely, justify the decisions taken, and demonstrate the resulting improvement.
Outcomes
Six specific capabilities, each applicable to the analysis of a live production system.
Table of contents
The book progresses from measurement theory through each subsystem into profiling and tracing, concluding with production case studies and eBPF.
Readership
Responsible for service level objectives, and required to establish the cause of a breach and the measures that will prevent recurrence.
Responsible for fleet sizing and architectural approval, and requiring capacity figures that remain valid under load.
Already familiar with the tooling, and seeking the formal basis behind it: queueing theory, bottleneck analysis and measurement method.
Working on code that performs well in isolation but poorly in production. The book addresses the causes of that discrepancy.
Experienced in operational troubleshooting, and seeking a formal vocabulary and a repeatable procedure.
Familiarity with the Linux command line and a working understanding of processes and memory. Prior kernel development experience is not required.
Where to buy
Available in paperback, ebook and audiobook. Purchasing directly from the publisher provides access to the combined paperback and ePub edition.
Supplied directly by BPB Publications, who hold worldwide rights and ship internationally. The combined paperback and ePub edition is available only from this source, providing both a print copy and a searchable digital edition.
Buy from BPB OnlineListed on 16 Amazon marketplaces worldwide, in paperback and Kindle editions.
Buy on AmazonAmazon links are affiliate links where an Associates programme is available. As an Amazon Associate the author earns from qualifying purchases, at no additional cost to the purchaser.
Questions
Common questions regarding the book's scope, formats and availability.
Preview
The free preview covers the opening material and the complete table of contents, allowing the level and approach to be assessed prior to purchase.
Amazon, Kindle and audiobook editions are listed above.