Site Reliability Engineering

Go beyond basic Linux tuning and learn how performance really works under the hood. Explore kernel internals, hardware behavior, benchmarking and production optimization with practical tools, real trade-offs and copy-paste-ready examples. Build the skills to diagnose bottlenecks and solve performance problems with confidence.

When a 2,048-GPU training cluster running a 400-billion-parameter model sees its iteration time double without triggering a single alert, firing a CPU panic, or dropping a single packet, you don't have a machine health problem : you have a coordination failure. Beyond Uptime is the definitive field guide for Site Reliability Engineers, platform leads, and cloud architects tasked with keeping frontier AI compute running at peak efficiency. Moving beyond deceptive "GPU Utilization" charts, this book uncovers the hidden mechanics of barrier synchronization, lossless network congestion cascades (PFC/ECN), silent PCIe downtraining, and Silent Data Corruption (SDC). Packed with eBPF tracing methods, streaming gNMI telemetry rules, statistical tail analytics (MAD/CUSUM), and dynamic circuit-breaker strategies, Beyond Uptime gives you the battle-tested framework needed to isolate stragglers in minutes rather than days. Stop paying the silent tax on your AI infrastructure!

In this book, one estate runs through all fourteen chapters. There are eight services across two regions, a hybrid link to an older data centre, a storefront that real customers use, and one service backed by a language model. We use OpenObserve for that, bring all the signals into it, shape the data until long-term storage becomes affordable, and then spend the rest of the time diagnosing the estate when it breaks. There won't be a feature tour here. The features don't stand the test of time, and the screenshots are even worse. What you'll find is a method. First, scope the flag, orient before acting, form a hypothesis you can disprove, test it against evidence rather than instinct, repair, verify, and add the prevention that stops the same failure twice. That loop works whatever platform you use.

Build AI agents that actually hold up in production. This practical guide covers agent architecture, tool use, multi-agent systems, security, reliability, observability and operations, with working code throughout. Learn how to build agents that are dependable, cost-aware and ready for the real world.

AI agents are moving from experiments into real production systems. This book shows experienced engineers how to design, deploy and operate reliable agent systems that can scale. From architecture and infrastructure to security, reliability and governance, it provides practical patterns and production-ready examples for building autonomous systems that work in the real world.

When systems fail, theory meets reality. This book explores how to design, build and operate software that holds up under pressure, from unexpected outages to subtle failures buried deep in the architecture. Learn the principles and practical techniques that help engineers make better decisions, reduce risk and build systems that keep working when things go wrong.

Build a practical foundation in Site Reliability Engineering. Clear diagrams connect service maps, reliability objectives, observability, incident response, safe releases, and recovery.

Mastering Keepalived is a practical guide to building reliable, highly available infrastructure with VRRP, failover and health checking. Learn how to design, configure, troubleshoot and secure Keepalived in real production environments, from core networking concepts to advanced HA architectures and day-to-day operations.

A practical guide to building reliable LLM integrations that hold up in production. Learn how to design provider-agnostic APIs, handle failures, add observability and security, and keep your architecture maintainable. With a complete Python codebase, this book turns proven patterns into working software.

Running ClickHouse at petabyte scale takes more than knowing SQL. This practical guide shows experienced engineers how to design, deploy and operate production clusters, from data modeling and ingestion to Kubernetes, query tuning, observability, disaster recovery and cost control, with real-world patterns and concrete configurations throughout.

Learn real DevOps by building a production-style environment on your laptop. No cloud bills or subscriptions. Just systems you control end-to-end.

AI models do not have to be huge, slow or expensive. Distilling Intelligence explores how to build smaller models that perform at scale, then secure the APIs that serve them. From compression and distributed serving to extraction attacks, observability and incident response, this book covers what it takes to run AI reliably in the real world.

A brilliant engineer with no evidence is reduced to guessing, while an ordinary engineer with good evidence looks like a genius. Everything in this book is an attempt to get you to think differently. You don't need a huge team to do this. You don't need to get anyone's approval for a budget, and you don't have to rewrite any services. What you need is a system that tells you the truth about itself, and you need it before the incident rather than during it.As a team, we'll build that system together on one small storefront, and we'll do it the way real teams build, which is to say not perfectly but in order.

Learn Argo Rollouts through hands-on Kubernetes labs. Build canary and blue-green deployments, control traffic with Istio, validate releases with Prometheus, run Experiments, automate promotion and rollback, and manage progressive delivery with Argo CD and GitOps.

Somewhere there is a server with two GPUs for which no operational process exists. This book turns that into an operable platform — hands-on, in 23 labs: vLLM, KServe, LiteLLM, the NVIDIA GPU Operator, Keycloak, OpenBao, ArgoCD, pgvector. At the end your developers write code with local and external models through a governed gateway.