TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
The Pragmatic Engineer published excerpts from Sam Newman’s 2026 book, Building Resilient Distributed Systems, on disaster recovery and AI-powered software resilience. The material covers recovery targets, infrastructure constraints, model and vendor uncertainty, and the trade-offs between recovery speed, data loss, cost and complexity.
The Pragmatic Engineer has published excerpts from Sam Newman’s 2026 book, Building Resilient Distributed Systems, outlining how software teams can plan for major outages and manage risks in AI-powered systems. The report focuses on disaster recovery targets, infrastructure availability, and the uncertainty created by changing models and vendors.
Newman’s excerpt explains that disaster recovery planning addresses events that can disable a large portion of a system at once, including fires, floods, power failures and problems in cloud infrastructure. It presents recovery point objective, or RPO, as the maximum acceptable span of potential data loss, and recovery time objective, or RTO, as the acceptable period of service downtime after an incident.
The report describes several recovery approaches, from backup and restore to pilot-light and warm-failover configurations, as well as multisite active-active systems. These options differ in cost, operational complexity and recovery speed. Newman’s excerpt says that selecting an approach requires setting requirements for the particular system rather than assuming one pattern suits every service.
The publication also covers resilience issues specific to AI-powered software. Its summary points to dependence on model providers, changes in model behavior or availability, and the need for observability. It highlights multi-model and multi-vendor strategies, transparent cost tracking and regular vendor analysis as ways teams can manage uncertainty. The report says ongoing GPU and CPU shortages also make it unsafe to assume that infrastructure will always be available when needed.
Recovery Targets Shape System Resilience
The material matters to engineering teams because recovery plans determine how much data a business may lose and how long its services may remain unavailable after a severe incident. Clear RPO and RTO targets make those expectations measurable and help teams compare recovery designs against actual operational needs.
For AI services, resilience also depends on more than keeping servers online. A system may rely on a particular model provider, model version or scarce accelerator capacity. Newman’s topics point to a broader planning challenge: teams need visibility into dependencies and alternatives, while recognizing that redundancy can add cost and complexity. The excerpts identify practices and trade-offs, but do not provide evidence that any single approach will prevent outages.
As an affiliate, we earn on qualifying purchases.
From Site Outages to Model Dependencies
The source is a report and excerpt published by The Pragmatic Engineer, whose editor says Newman’s book took three years to complete, rather than the two initially planned. The editor said Newman revisited the material to update it where AI was relevant to resilient system design. The book is published by O’Reilly Media, and the excerpts are used with permission.
The selection covers parts of two chapters rather than the entire book. Its disaster recovery discussion uses “site” as a broad term for the location of computing infrastructure, whether a public-cloud region, data center or comparable environment. The report’s AI coverage connects that established recovery problem to provider and infrastructure dependencies that can change over time.
“The recovery point objective (RPO) defines the maximum amount of time that can have passed since the data was last backed up or synchronized.”
— Sam Newman, in an excerpt from Building Resilient Distributed Systems
AI system resilience monitoring tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Excerpts Leave Key Details Open
The published material does not specify recovery targets for a particular company or system, nor does it name a universally preferred disaster recovery pattern. Costs, acceptable downtime, data-loss tolerance and technical constraints vary by service. The excerpt supplied here ends during its discussion of RTO, so it does not include the full treatment of recovery procedures.
The report also does not quantify how often AI-related outages occur or establish that AI code generation is causing a general rise in production incidents. Its editor describes more deployments and bugs as contributing to outages, but the material provides no incident dataset or comparison period. It likewise offers no named provider examples, availability figures or measured results for multi-vendor and multi-model arrangements.
multi-region cloud failover solutions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Teams Must Set Their Own Recovery Plans
The report presents planning choices rather than a specific upcoming policy or product change. Engineering teams applying its guidance would first define acceptable data loss and downtime, then assess which recovery pattern fits their service and resources. They would also need to account for vendor constraints, infrastructure availability and the operational cost of maintaining alternatives.
The Pragmatic Engineer says the excerpts cover only part of the book’s 16 chapters and directs readers to the publisher for the full text. No further release, implementation milestone or follow-up evidence is specified in the source material.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the news?
The Pragmatic Engineer published excerpts from Sam Newman’s 2026 book about disaster recovery and resilience in distributed and AI-powered systems.
What do RPO and RTO mean?
RPO is the maximum acceptable period of potential data loss. RTO is the maximum acceptable time a service can remain unavailable after an incident.
Which recovery approaches does the report describe?
The report lists backup and restore, pilot-light failover, warm failover and multisite active-active approaches. They involve different costs, levels of complexity and recovery characteristics.
What AI-specific resilience concerns are covered?
The report highlights reliance on model providers, model changes, observability and shortages of computing resources such as GPUs and CPUs. It also discusses multi-model and multi-vendor planning, cost transparency and regular vendor review.
Does the report establish that AI is causing more outages?
No. The editor links more deployments and bugs with outages, but the supplied material provides no incident data or comparison baseline to establish the scale or cause of any increase.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
