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, covering disaster recovery and resilience in AI-powered software. The material outlines recovery targets and infrastructure patterns, while discussing risks from changing AI models, provider constraints and uncertainty.
The Pragmatic Engineer has published excerpts from Sam Newman’s new book, Building Resilient Distributed Systems, focusing on disaster recovery and the reliability challenges of AI-powered software. The report describes practical recovery targets and patterns, and discusses how changing models, vendor constraints and uncertainty affect plans for keeping systems running.
Newman’s excerpt frames disaster recovery (DR) as preparation for major failures that can affect an entire site, including fires, floods, power problems or failures in cloud infrastructure. These events differ from the failure of an individual service instance: a widespread outage can make other resilience measures ineffective. The book presents DR as a set of procedures for restoring service after such events, with choices shaped by the cost and complexity an organization can accept.
Two measures help teams set requirements. The recovery point objective (RPO) sets the maximum acceptable period of potential data loss; hourly backups, for example, can leave up to an hour of data at risk. The recovery time objective (RTO) sets how long a system may remain unavailable. Newman says RTO includes recovering data, making computing resources available and redirecting traffic to an alternative site.
The report also previews DR approaches ranging from backup and restore to pilot-light, warm-failover and multisite active-active configurations. It says the book addresses infrastructure availability amid CPU and GPU constraints, along with resilience concerns specific to AI-powered services: dependence on model providers, changes to models, observability, and uncertainty about vendors. Suggested tactics in the report include using multiple models or vendors, making costs visible and reviewing vendors regularly.
Recovery Plans for AI Services
The material connects familiar continuity planning with newer dependencies in AI-powered software. A service may rely not just on its own application and data, but also on outside model providers and computing resources that may not be available when needed. That makes it useful for engineering teams to identify which components are critical and what fallback options they can actually support.
Newman’s framing also makes the trade-off explicit: shorter recovery times and lower tolerance for data loss generally require more spending and operational complexity. Teams need to set objectives that fit their service and users, rather than assume every system needs the same level of redundancy. In the AI setting, observability and vendor review can help teams detect changes and understand dependencies, but the report does not claim that any single measure prevents outages.
Products Worth Considering
LG UltraGear 34GX90SA-W 34-inch WQHD Curved OLED Gaming Monitor
Samsung Odyssey G4 25-inch FHD Gaming Monitor
LG UltraGear 32GX850A-B 32-inch 4K UHD Glossy OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
From Site Outages to Model Risk
The report is an excerpt feature based on two chapters of Newman’s book, rather than an account of a particular outage or a new industry-wide measurement. The Pragmatic Engineer says Newman spent three years writing the book, one year longer than his original plan, and revised its material to address AI’s relevance to resilient system design.
For the DR discussion, Newman uses site as a broad term covering a cloud region, data center or other location where computing infrastructure operates. The excerpt says the patterns can apply to public- and private-cloud settings. The report also notes that infrastructure availability and provider limits matter when planning recovery, particularly where CPU and GPU capacity is constrained.
The publication identifies the excerpts as coming from Building Resilient Distributed Systems, copyright 2026 Sam Newman and published by O’Reilly Media, Inc., with permission. The report’s author says the book recommendation was unpaid and that links were not affiliate links.
““As you reduce your period of acceptable downtime or data loss, cost and complexity increase.””
— Sam Newman, in the book excerpt published by The Pragmatic Engineer
Limits of the Published Excerpts
The report does not provide a case study with measured recovery results or recommend one DR pattern for all organizations. It also gives no specific availability figures, recovery targets or cost estimates for the approaches described. Readers would need to determine suitable RPO and RTO targets from their own service requirements and infrastructure.
The material raises concerns about model changes, provider resilience and hardware constraints, but does not quantify how often these issues cause incidents or establish that AI-generated code has caused a general increase in outages. The report’s opening says outages and production bugs are becoming more common and links that trend in part to more code deployments associated with AI agents; it offers no supporting data or baseline in the supplied material. The scale of that trend and the best safeguards remain unestablished here.
Applying the Recovery Guidance
The next step for teams applying the excerpt is to define acceptable data loss and downtime, then choose a recovery pattern that matches those limits and the resources available. That work includes checking whether a fallback site can obtain the necessary compute, rehearsing how data and traffic would be restored, and identifying the service’s dependence on external models and vendors.
The Pragmatic Engineer says its feature covers only portions of two chapters from a book with 16 chapters. Its report points readers to the publisher for the full book; the source material does not specify a further release event or a scheduled follow-up.
Key Questions
What is the news development?
The Pragmatic Engineer published excerpts from Sam Newman’s 2026 book on distributed-system resilience, including disaster recovery and AI-related reliability issues.
What do RPO and RTO mean?
RPO is the maximum period of data loss a system can accept. RTO is the maximum acceptable time the system stays unavailable after an incident.
Which recovery approaches does the report mention?
The excerpt identifies backup and restore, pilot light, warm failover and multisite active-active as options, with different costs, complexity and recovery characteristics.
What AI-related risks does the report discuss?
It points to model-provider resilience, models changing, observability and uncertainty about vendors. It also discusses CPU and GPU availability as a constraint on recovery planning.
Does the report show that AI coding tools are causing more outages?
No supporting statistics or baseline are included in the supplied report. Its opening links more deployments associated with AI agents to more bugs and incidents, but that is presented without quantified evidence in the material provided.
Source: rss
