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 reliability challenges in AI-powered software. The material discusses recovery targets, failover options, constrained infrastructure, model-provider risks and observability; it is guidance from a book excerpt, not evidence of a measured rise in outages.
The Pragmatic Engineer has published excerpts from software engineer Sam Newman’s 2026 book, Building Resilient Distributed Systems, outlining how teams can plan for major outages and reliability risks in AI-powered software. The material focuses on setting recovery targets, choosing disaster-recovery patterns and accounting for changing AI models and provider dependencies.
Newman’s excerpt frames disaster recovery (DR) as preparation for failures that affect an entire site, such as a cloud region or data center, rather than an individual service instance. It describes events including fire, flooding, power problems and failures in cloud infrastructure as risks outside an engineering team’s direct control. The report says the book covers recovery objectives and patterns including backup-and-restore, pilot light, warm failover and multisite active-active arrangements.
Two measures help teams state their recovery requirements. A recovery point objective (RPO) sets the maximum acceptable amount of data loss, expressed as the time since data was last backed up or synchronized. A recovery time objective (RTO) sets how long a service can remain unavailable after an incident. Newman notes that returning a system to service may involve restoring data, making computing resources available and redirecting traffic.
The report also highlights challenges specific to AI-powered software: dependence on model providers, models changing over time, and the need for observability. The listed approaches to uncertainty include using multiple models or vendors, making costs transparent and reviewing vendors regularly. The excerpt says ongoing CPU and GPU shortages mean teams cannot assume infrastructure will always be available, and calls for prioritizing critical services.
Recovery Targets Shape Outage Plans
For teams running customer-facing software, a disaster plan is only useful if it reflects how much downtime and data loss the business can tolerate. RPO and RTO turn that tolerance into targets that can guide engineering decisions, recovery procedures and investment. The excerpt stresses a trade-off: reducing acceptable downtime or data loss can raise both cost and operational complexity.
AI dependencies add another source of uncertainty. If a system relies on a particular model, provider or supply of specialized computing capacity, changes or shortages may affect service continuity. Planning for provider and model variability, and monitoring system behavior, can help teams understand their dependencies and prioritize the functions they need to restore first. These are recommendations in the book excerpt; the report does not quantify their effect on outages.
As an affiliate, we earn on qualifying purchases.
From Service Failures to Site Loss
The published material comes from Chapter 10 of Newman’s book, alongside discussion of AI-related resilience topics drawn from other parts of the book. The Pragmatic Engineer says Newman spent three years writing it, rather than the two he had planned, and revised sections to address AI. The report identifies the book as published by O’Reilly Media in 2026 and says the excerpt is used with permission.
The source introduces “site” as a general term for the location of computing infrastructure, covering a public-cloud region as well as a data center in private-cloud or on-premises environments. It presents disaster recovery as distinct from handling failures within a single service instance: site-wide events can defeat protections that work at a smaller scale. The source’s excerpt begins explaining recovery objectives but does not provide detailed implementation instructions for every pattern it lists.
““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, author of Building Resilient Distributed Systems
As an affiliate, we earn on qualifying purchases.
Outage Trends Need More Evidence
The report says outages are becoming more common and links this in part to more code deployments associated with AI agents and to more bugs reaching production. However, the supplied material provides no outage statistics, time series or comparison baseline to establish the size or cause of that trend. It should be treated as the report’s characterization, not as a quantified finding.
The excerpt also does not specify which recovery pattern suits a particular system, what its implementation would cost, or what recovery targets organizations should adopt. Those decisions depend on each service’s requirements and infrastructure. The source identifies multiple-model and multi-vendor approaches as tactics for managing uncertainty but does not provide comparative results showing how well they perform.
As an affiliate, we earn on qualifying purchases.
Teams Must Set Their Own Targets
The immediate next step for a team applying the guidance is to define acceptable data loss and downtime for its own services, then assess whether its current backup, failover and traffic-redirection plans can meet those targets. The excerpt presents a range of DR patterns but leaves the selection to the organization’s needs, resources and tolerance for cost and complexity.
For AI-dependent systems, the report points toward continued attention to provider arrangements, model changes, infrastructure constraints, transparent costs and observability. The Pragmatic Engineer says the published material covers only a portion of the book’s 16 chapters. It does not give a schedule for further excerpts or report any new industry-wide resilience measures.
distributed system resilience tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the news development?
The Pragmatic Engineer published authorized excerpts from Sam Newman’s 2026 book, Building Resilient Distributed Systems, covering disaster recovery and reliability concerns for AI-powered software.
What do RPO and RTO mean?
RPO is the maximum acceptable time between the last backup or synchronization and an incident, and indicates the potential data-loss window. RTO is the maximum acceptable time a service can be unavailable while it is restored.
Which disaster-recovery approaches does the excerpt list?
It names backup-and-restore, pilot light, warm failover and multisite active-active as patterns. The supplied excerpt does not give enough detail to prescribe one for a particular system.
What AI-specific risks does the report identify?
It points to model-provider resilience, models changing over time, infrastructure availability and observability. It also lists multiple models or vendors, transparent costs and regular vendor analysis as ways to handle uncertainty.
Does the source prove AI coding tools are causing more outages?
No. The report says increased deployments and bugs associated with AI-generated code are contributing to more outages, but the supplied material includes no data or comparison baseline to independently establish the trend or quantify the causes.
Source: rss
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
