TL;DR
Get travel and outdoor gear 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 has published excerpts from Sam Newman’s 2026 book, Building Resilient Distributed Systems, covering disaster recovery and resilience for AI-powered software. The material outlines recovery targets, failover approaches and risks such as changing models and limited computing capacity; it does not provide outage statistics or claim that AI has caused a measured rise in incidents.
The Pragmatic Engineer has published excerpts from Sam Newman’s new book, Building Resilient Distributed Systems, addressing disaster recovery and the design of AI-powered software. The material sets out practical recovery objectives and failover patterns while discussing risks tied to model providers, model changes and the availability of computing resources.
Newman’s excerpts explain two measures used to plan recovery after a major outage. A recovery point objective (RPO) sets the maximum acceptable period of potential data loss, based on how recently data was backed up or synchronized. A recovery time objective (RTO) sets how long a service can remain unavailable while teams restore data, obtain computing resources and redirect traffic.
The book describes a range of disaster-recovery patterns: backup and restore, pilot-light failover, warm failover and multisite active-active operation. These approaches differ in cost, complexity and how quickly services can return. The excerpt frames the choice as a trade-off: tighter limits on downtime or data loss generally require greater spending and operational effort.
The report also highlights concerns specific to AI-powered systems. Newman’s material addresses resilience when a model provider is unavailable, the challenge of models changing, and the need for observability. It recommends considering multiple models or vendors, making costs transparent and reviewing vendors regularly. The report says ongoing GPU and CPU shortages mean teams should not assume infrastructure will always be available, and should prioritize critical services.
Recovery Targets Shape Outage Readiness
The guidance matters because a recovery plan is only useful if it reflects the service’s actual tolerance for disruption and data loss. Defining RPO and RTO gives engineering teams concrete targets against which to select backups, failover capacity and recovery procedures. Without those targets, organizations may pay for more resilience than they need or discover during an incident that recovery takes longer than users and the business can accept.
AI adds dependencies that conventional recovery planning may not fully capture. A service can rely on external model providers and on access to specialized computing resources; either may constrain its operation. Planning for provider failure and model changes, and monitoring system behavior, can help teams identify where their fallback plans depend on assumptions that might not hold. The excerpts offer engineering guidance, not evidence that any particular company has adopted these measures or that they prevent outages.
As an affiliate, we earn on qualifying purchases.
From Site Outages to Model Risks
The excerpts come from two chapters of Newman’s book, published by O’Reilly Media in 2026. The Pragmatic Engineer report says Newman took three years to complete the book, rather than the two he had initially planned, and reworked sections to address AI. The published material includes discussion of catastrophic failures such as fire, flooding, electrical problems and cloud infrastructure errors.
Newman uses the term site to cover locations where computing infrastructure runs, including cloud regions and data centers. The report introduces the book as timely for software engineers amid concern about more code deployments and production bugs as AI agents write code. That is the report’s characterization; it gives no underlying incident data or comparison period establishing how much outages or bugs have changed.
“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
As an affiliate, we earn on qualifying purchases.
Outage Trends Remain Unquantified
The report does not provide data on whether outages have become more frequent, how many incidents are linked to AI-generated code, or how those figures compare with an earlier period. Its comments about more deployments, bugs and incidents are presented as the report’s framing, not as quantified findings. The supplied excerpt also ends during its explanation of RTO, so it does not show every detail of the book’s disaster-recovery discussion.
It is also unclear which recovery patterns or AI-provider strategies Newman considers appropriate for particular systems; the material emphasizes that costs, complexity and acceptable disruption vary by service. No company case studies, implementation results or performance comparisons are included in the supplied report.
As an affiliate, we earn on qualifying purchases.
Teams Must Set Service-Specific Targets
The immediate next step for engineering teams applying the guidance is to define acceptable data loss and downtime for each service, then test whether their chosen recovery approach can meet those RPO and RTO targets. That work includes accounting for data restoration, available capacity and traffic redirection, not just the existence of backups.
For AI-powered services, the report points to ongoing review of model and infrastructure dependencies, vendor constraints, costs and monitoring. It does not announce a formal follow-up or a specific implementation deadline. The excerpts are from Newman’s book, which the report says is available from its publisher.
distributed system monitoring tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What new development does the report describe?
The Pragmatic Engineer published excerpts from Sam Newman’s 2026 book, Building Resilient Distributed Systems, on disaster recovery and AI-system resilience.
What do RPO and RTO measure?
RPO measures the maximum acceptable period of potential data loss. RTO measures the acceptable time a system can remain unavailable after an incident.
Which disaster-recovery approaches are covered?
The excerpts name backup and restore, pilot-light failover, warm failover and multisite active-active. They differ in cost, complexity and recovery speed.
What AI-related risks does the material discuss?
It addresses model-provider resilience, model changes, observability and constraints on GPU and CPU availability. The report also discusses using multiple models or vendors and reviewing vendor relationships.
Does the report prove AI has caused more outages?
No. The report describes concern about more deployments, bugs and incidents, but supplies no incident statistics, comparison baseline or causal evidence showing that AI-generated code has increased outages.
Source: rss
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
