TL;DR
Get business pricing on networking and server gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
The Pragmatic Engineer published excerpts from Sam Newman’s 2026 book, Building Resilient Distributed Systems, covering disaster recovery and resilience for AI-powered software. The material emphasizes setting recovery targets, weighing recovery costs, and planning for changing models, vendor constraints and infrastructure shortages.
The Pragmatic Engineer has published excerpts from Sam Newman’s new book, Building Resilient Distributed Systems, outlining ways software teams can plan for major outages and the risks of relying on AI services. The material covers disaster-recovery targets and patterns, as well as model-provider resilience, observability and vendor uncertainty.
In the report, the publication says Newman spent three years writing the book, one year longer than he initially planned, and revised its material to account for AI. The excerpts span two chapters in a book with 16 chapters, according to the report. The publisher is O’Reilly Media; the excerpts are used with permission.
The disaster-recovery chapter describes two measurements for setting recovery requirements. Recovery point objective (RPO) is the maximum acceptable period of potential data loss, based on how recently data was backed up or synchronized. Recovery time objective (RTO) is the acceptable time a system can remain unavailable after an incident, including the time needed to restore data, make computing resources available and redirect traffic.
The report also lists backup-and-restore, pilot-light failover, warm failover and multisite active-active as disaster-recovery patterns. These approaches involve trade-offs in cost, complexity and recovery. The AI-focused material raises questions about changes to models, provider resilience and the need to observe systems in operation, themes also relevant to how software development is changing with AI. It also recommends attention to vendor constraints, infrastructure availability and the prioritization of critical services.
Recovery Plans Must Include AI Dependencies
AI-powered software can depend on external models and providers as well as the infrastructure that runs an organization’s own services. The report’s central practical point is that resilience planning needs to account for those dependencies, not only failures inside individual software components. A provider change or service interruption can affect an application even when its own systems remain available.
For readers responsible for production systems, RPO and RTO make recovery expectations measurable. Teams can use them to discuss how much data loss and downtime are acceptable, then select a recovery design that fits those limits. The excerpts do not prescribe one approach for every system: faster recovery and lower data loss can require more spending and operational complexity.
The report also points to a supply-side risk. It says ongoing GPU and CPU shortages mean teams cannot assume infrastructure will always be available. That makes vendor capacity and service priorities relevant to planning, particularly for services an organization considers critical.
As an affiliate, we earn on qualifying purchases.
From Site Outages to Model Changes
The book’s disaster-recovery chapter addresses failures that affect a large part of a system, such as a fire, flood, electrical supply problem or error in underlying cloud infrastructure. These events differ from failures limited to an individual service instance: they can make an entire site unavailable. Newman uses “site” as a general term covering a cloud region, data center or comparable location.
The report places that material alongside a discussion of AI-powered software. It says the book considers model-provider resilience, the challenge of models changing, and the need for observability. For handling uncertainty, the report identifies tactics including using more than one model or vendor, making costs transparent and reviewing vendors frequently. These are presented as topics in the excerpted book, not as evidence that any particular design guarantees uninterrupted service.
The publication frames the subject as timely in light of more software deployments and production bugs. That is the report’s explanation for the book’s relevance; it does not provide outage statistics or data establishing how much AI-assisted coding has changed incident rates.
“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.
Specific Recovery Targets Remain System-Dependent
The excerpts do not provide recommended RPO or RTO values for particular types of service. Those targets depend on how much data loss and downtime an organization can accept, and the report does not compare the cost or performance of the listed recovery patterns in specific deployments.
It is also unclear from the material provided how often AI providers change models, how frequently shortages affect a given customer, or what service guarantees individual vendors offer. The report gives no incident data to quantify its claims about outages, deployments or bugs. Its discussion is guidance and book material, rather than a documented evaluation of specific systems or vendors.
cloud backup and restore solutions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Teams Can Set Targets Before Choosing Patterns
The next step for a team applying the material is to define acceptable data loss and downtime, then assess how its services and suppliers could meet those targets during a site-wide failure. That work includes identifying which services are critical, understanding infrastructure constraints, and deciding whether recovery should rely on backups, standby capacity or multiple active sites.
For AI-dependent systems, teams can also examine how applications respond to model or provider changes, what they can observe during a disruption, and whether using multiple vendors is appropriate. The report does not announce a rollout or an industry standard. It directs readers to Newman’s book for fuller treatment; the supplied material gives no specific follow-up date or new milestone.
failover and high availability infrastructure
As an affiliate, we earn on qualifying purchases.
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, Building Resilient Distributed Systems, including material on disaster recovery and AI-powered software.
What do RPO and RTO mean?
RPO is the maximum period of potential data loss a system can tolerate. RTO is the acceptable time for the system to remain unavailable after an incident.
Which disaster-recovery approaches does the report name?
The excerpts list backup-and-restore, pilot-light failover, warm failover and multisite active-active. They involve different trade-offs in cost, complexity and recovery.
What AI-related risks does the material discuss?
It addresses model-provider resilience, model changes, observability and vendor uncertainty, alongside infrastructure constraints such as GPU and CPU availability.
Does the report give universal recovery targets?
No. The supplied excerpts explain RPO and RTO but do not set universal targets. Appropriate limits depend on each system’s acceptable downtime, potential data loss, costs and operational needs.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
