Reliability
Reliability is the ability of a system or process to perform its intended functions consistently and dependably under specified conditions and over time.
Inside AI
Reliability is the degree to which a system performs its intended functions consistently and correctly under specified conditions.
- How it is used
- Frames whether AI predictions and consequential automated decisions can be trusted in practice.
- How deep it goes
- Explains reliability through error consequences and validation needs, without providing a testing procedure or reliability metric.
- Examples
- Uses wrongly declined online payments and military automation incidents to show the costs of unreliable decisions.
- What you can do after
- Evaluate/understand AI reliability by considering both missed detections and harmful mistaken actions before trusting consequential decisions.
Apache Kafka in Action
Reliability is the likelihood that a system continues to perform correctly over a specified period and operating conditions.
- How it is used
- Frames Kafka reliability as preserving durable, available, and consistent messaging when brokers or other components fail.
- How deep it goes
- Defines three concerns and explains a page-cache acknowledgment trade-off, while pointing toward detailed treatment of ACKs, transactions, and replication.
- Examples
- Uses a three-broker failure scenario, page-cache writes acknowledged before disk commit, and a vertical-versus-horizontal scaling comparison.
- What you can do after
- Operate/configure a Kafka deployment with separated brokers and adequate capacity, then monitor replication health and client problems.
Effective Data Analysis
Reliability is the consistency or stability of a measurement across repeated observations, items, or conditions.
- How it is used
- Supports trustworthy analytics, stakeholder decisions, and evidence-based practices by checking whether measurements remain consistent.
- How deep it goes
- Defines three reliability forms and gives procedural detail for test-retest analysis, while illustrating internal consistency through Cronbach’s alpha.
- Examples
- Uses depression scales, NPS responses, qualitative topic coding, employee well-being questionnaires, and weather data for illustration.
- What you can do after
- Evaluate measures by selecting test-retest, inter-rater, or internal-consistency checks suited to the measurement and its intended use.
Building Reliable AI Systems
The ability of an AI system to behave dependably and maintain accurate, safe performance under its intended operating conditions.
- How it is used
- Frames the gap between benchmark capability and dependable answers, actions, and operations in deployed AI systems.
- How deep it goes
- Defines reliability through output, action, efficiency, fairness, and long-term quality dimensions, then identifies system-level techniques without specifying a single metric.
- Examples
- Contrasts SWE-bench results with unfamiliar codebases and uses a multipart retail query to illustrate coordinated, fault-tolerant agents.
- What you can do after
- Build it from scratch across output, agent, and operations layers, then evaluate and monitor production quality.
Read how Building Reliable AI Systems explains reliability →
Generative AI in Action
Reliability is the ability of a system to perform its required functions consistently and correctly over time and under expected conditions.
- How it is used
- Frames reliability as a production requirement for GenAI applications that must meet business needs despite failures and changing demand.
- How deep it goes
- Identifies operational practices including monitoring, validation, resilience planning, and load testing, without specifying numerical targets or a failover procedure.
- Examples
- Uses monitoring with Prometheus and Grafana and load simulation with Apache JMeter or Locust as concrete operational options.
- What you can do after
- Operate/configure it by planning failure responses, monitoring behavior, validating model outputs, and testing application performance under load.
Starting Data Analytics with Generative AI and Python
Reliability is the degree to which a system produces accurate and timely results consistently under varying conditions.
- How it is used
- Measures whether data-analysis systems deliver accurate, timely results consistently across conditions.
- How deep it goes
- Defines reliability and assesses failure or inaccuracy frequency and severity, without prescribing a calculation.
- Examples
- Contrasts observatory, hospital-monitoring, and chemical-plant systems to show how reliability priorities vary by domain.
- What you can do after
- Evaluate reliability by checking accuracy, timeliness, and the frequency and severity of failures or inaccuracies.
Read how Starting Data Analytics with Generative AI and Python explains reliability →
Machine Learning System Design
Reliability is the ability of a system to continue performing its required functions consistently despite faults, failures, or changing load.
- How it is used
- Guides production inference-platform and pipeline choices by weighing continued service against cost and failure consequences.
- How deep it goes
- Examines failure scenarios and tradeoffs but stops short of specifying technical recovery mechanisms or reliability calculations.
- Examples
- Uses unexpected load spikes, model bugs, hardware failures, and a hot-season platform outage to frame reliability decisions.
- What you can do after
- Evaluate inference designs by balancing reliability, operating cost, performance, scalability, flexibility, and security requirements.
Read how Machine Learning System Design explains reliability →
Effective Platform Engineering
System reliability is the ability of a software system to perform its required functions consistently over an expected period and operating condition.
- How it is used
- Measures platform value through system stability, observability coverage, incident performance, and reduced failures.
- How deep it goes
- Connects reliability to SRE practices, SLOs, SLIs, and error budgets without detailing their implementation.
- Examples
- Uses Epetech’s reduced downtime, fewer critical incidents, higher availability, faster recovery, and improved monitoring as evidence.
- What you can do after
- Evaluate reliability using availability, incident, recovery, telemetry, failure, and defect measures.
Read how Effective Platform Engineering explains reliability →
Architecting an Apache Iceberg Lakehouse
Reliability is the ability of a system to perform consistently and preserve correct results despite failures, retries, and interruptions.
- How it is used
- Guides ingestion architecture by matching delivery guarantees, retry behavior, source capabilities, and table-commit coordination to operational requirements.
- How deep it goes
- Explains delivery semantics, checkpointing, atomic commits, and recovery coordination, while deferring detailed resumability mechanics.
- Examples
- Uses streaming ingestion, batch archives, replayable sources, message queues, checkpointing, and Iceberg snapshot commits as concrete cases.
- What you can do after
- Evaluate ingestion guarantees and coordinate retries, checkpoints, commits, and recovery across processing and storage layers.
Read how Architecting an Apache Iceberg Lakehouse explains reliability →
Infrastructure as Code, Patterns and Practices: With examples in Python and Terraform
The capability of a system to perform its required functions consistently under stated conditions for a stated period.
- How it is used
- Uses active-passive infrastructure to limit disruption during IaC changes and protect business-critical applications.
- How deep it goes
- Defines reliability and works through regional standby construction and load-balancer failover, but only previews active-active refactoring.
- Examples
- Uses Python configuration files, separate regional environments, and weighted load-balancer services as concrete examples.
- What you can do after
- Operate/configure an active-passive deployment and redirect all production traffic to its standby environment after failure.
How Large Language Models Work
Reliability is the degree to which a system consistently produces dependable behavior and results under expected conditions.
- How it is used
- Frames reliability as a deployment criterion for deciding whether LLM applications can safely replace existing services.
- How deep it goes
- Explains accuracy failures, unexpected inputs, liability, and remediation needs, but stops short of specifying reliability measurement methods.
- Examples
- Uses FAQs, email tickets, phone support, customer retention, and trial deployments to illustrate reliability decisions.
- What you can do after
- Evaluate an LLM deployment by weighing inaccurate outputs, liability, remediation effort, and customer-impact risks.
Read how How Large Language Models Work explains reliability →
Think Distributed Systems
The ability of a system to continue providing its intended service correctly over time and under expected conditions.
- How it is used
- Frames reliability as the ability to remain responsive while failures affect a distributed system.
- How deep it goes
- Develops a definition, emergent-system model, responsiveness measures, and application-specific guarantees.
- Examples
- Uses a company analogy and component-interaction scenarios involving crashes, absence, load, and system behavior.
- What you can do after
- Evaluate reliability by defining service-level objectives, measuring error rates, and setting acceptable error budgets.
Amazon Web Services in Action, Third Edition: An in-depth guide to AWS
Reliability is the ability of a system to perform its intended functions consistently over time.
- How it is used
- Uses managed AWS services and deployment automation to keep web applications available while infrastructure or application versions change.
- How deep it goes
- Traces reliability from service defaults through rolling updates and bootstrap failures, then points toward AMIs as mitigation.
- Examples
- Demonstrates reliability through an Etherpad rolling update, bootstrap scripts, repository outages, AMIs, and cloud-init logs.
- What you can do after
- Operate/configure CloudFormation rolling updates and recognize when AMIs can reduce deployment-script failure risk.
Related concepts
- inter-rater reliabilityincludes · Inter-rater reliability is the degree of agreement or consistency among independent raters evaluating the same items, subjects, or events.
- test-retest reliabilityincludes · Test-retest reliability is the consistency of a measure when the same individuals are assessed repeatedly over time while the measured construct remains stable.
- internal consistencyincludes · Internal consistency is the degree to which components of a measure or artifact produce mutually consistent, noncontradictory results or statements.
- redundancyproduced by · Redundancy is the presence of duplicate or overlapping information, components, or resources, whether intentional or unnecessary.
- replicationaddressed by · Replication is the creation or maintenance of multiple copies of data, state, or functionality across separate locations or components.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.