Reliability Glossary
Browse 91 reliability terms defined in plain English, from the cultural dictionary of computing.
91 Reliability Terms
- ACID
- An acronym for the four properties that guarantee reliable database transactions: Atomicity (all or nothing), Consistency (data stays valid), Isolation...
- Active-Active
- Active-Active is a high-availability architecture pattern where two or more servers or data centers simultaneously handle production traffic, as opposed to...
- AI Benchmark
- An evaluation concept used to measure, inspect, or compare model behavior through ai benchmark. It influences how models are trained, evaluated, or served, and...
- AI Fairness
- An evaluation concept used to measure, inspect, or compare model behavior through ai fairness. It influences how models are trained, evaluated, or served, and...
- AI Hallucination
- An evaluation concept used to measure, inspect, or compare model behavior through ai hallucination. It influences how models are trained, evaluated, or served,...
- AI Hallucination Rate
- A measured estimate of how often an AI system produces false, invented, or unsupported claims in a given task setup.
- Availability Zone
- A physically separate data center within a cloud region, with independent power, cooling, and networking. Deploying across multiple AZs provides high...
- blast radius
- The scope of impact when a failure, misconfiguration, or security breach occurs — how many users, services, or systems are affected. Good architecture...
- blue-green deployment
- A deployment strategy that maintains two identical production environments: 'blue' (current live) and 'green' (new version). Traffic is switched from blue to...
- Bullet Proof
- A casual way to say something seems extremely robust, reliable, or hard to break. In engineering slang, it is often used optimistically, sometimes more...
- Canary Analysis
- Automated statistical comparison between a canary deployment (new version, small traffic percentage) and the baseline (current version) to detect regressions...
- canary release
- A deployment strategy where a new version is rolled out to a small subset of users or servers first, then gradually expanded if metrics look healthy. Named...
- Chaos Budget
- The amount of risk a team is willing to take with chaos experiments, analogous to error budgets. When reliability is high (error budget surplus), the chaos...
- Chaos Monkey
- A tool created by Netflix that randomly terminates production instances to ensure that engineers build resilient services that can tolerate unexpected...
- Chaos Monkey Culture
- A reliability mindset that values failure testing, resilience drills, and systems that can tolerate component loss under realistic conditions.
- Chaos Testing
- Deliberately injecting failures (network partitions, server crashes, disk full, clock skew, high latency) into a system to verify it handles them gracefully....
- circuit breaker
- A design pattern that prevents an application from repeatedly trying to execute an operation that is likely to fail, by 'tripping' after a threshold of...
- Core Service
- A service that plays a central role in a platform and is widely depended on by other services, jobs, or customer-facing features. Core services usually require...
- Dead Letter Queue
- A special queue where messages that can't be processed (after exceeding retry limits) are stored for later analysis instead of being discarded. Prevents poison...
- Defensive Coding
- Writing code to guard against bad input, misuse, edge cases, and unexpected states rather than assuming everything upstream behaves perfectly. It often...
- Defensive Programming
- A coding approach that assumes inputs, state, and dependencies may be invalid or hostile and therefore validates, bounds, and handles errors carefully. In...
- Downtime
- A period when a system, service, or component is unavailable or not functioning as intended for users. It is one of the most direct signals of operational...
- Downtime Culture
- A team's shared attitude toward outages, maintenance windows, and service interruptions, including what is tolerated, communicated, and learned from. Downtime...
- Error Boundary
- An Error Boundary is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
- error budget
- The acceptable amount of unreliability for a service over a given period, calculated as 1 minus the SLO. A service with a 99.9% availability SLO has a 0.1%...
- Error Budget
- The allowed amount of unreliability based on SLO targets — if your SLO is 99.9% availability, your error budget is 0.1% (about 43 minutes per month). When the...
- Error Budget Culture
- A culture that takes error budgets seriously as a way to balance reliability and change velocity. In healthy teams, error budget culture means service quality...
- Error Code
- An Error Code is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code easier...
- Error Correction
- An Error Correction is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
- Error Handling
- An Error Handling is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
- Error Recovery
- An Error Recovery is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
- Error Type
- An Error Type is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code easier...
- Evaluation Metric
- An evaluation concept used to measure, inspect, or compare model behavior through evaluation metric. It influences how models are trained, evaluated, or...
- Factuality
- An evaluation concept used to measure, inspect, or compare model behavior through factuality. It influences how models are trained, evaluated, or served, and...
- Failover
- Failover is the automatic or manual process of switching to a redundant system, server, network, or component when the primary one fails, ensuring continuous...
- Failure Analysis
- The systematic investigation of why something failed, with the goal of understanding causes, contributing factors, and improvements. In engineering culture...
- Failure Mode
- A specific way in which a system, process, or component can fail. In engineering culture, thinking in failure modes helps teams prepare for real breakage...
- Fairness Metric
- An evaluation concept used to measure, inspect, or compare model behavior through fairness metric. It influences how models are trained, evaluated, or served,...
- Fault Tolerant Culture
- A culture that assumes failures will happen and therefore values resilience, redundancy, graceful degradation, and calm recovery rather than magical thinking....
- Five Nines
- A measure of system availability equal to 99.999%, allowing no more than 5 minutes and 15 seconds of downtime per year.
- Flake
- A test, build, or job that passes and fails intermittently without a meaningful code change, usually because of timing, environment variance, hidden...
- Flaky
- Unreliable or inconsistent in a way that passes sometimes and fails other times without a meaningful code change. The term is most often used for tests,...
- Game Day
- A planned exercise where teams simulate failures, incidents, or unusual conditions to practice response and test resilience. In operations culture game days...
- Hallucination Mitigation
- An evaluation concept used to measure, inspect, or compare model behavior through hallucination mitigation. It influences how models are trained, evaluated, or...
- Health Check
- An endpoint (typically /health or /healthz) that reports whether a service is running and ready to accept traffic. Used by load balancers, orchestrators, and...
- Heroic Engineering
- A culture that relies on last-minute heroics, individual rescue acts, and unsustainable effort instead of resilient systems and planning.
- Human Evaluation
- An evaluation concept used to measure, inspect, or compare model behavior through human evaluation. It influences how models are trained, evaluated, or served,...
- idempotency
- The property where performing an operation multiple times produces the same result as performing it once. In APIs, idempotent operations are safe to retry...
- Idempotency Key
- A unique token sent with an API request to ensure that retrying the same operation produces the same result without side effects. Critical for payment APIs and...
- Idempotent
- An operation that produces the same result whether executed once or multiple times. Critical in distributed systems where network failures mean you might...
- Idempotent Pipeline
- A data pipeline that produces the same output regardless of how many times it's run with the same input. Achieved through techniques like UPSERT/MERGE,...
- Incident Post-Mortem
- A structured analysis written after an outage or operational failure to document timeline, impact, root causes, contributing factors, and follow-up actions....
- Just Works
- A simpler shorthand for saying a tool or system behaves with satisfying reliability and low friction. In engineering slang, people love the phrase because...
- LLM Reliability
- The extent to which a large language model system behaves consistently and acceptably across repeated use, edge cases, and operational conditions. Reliability...
- Load Shedding
- The deliberate rejection of a portion of incoming requests when a system is under extreme load, preserving capacity to successfully serve the remaining traffic...
- MTBF
- Mean Time Between Failures. The average time a repairable system operates between successive failures, calculated as total uptime divided by the number of...
- MTTF
- Mean Time To Failure. The average time a non-repairable component or system operates before it fails for the first time.
- MTTR
- Mean Time To Recovery (or Repair). The average elapsed time from when a failure is detected to when the system is fully restored to normal operation.
- Nines
- Shorthand for uptime percentages. 'Three nines' = 99.9%, 'five nines' = 99.999%. Each additional nine is exponentially harder and more expensive to achieve....
- Oncall Burnout
- Exhaustion caused by repeated pager duty, interrupted sleep, constant incident pressure, or poor production hygiene.
- On-Call Culture
- The norms and expectations surrounding pager duty, incident ownership, and how teams support engineers who handle production issues.
- Ounce of Prevention
- A proverb-style reminder that a little preventive work can save a lot of later pain. In engineering slang, it usually gets invoked when someone is arguing for...
- Outbox Pattern
- A pattern for reliable event publishing in microservices: instead of directly publishing to a message broker (which can fail independently of the database),...
- Packet Loss
- The failure of one or more network packets to reach their destination, caused by congestion, faulty hardware, signal degradation, or deliberate QoS dropping....
- postmortem
- A structured document written after an incident that analyzes what happened, why, and how to prevent recurrence. Good postmortems are blameless (focusing on...
- Postmortem Culture
- The way a team handles incident reviews, including whether postmortems are blameless, useful, and tied to real follow-through.
- Post-Mortem Culture
- The shared norms around how an organization investigates failures after they happen, including whether reviews are blameful or learning-oriented and whether...
- Production Incident
- A failure or degradation in a live system that affects real users, business operations, or service reliability. The term usually implies formal response,...
- Production Ready
- Good enough to run for real users under real operational conditions, with acceptable reliability, observability, security, and support characteristics. Teams...
- Replication
- The process of copying and maintaining data across multiple database instances or nodes so that every replica holds the same information, improving...
- Retry Storm
- A cascading failure where many clients simultaneously retry failed requests, overwhelming an already struggling service and preventing it from recovering —...
- Robustness
- The ability of an AI system to remain useful and stable across noisy inputs, edge cases, shifting conditions, and small perturbations. Robustness matters...
- Safety-Critical
- Describing a system whose failure could result in loss of life, significant injury, or major environmental damage, requiring rigorous verification, formal...
- Service Level
- A defined standard for performance, availability, response time, or other operational commitments in a service relationship. Service levels matter because...
- Single Point of Failure Slang
- Informal language for any person, service, component, or process whose failure can take down a larger system. In engineering slang, calling something a single...
- Site Reliability
- The discipline of keeping services reliable through engineering approaches to operations, including automation, observability, incident response, and...
- Six Nines
- A shorthand boast or target for extremely high availability, roughly 99.9999 percent uptime. In engineering slang, six nines often carries a skeptical tone...
- SLA
- Service Level Agreement — a formal commitment defining the expected level of service (usually uptime). 99.9% sounds great until you realize it allows 8.7 hours...
- Sled Dog
- A dependable system or person that keeps pulling heavy operational load without much glamour. In engineering slang, a sled dog is respected because the boring...
- SLI
- Service Level Indicator — a carefully chosen metric that quantitatively measures some aspect of a service's reliability, such as request latency, error rate,...
- SLI/SLO/SLA
- Three related reliability concepts: SLI (Service Level Indicator) is the metric measured (e.g., request latency p99). SLO (Service Level Objective) is the...
- SLO
- Service Level Objective — a target value or range for a service's reliability, expressed as a percentage (e.g., 99.9% availability). SLOs provide a...
- Software Testing
- The practice of executing software under controlled conditions to verify that it behaves as expected and to discover defects. Encompasses multiple levels —...
- SRE
- Site Reliability Engineering (SRE) is a discipline, originated at Google, that applies software engineering principles to infrastructure and operations. SRE...
- Status Page Culture
- The norms around how openly and accurately a company communicates outages, degradation, and incident progress to customers.
- Structured Output AI
- An AI pattern where responses are constrained to a defined structure such as JSON, a schema, or typed fields.
- TCP
- TCP, or Transmission Control Protocol, is a connection-oriented transport protocol that provides reliable, ordered delivery of data between applications over...
- Transaction
- A sequence of database operations treated as a single atomic unit — either all succeed (commit) or all are rolled back. Transactions provide ACID guarantees:...
- Trust but Verify
- A principle that accepts cooperation or assumptions provisionally but still requires evidence, checks, or monitoring before relying on them completely....
- Warm Standby
- A redundancy approach where a secondary system is running and receiving data updates but is not serving live traffic, allowing it to be promoted to primary...
- Webhook Retry
- The retry policy used when a webhook delivery fails (non-2xx response or timeout). Good implementations use exponential backoff (e.g., 1min, 5min, 30min, 2hr,...
Related Topics
- Sre (17 terms in common)
- Operations (12 terms in common)
- Ai (10 terms in common)
- Evaluation (9 terms in common)
- Devops (7 terms in common)
- Debugging (6 terms in common)
- Backend (6 terms in common)
- Slang (6 terms in common)
- Testing (5 terms in common)
- Metrics (5 terms in common)
- Learning (4 terms in common)
- Infrastructure (4 terms in common)
- Pattern (4 terms in common)
- Api (3 terms in common)
- Deployment (3 terms in common)
- Architecture (3 terms in common)
- Culture (3 terms in common)
- Robustness (2 terms in common)
- Incident Response (2 terms in common)
- Availability (2 terms in common)