Reference

DevOps, Cloud & AI Glossary

Plain-English definitions for the terms that come up constantly across DevOps, Kubernetes, cloud infrastructure, APIs, and AI engineering. Every term links to a full guide when one exists — click a heading below to jump straight to a section.

DevOps & CI/CD

CI/CD #: Continuous Integration and Continuous Delivery/Deployment — the practice of automatically building, testing, and shipping code on every change instead of in large manual batches. Read more →
GitOps#: An operating model where a Git repository is the single source of truth for infrastructure and deployments — an automated controller continuously reconciles the live system to match what's committed, and every change ships as a pull request.
Infrastructure as Code (IaC)#: Defining servers, networks, and other infrastructure in version-controlled configuration files instead of clicking through a console, so changes are reviewable, repeatable, and auditable. Read more →
Immutable Infrastructure#: A pattern where servers or containers are never modified in place after deployment — any change ships as a brand-new image or instance that replaces the old one, eliminating configuration drift.
Configuration Drift#: The gradual divergence between a system's actual state and the configuration it was supposed to have, usually caused by manual, undocumented changes made outside the normal deployment process.
Blue-Green Deployment#: A release strategy that runs two identical production environments — one live (blue), one idle (green) — and switches traffic to the new version instantly, with an immediate rollback path if something's wrong.
Canary Release#: Rolling out a new version to a small slice of real traffic first, watching error rates and latency, and only widening the rollout once it proves healthy.
Idempotency#: A property of an operation that produces the same result no matter how many times it's applied — critical for safe retries in APIs, deployments, and distributed systems where a request might be sent more than once.
Observability #: The ability to understand a system's internal state from the outside, through logs, metrics, and traces — going beyond monitoring's known failure modes to answer questions you didn't anticipate asking. Read more →
SRE (Site Reliability Engineering)#: An engineering discipline that applies software practices to operations, using error budgets and SLOs to balance reliability against the pace of shipping new features. Read more →
Chaos Engineering#: Deliberately injecting failures into a production or production-like system to verify it actually survives the outages it's supposed to be resilient to, before a real one finds the gap for you.

Containers & Kubernetes

Container #: An isolated process that packages an application with its dependencies, sharing the host machine's kernel rather than virtualizing an entire OS — what makes containers fast to start and light on resources compared to VMs. Read more →
Kubernetes #: An open-source system for running containers across a cluster of machines — it schedules workloads, restarts failed containers, scales replicas, and rolls out updates without downtime. Read more →
Pod#: The smallest deployable unit in Kubernetes: one or more containers that share a network namespace and storage, always scheduled together on the same node.
Kubernetes Service#: A stable network endpoint that load-balances traffic across a changing set of Pods, so other workloads never need to know individual Pod IP addresses.
Ingress#: A Kubernetes resource that manages external HTTP(S) access to services inside a cluster, typically handling routing, TLS termination, and virtual hosting through a single load balancer.
Helm Chart#: A packaged, templated bundle of Kubernetes manifests — the standard way to install and version complex applications on a cluster with a single command instead of hand-applying dozens of YAML files.
Sidecar Pattern#: Running a helper container alongside a main application container in the same Pod to add a cross-cutting concern — logging, a proxy, TLS — without modifying the main application's code.
Service Mesh#: An infrastructure layer, usually built from sidecar proxies, that handles service-to-service traffic — retries, mutual TLS, load balancing, and observability — without any of that logic living in application code.
Horizontal Pod Autoscaler (HPA)#: A Kubernetes controller that automatically adds or removes Pod replicas based on observed CPU, memory, or custom metrics, so capacity tracks real demand instead of a fixed, guessed number.

Cloud & Infrastructure

CDN (Content Delivery Network)#: A distributed network of servers that caches and serves content from a location physically close to each visitor, cutting latency compared to serving every request from one origin server. Read more →
Edge Computing#: Running code close to the user — on CDN points of presence or regional nodes — instead of a single centralized server, trading some flexibility for much lower latency.
Serverless #: A model where you deploy functions or services without managing servers — the platform handles provisioning and scaling, and you're billed for actual usage rather than idle capacity. Read more →
Cold Start#: The extra latency a serverless function or scale-to-zero database incurs on its first request after being idle, while the platform provisions a fresh execution environment.
Multi-Cloud#: Running workloads across more than one cloud provider — deliberately, for redundancy or best-of-breed services, as opposed to being locked into a single vendor's ecosystem.
Vendor Lock-In#: The cost and difficulty of switching away from a cloud provider or tool once you've built deeply against its proprietary APIs, formats, or managed services.
Load Balancer #: A component that distributes incoming traffic across multiple backend servers, so no single instance is overwhelmed and traffic keeps flowing if one instance fails. Read more →
Reverse Proxy #: A server that sits in front of one or more backend services, forwarding client requests to them — used for load balancing, TLS termination, caching, and hiding backend topology. Read more →
VPC (Virtual Private Cloud)#: An isolated, private slice of a public cloud provider's network where you control IP ranges, subnets, and routing — the boundary that keeps your resources network-isolated from other tenants.
IAM (Identity and Access Management)#: The system of users, roles, and policies that controls who — human or machine — can do what to which resources in a cloud account, ideally scoped to least privilege. Read more →
Zero Trust #: A security model that assumes no request is trustworthy by default, network location included, and instead verifies every request's identity and authorization explicitly, every time. Read more →

APIs & Architecture

REST #: An architectural style for web APIs built around resources, standard HTTP methods, and statelessness — the default choice for most public and web-facing APIs. Read more →
GraphQL #: A query language for APIs that lets clients request exactly the fields they need in one round trip, instead of over- or under-fetching from fixed REST endpoints. Read more →
gRPC #: A high-performance RPC framework using Protocol Buffers over HTTP/2, favored for fast, strictly-typed internal service-to-service communication over REST or GraphQL. Read more →
Webhook#: A callback: instead of a client repeatedly polling for updates, a server sends an HTTP request to a client-provided URL the moment an event happens.
API Gateway#: A single entry point that sits in front of one or more backend services, handling routing, authentication, and rate limiting so individual services don't each reimplement them.
Rate Limiting #: Capping how many requests a client can make in a given time window, to protect a service from being overwhelmed and to keep usage fair across clients. Read more →
Message Queue #: A component that decouples producers from consumers by holding messages until a consumer is ready to process them — smoothing traffic spikes and letting services fail independently. Read more →
Event-Driven Architecture#: A design where services communicate by publishing and reacting to events rather than calling each other directly, reducing tight coupling at the cost of harder-to-trace flows.
Circuit Breaker#: A pattern that stops calling a failing downstream dependency after enough errors, failing fast instead of piling up timeouts, and periodically retries to see if it has recovered.

AI & Machine Learning

LLM (Large Language Model)#: A neural network trained on huge amounts of text to predict the next token, which turns out to be a general-purpose engine for generating, summarizing, and reasoning about language. Read more →
Embeddings #: Numeric vectors that represent the meaning of text, images, or other data, positioned so that similar things end up close together in vector space — the foundation of semantic search. Read more →
RAG (Retrieval-Augmented Generation)#: Feeding an LLM relevant documents retrieved from a knowledge base at query time, so it answers grounded in your actual data instead of only what it memorized during training. Read more →
Fine-Tuning#: Further training a pre-trained model on a smaller, task-specific dataset to specialize its behavior, as opposed to prompting a general-purpose model at inference time.
Prompt Engineering#: Deliberately structuring the instructions and context given to an LLM to reliably get the output you want, without changing the model's weights.
Hallucination#: When an LLM generates output that's fluent and confident but factually wrong or unsupported by its input — a known failure mode, not a bug in the traditional sense.
Vector Database #: A database purpose-built to store embeddings and run fast nearest-neighbor similarity search over them — the retrieval half of most RAG systems. Read more →
Context Window#: The maximum amount of text, measured in tokens, an LLM can process in a single request — everything beyond it gets truncated or must be summarized/retrieved instead.
Model Context Protocol (MCP)#: An open protocol that standardizes how AI applications connect to external tools and data sources, so a model can call a shared set of tools instead of every app inventing its own integration. Read more →
Vibe Coding #: Building software primarily by describing intent to an AI coding assistant and iterating on what it produces, rather than writing most of the code by hand. Read more →
MLOps #: The practices and tooling for deploying, monitoring, and retraining machine learning models reliably in production — CI/CD's counterpart for the ML lifecycle. Read more →
AI Agent#: A system that uses an LLM in a loop to plan, call tools, and take multi-step actions toward a goal, rather than producing a single one-shot response.

Linux & Systems

OOM Killer #: A Linux kernel mechanism that forcibly kills a process when the system is critically low on memory, choosing a victim by a scoring heuristic rather than at random. Read more →
Load Average #: A rolling measure of how many processes are running or waiting for CPU time over the last 1, 5, and 15 minutes — high relative to core count signals contention, not necessarily CPU usage itself. Read more →
systemd#: The init system and service manager used by most modern Linux distributions — it starts services at boot, restarts them on failure, and manages dependencies between them.
Cgroups#: A Linux kernel feature that limits and accounts for the CPU, memory, and I/O a group of processes can use — one of the two primitives (with namespaces) that containers are built on.
Linux Namespaces#: A kernel feature that isolates what a group of processes can see — its own process tree, network stack, or filesystem mounts — the other primitive containers rely on, alongside cgroups.
Symlink#: A symbolic link: a file that's really just a pointer to another file or directory's path, resolved transparently by the filesystem when accessed.
WebAssembly (Wasm)#: A portable, low-level binary format that runs at near-native speed in browsers and increasingly on servers and at the edge, letting languages like Rust and Go run outside their usual runtime. Read more →

Browse All Articles Troubleshooting Cheat Sheet

DevOps & CI/CD

CI/CD #

Continuous Integration and Continuous Delivery/Deployment — the practice of automatically building, testing, and shipping code on every change instead of in large manual batches. Read more →

GitOps#

An operating model where a Git repository is the single source of truth for infrastructure and deployments — an automated controller continuously reconciles the live system to match what's committed, and every change ships as a pull request.

Infrastructure as Code (IaC)#

Defining servers, networks, and other infrastructure in version-controlled configuration files instead of clicking through a console, so changes are reviewable, repeatable, and auditable. Read more →

Immutable Infrastructure#

A pattern where servers or containers are never modified in place after deployment — any change ships as a brand-new image or instance that replaces the old one, eliminating configuration drift.

Configuration Drift#

The gradual divergence between a system's actual state and the configuration it was supposed to have, usually caused by manual, undocumented changes made outside the normal deployment process.

Blue-Green Deployment#

A release strategy that runs two identical production environments — one live (blue), one idle (green) — and switches traffic to the new version instantly, with an immediate rollback path if something's wrong.

Canary Release#

Rolling out a new version to a small slice of real traffic first, watching error rates and latency, and only widening the rollout once it proves healthy.

Idempotency#

A property of an operation that produces the same result no matter how many times it's applied — critical for safe retries in APIs, deployments, and distributed systems where a request might be sent more than once.

Observability #

The ability to understand a system's internal state from the outside, through logs, metrics, and traces — going beyond monitoring's known failure modes to answer questions you didn't anticipate asking. Read more →

SRE (Site Reliability Engineering)#

An engineering discipline that applies software practices to operations, using error budgets and SLOs to balance reliability against the pace of shipping new features. Read more →

Chaos Engineering#

Deliberately injecting failures into a production or production-like system to verify it actually survives the outages it's supposed to be resilient to, before a real one finds the gap for you.

Containers & Kubernetes

Container #

An isolated process that packages an application with its dependencies, sharing the host machine's kernel rather than virtualizing an entire OS — what makes containers fast to start and light on resources compared to VMs. Read more →

Kubernetes #

An open-source system for running containers across a cluster of machines — it schedules workloads, restarts failed containers, scales replicas, and rolls out updates without downtime. Read more →

Pod#

The smallest deployable unit in Kubernetes: one or more containers that share a network namespace and storage, always scheduled together on the same node.

Kubernetes Service#

A stable network endpoint that load-balances traffic across a changing set of Pods, so other workloads never need to know individual Pod IP addresses.

Ingress#

A Kubernetes resource that manages external HTTP(S) access to services inside a cluster, typically handling routing, TLS termination, and virtual hosting through a single load balancer.

Helm Chart#

A packaged, templated bundle of Kubernetes manifests — the standard way to install and version complex applications on a cluster with a single command instead of hand-applying dozens of YAML files.

Sidecar Pattern#

Running a helper container alongside a main application container in the same Pod to add a cross-cutting concern — logging, a proxy, TLS — without modifying the main application's code.

Service Mesh#

An infrastructure layer, usually built from sidecar proxies, that handles service-to-service traffic — retries, mutual TLS, load balancing, and observability — without any of that logic living in application code.

Horizontal Pod Autoscaler (HPA)#

A Kubernetes controller that automatically adds or removes Pod replicas based on observed CPU, memory, or custom metrics, so capacity tracks real demand instead of a fixed, guessed number.

Cloud & Infrastructure

CDN (Content Delivery Network)#

A distributed network of servers that caches and serves content from a location physically close to each visitor, cutting latency compared to serving every request from one origin server. Read more →

Edge Computing#

Running code close to the user — on CDN points of presence or regional nodes — instead of a single centralized server, trading some flexibility for much lower latency.

Serverless #

A model where you deploy functions or services without managing servers — the platform handles provisioning and scaling, and you're billed for actual usage rather than idle capacity. Read more →

Cold Start#

The extra latency a serverless function or scale-to-zero database incurs on its first request after being idle, while the platform provisions a fresh execution environment.

Multi-Cloud#

Running workloads across more than one cloud provider — deliberately, for redundancy or best-of-breed services, as opposed to being locked into a single vendor's ecosystem.

Vendor Lock-In#

The cost and difficulty of switching away from a cloud provider or tool once you've built deeply against its proprietary APIs, formats, or managed services.

Load Balancer #

A component that distributes incoming traffic across multiple backend servers, so no single instance is overwhelmed and traffic keeps flowing if one instance fails. Read more →

Reverse Proxy #

A server that sits in front of one or more backend services, forwarding client requests to them — used for load balancing, TLS termination, caching, and hiding backend topology. Read more →

VPC (Virtual Private Cloud)#

An isolated, private slice of a public cloud provider's network where you control IP ranges, subnets, and routing — the boundary that keeps your resources network-isolated from other tenants.

IAM (Identity and Access Management)#

The system of users, roles, and policies that controls who — human or machine — can do what to which resources in a cloud account, ideally scoped to least privilege. Read more →

Zero Trust #

A security model that assumes no request is trustworthy by default, network location included, and instead verifies every request's identity and authorization explicitly, every time. Read more →

APIs & Architecture

REST #

An architectural style for web APIs built around resources, standard HTTP methods, and statelessness — the default choice for most public and web-facing APIs. Read more →

GraphQL #

A query language for APIs that lets clients request exactly the fields they need in one round trip, instead of over- or under-fetching from fixed REST endpoints. Read more →

gRPC #

A high-performance RPC framework using Protocol Buffers over HTTP/2, favored for fast, strictly-typed internal service-to-service communication over REST or GraphQL. Read more →

Webhook#

A callback: instead of a client repeatedly polling for updates, a server sends an HTTP request to a client-provided URL the moment an event happens.

API Gateway#

A single entry point that sits in front of one or more backend services, handling routing, authentication, and rate limiting so individual services don't each reimplement them.

Rate Limiting #

Capping how many requests a client can make in a given time window, to protect a service from being overwhelmed and to keep usage fair across clients. Read more →

Message Queue #

A component that decouples producers from consumers by holding messages until a consumer is ready to process them — smoothing traffic spikes and letting services fail independently. Read more →

Event-Driven Architecture#

A design where services communicate by publishing and reacting to events rather than calling each other directly, reducing tight coupling at the cost of harder-to-trace flows.

Circuit Breaker#

A pattern that stops calling a failing downstream dependency after enough errors, failing fast instead of piling up timeouts, and periodically retries to see if it has recovered.

AI & Machine Learning

LLM (Large Language Model)#

A neural network trained on huge amounts of text to predict the next token, which turns out to be a general-purpose engine for generating, summarizing, and reasoning about language. Read more →

Embeddings #

Numeric vectors that represent the meaning of text, images, or other data, positioned so that similar things end up close together in vector space — the foundation of semantic search. Read more →

RAG (Retrieval-Augmented Generation)#

Feeding an LLM relevant documents retrieved from a knowledge base at query time, so it answers grounded in your actual data instead of only what it memorized during training. Read more →

Fine-Tuning#

Further training a pre-trained model on a smaller, task-specific dataset to specialize its behavior, as opposed to prompting a general-purpose model at inference time.

Prompt Engineering#

Deliberately structuring the instructions and context given to an LLM to reliably get the output you want, without changing the model's weights.

Hallucination#

When an LLM generates output that's fluent and confident but factually wrong or unsupported by its input — a known failure mode, not a bug in the traditional sense.

Vector Database #

A database purpose-built to store embeddings and run fast nearest-neighbor similarity search over them — the retrieval half of most RAG systems. Read more →

Context Window#

The maximum amount of text, measured in tokens, an LLM can process in a single request — everything beyond it gets truncated or must be summarized/retrieved instead.

Model Context Protocol (MCP)#

An open protocol that standardizes how AI applications connect to external tools and data sources, so a model can call a shared set of tools instead of every app inventing its own integration. Read more →

Vibe Coding #

Building software primarily by describing intent to an AI coding assistant and iterating on what it produces, rather than writing most of the code by hand. Read more →

MLOps #

The practices and tooling for deploying, monitoring, and retraining machine learning models reliably in production — CI/CD's counterpart for the ML lifecycle. Read more →

AI Agent#

A system that uses an LLM in a loop to plan, call tools, and take multi-step actions toward a goal, rather than producing a single one-shot response.

Linux & Systems

OOM Killer #

A Linux kernel mechanism that forcibly kills a process when the system is critically low on memory, choosing a victim by a scoring heuristic rather than at random. Read more →

Load Average #

A rolling measure of how many processes are running or waiting for CPU time over the last 1, 5, and 15 minutes — high relative to core count signals contention, not necessarily CPU usage itself. Read more →

systemd#

The init system and service manager used by most modern Linux distributions — it starts services at boot, restarts them on failure, and manages dependencies between them.

Cgroups#

A Linux kernel feature that limits and accounts for the CPU, memory, and I/O a group of processes can use — one of the two primitives (with namespaces) that containers are built on.

Linux Namespaces#

A kernel feature that isolates what a group of processes can see — its own process tree, network stack, or filesystem mounts — the other primitive containers rely on, alongside cgroups.

Symlink#

A symbolic link: a file that's really just a pointer to another file or directory's path, resolved transparently by the filesystem when accessed.

WebAssembly (Wasm)#

A portable, low-level binary format that runs at near-native speed in browsers and increasingly on servers and at the edge, letting languages like Rust and Go run outside their usual runtime. Read more →