Opscraft
Glossary

Every term, in plain words

628 terms used across the lessons, each explained without jargon. In lessons, terms with a dotted underline open the same explanation when you tap them.

A

A record · address record
A DNS record that maps a name to an IPv4 address.
AAAA · IPv6 address record
A DNS record that maps a name to an IPv6 address.
ABAC · attribute-based access control
Granting access by matching tags, such as "people tagged team=orders can manage resources tagged team=orders", instead of listing every resource.
Absolute path
A file location written from the top of the tree, starting with /, such as /var/log/app.log. It means the same thing whichever folder you are in.
Access Analyzer
An AWS tool that finds resources shared outside your account or organisation, unused permissions, and problems in policies, and can generate policies from actual activity.
access key
A long-lived ID and secret pair that lets a program call AWS as an IAM user. They leak easily, so roles and short-lived credentials are preferred.
ACK · acknowledgement
A TCP message saying "I received your data up to this point." Every TCP segment after the first carries one.
ACL · access control list
A list of who is allowed to do what with a resource. In AWS, S3 ACLs are a legacy feature and network ACLs filter subnet traffic.
ACM · AWS Certificate Manager
The AWS service that issues free public TLS certificates for use on AWS load balancers and CloudFront, and renews them automatically.
admission controller
A check that runs when an object is submitted to the API server and can reject or modify it before it is saved. Policy tools such as Gatekeeper and Kyverno work this way.
affinity
Rules that attract pods to certain nodes or to each other, or keep them apart (anti-affinity), for example to spread copies across zones.
AKS · Azure Kubernetes Service
Microsoft Azure's managed Kubernetes, the equivalent of EKS.
ALB · Application Load Balancer
AWS's layer-7 load balancer for HTTP and HTTPS. It can route by hostname, path and headers, terminate TLS and integrate with WAF.
alert
An automatic notification when a measurement crosses a threshold, such as "error rate above 5% for 10 minutes". Good alerts are about user impact and come with a runbook.
Alertmanager
The Prometheus component that receives alerts and groups, silences and routes them to Slack, email or paging tools.
alias record
A Route 53 feature that behaves like a CNAME but works at the zone apex and points straight at AWS resources such as load balancers and CloudFront, with no extra lookup charge.
AMI · Amazon Machine Image
A saved disk image used to launch EC2 instances, containing the operating system and any pre-installed software.
anycast
Announcing the same IP address from many locations so each user reaches the nearest one. Root DNS servers and CDNs use it.
API · application programming interface
A defined way for programs to talk to each other, usually by sending HTTP requests to documented URLs and getting JSON back.
API Gateway
AWS's managed service for publishing HTTP, REST and WebSocket APIs in front of Lambda functions and other backends.
API server
The front door of Kubernetes. Every command, controller and node talks to the cluster through it; it checks permissions and stores changes in etcd.
ARC · Actions Runner Controller
A Kubernetes operator that runs GitHub Actions self-hosted runners as pods, creating a fresh one for each job and scaling with demand.
Argo CD
A CNCF GitOps tool for Kubernetes with a rich web UI, showing each application's sync and health status and syncing it from Git.
Argo Rollouts
A Kubernetes controller from the Argo project that adds canary and blue/green deployments with automated metric analysis.
ARN · Amazon Resource Name
The unique ID of an AWS resource, such as arn:aws:s3:::my-bucket or arn:aws:iam::123456789012:role/deploy.
ARP · Address Resolution Protocol
How a machine finds the hardware (MAC) address that belongs to an IP address on the same local network, by shouting "who has 10.0.1.5?"
artifact
A file produced by a build that you keep and deploy, such as a JAR, a ZIP or a container image.
artifact repository
A server that stores your built packages and caches public ones (Maven Central, npm, PyPI), giving builds one controlled, fast source. Nexus and Artifactory are popular products.
Aurora
AWS's cloud-built MySQL- and PostgreSQL-compatible database, with storage copied six ways across three Availability Zones, fast failover and up to 15 read replicas.
authoritative name server
The DNS server that holds the official records for a domain and gives definitive answers, like the registrar's office for a street.
Auto Scaling group
A group of EC2 instances that AWS keeps at a target size, replacing unhealthy ones and adding or removing instances based on load or schedules.
autovacuum
PostgreSQL's background cleanup of dead rows left by updates and deletes; it also prevents transaction ID wraparound.
AWS · Amazon Web Services
Amazon's cloud platform: rentable servers, storage, databases, networking and hundreds of other services, paid for by use.
AWS App Runner
An AWS service that runs a web application or container straight from source code or an image, with no servers, load balancers or clusters to manage.
AWS Backup
A central AWS service that schedules and manages backups of databases, disks and file systems across accounts and regions.
AWS Config
An AWS service that records how your resources are configured over time and checks them against rules, such as "no public S3 buckets".
AWS Load Balancer Controller
A Kubernetes add-on that creates and configures ALBs and NLBs on AWS from Ingress and Service objects.
AWS Organizations · AWS Organizations
The AWS service for managing many accounts centrally: grouping them into organisational units, consolidated billing and organisation-wide policies.
AWS RAM · Resource Access Manager
An AWS service for sharing resources such as subnets or Transit Gateways with other accounts.
AZ · Availability Zone
One or more separate data centres within a region, with their own power and networking. Spreading across two or three AZs keeps an app running if one fails.
Azure
Microsoft's cloud platform, AWS's main competitor in many enterprises.

B

Backstage
An open-source framework (from Spotify, now CNCF) for building developer portals with a software catalogue, templates, docs and plugins.
base image
The starting image a Dockerfile builds on, named in FROM, such as eclipse-temurin:21-jre or alpine. Its packages and vulnerabilities become part of your image.
Bedrock
AWS's managed service for calling foundation models (LLMs and others) through an API, without hosting them yourself.
BGP · Border Gateway Protocol
The protocol internet providers and large networks use to tell each other which IP ranges they can reach. AWS Direct Connect and VPNs use it to exchange routes.
bind mount
Mapping a folder from the host machine directly into a container.
Blameless
Reviewing a failure to improve the system rather than to punish a person. People share details honestly when they are not afraid of blame.
blue/green deployment
Running the new version (green) alongside the old one (blue), testing it, then switching all traffic at once — with an instant switch back if needed.
BOM · bill of materials
In Maven and Gradle, a special file that pins compatible versions for a family of libraries so you don't set each one yourself.
boto3
The official AWS SDK for Python, used to automate AWS from scripts.
Bottlerocket
A minimal, container-focused Linux OS from AWS for nodes, with a read-only root filesystem and whole-image updates.
branch
A movable label pointing at a commit, used to develop changes in parallel without disturbing the main line.
branch protection
Repository rules that stop direct or forced pushes to important branches and require reviews and passing checks before merging.
bucket
A named container for objects in S3, with its own access policy and settings. Bucket names are globally unique.
Budget alert
A cloud setting that emails you when spending passes an amount you choose — the simplest protection against a surprise bill.
BuildKit
Docker's modern build engine. It builds steps in parallel and supports cache mounts and secret mounts that don't end up in the image.
burn rate
How fast the error budget is being used up compared with the steady pace that would use exactly all of it by the end of the window. A burn rate of 10 means ten times too fast.

C

CA · certificate authority
An organisation (or internal system) that signs certificates, vouching that a public key really belongs to a name. Browsers and operating systems ship with a list of CAs they trust.
CAA · Certification Authority Authorization
A DNS record listing which certificate authorities may issue certificates for a domain.
cache invalidation
Removing cached copies before they expire, for example a CloudFront invalidation after a bad deploy.
cache key
What a cache uses to decide whether two requests are the same, such as host, path and selected headers or query strings.
cache stampede
Many requests missing the cache at the same moment (often when a popular key expires) and all hitting the database together.
cache-aside
The common caching pattern: read the cache, and on a miss read the database and store the result with an expiry time.
Cache-Control
The HTTP header that tells browsers and CDNs whether a response may be stored, by whom, for how long, and what to do once it's stale.
canary
Sending a small share of real traffic (say 5%) to a new version first and watching its errors and latency before rolling it out to everyone. Named after the canaries miners used to detect gas.
capabilities (Linux)
Root's powers split into small pieces (such as binding to low ports or changing network settings) that can be granted or removed individually. Containers should drop all they don't need.
cardinality
The number of distinct label combinations a metric has. Labels with endless values (user IDs, full URLs) create millions of series and overload monitoring systems.
CDK · Cloud Development Kit
A way to write AWS infrastructure in programming languages such as TypeScript or Python; it generates CloudFormation templates.
CDN · content delivery network
A worldwide network of cache servers that keeps copies of content close to users, so pages load faster. Amazon CloudFront is AWS's CDN.
certificate
A digital ID card for a server (or client). It contains a name such as shop.example.com and a public key, and is signed by a certificate authority that browsers trust.
cgroups · control groups
A kernel feature that limits and measures how much CPU, memory and I/O a group of processes can use. Container resource limits are cgroups underneath.
change set
A CloudFormation preview listing exactly what an update would add, modify or replace, before you apply it.
chaos engineering
Deliberately injecting failures (killing pods, adding latency, failing a zone) in a controlled way to find weaknesses before they cause real outages.
chart
A Helm package: a folder of templated Kubernetes manifests plus default settings, versioned and shareable.
Checkov
An open-source scanner that checks Terraform, CloudFormation, Kubernetes and other IaC files for security misconfigurations.
CI · continuous integration
Automatically building and testing every code change as soon as it is pushed, so problems are caught within minutes.
CI/CD · continuous integration and delivery
The whole automated path from a code push to a running release: build, test, package, scan, deploy.
CIDR · Classless Inter-Domain Routing
The address/number way of writing a range of IP addresses. The number after the slash says how many leading bits are fixed: 10.0.0.0/16 covers 65,536 addresses, 10.0.1.0/24 covers 256.
circuit breaker
A pattern that stops calling a dependency that keeps failing and returns an error quickly instead, then tries again later — like a fuse that trips to protect the rest of the house.
CIS Benchmark · Center for Internet Security Benchmark
Published, consensus security configuration checklists for systems such as Linux, Kubernetes, EKS and AWS accounts.
CLI · command-line interface
A tool you control by typing commands, such as aws, kubectl or git.
CloudFormation
AWS's infrastructure-as-code service: you describe resources in a JSON or YAML template and CloudFormation creates, updates and deletes them as one stack.
CloudFormation stack
A group of resources created from one CloudFormation template and managed together.
CloudFront
AWS's content delivery network (CDN), which caches content in locations worldwide close to users.
CloudTrail
AWS's audit log: a record of every API call made in your account — who did what, when and from where.
CloudWatch
AWS's built-in monitoring service: it collects metrics and logs, draws dashboards and raises alarms.
cluster
A group of machines managed as one system. A Kubernetes cluster has a control plane that makes decisions and worker nodes that run the applications.
Cluster Autoscaler
A Kubernetes add-on that adds nodes when pods can't be scheduled and removes nodes that are underused, by resizing node groups.
ClusterIP
The default Service type: a virtual IP reachable only inside the cluster.
CNAME · canonical name
A DNS record that says "this name is an alias of that other name; look that one up instead." It cannot be used at the top of a zone (the apex).
CNCF · Cloud Native Computing Foundation
The open-source foundation that hosts Kubernetes, Prometheus, Flux, Argo, Helm and many other cloud-native projects.
CNI · Container Network Interface
The standard plug-in interface for pod networking. On EKS the default CNI (the VPC CNI) gives every pod a real VPC IP address.
CodeArtifact
AWS's managed repository for software packages (Maven, npm, PyPI and others), with upstream caching of public registries.
CODEOWNERS
A file in a repository that names who must review changes to which paths, such as "the platform team must approve anything under .github/workflows/".
cold start
The extra delay when a serverless platform must create and initialise a new environment before running your code.
Commit
A saved snapshot of changes in Git, with an author, a time and a message explaining why. Commits form the project's history.
compliance
Meeting the rules set by regulators, customers or standards (such as ISO 27001, SOC 2 or PCI DSS) and being able to prove it.
composite action
A GitHub Action made by bundling several steps together, reused as a single step in other workflows.
Compute Optimizer
An AWS service that recommends better-fitting sizes for EC2, Auto Scaling groups, EBS, Lambda, Fargate and RDS based on actual usage.
ConfigMap
A Kubernetes object holding non-secret configuration (key-value pairs or files) that pods can read as environment variables or files.
confused deputy
A security problem where a trusted service is tricked into using its access on behalf of the wrong customer. AWS prevents it with External IDs and source conditions.
container
A packaged application running as an isolated process with its own filesystem, network and limits, but sharing the host's kernel. Lighter and faster to start than a virtual machine.
container runtime
The software that actually starts and stops containers on a machine. Kubernetes talks to runtimes such as containerd through the Container Runtime Interface (CRI).
continuous delivery · continuous delivery / deployment
Automatically taking each tested change towards production. Continuous delivery keeps every change ready to release (with a manual approval); continuous deployment releases it automatically.
control plane
The "brain" of Kubernetes: the API server, etcd, scheduler and controller manager, which store the desired state and work to make it real. On EKS, AWS runs it for you.
Control Tower
An AWS service that sets up a well-structured multi-account environment (a "landing zone") with logging, guardrails and an account factory.
CoreDNS
The DNS server inside a Kubernetes cluster that lets pods find Services by name, such as orders.payments.svc.cluster.local.
cosign
Tools from the Sigstore project for signing and verifying container images. Keyless signing uses the CI system's identity instead of a stored private key.
cost allocation tag
A resource tag (such as team or service) activated in AWS Billing so costs can be grouped and reported by it.
Cost Explorer
The AWS Billing tool for exploring spend by service, account, tag and usage type, with forecasts and recommendations.
counter (metric)
A metric that only goes up (such as total requests) and resets to zero when the app restarts. You look at its rate of increase.
CPS · continuation-passing style
The way Jenkins runs pipeline Groovy so a pipeline can pause and survive a Jenkins restart. It's why some normal Groovy code fails in pipelines and needs @NonCPS.
CrashLoopBackOff
A pod status meaning the container keeps crashing and Kubernetes is waiting longer and longer between restart attempts.
CRD · CustomResourceDefinition
A way to teach Kubernetes a new object type (such as HelmRelease or Certificate), which a matching controller or operator then acts on.
cron
The Linux scheduler that runs commands at set times, written as five fields: minute, hour, day of month, month, day of week.
CronJob
A Kubernetes object that creates Jobs on a schedule, like cron for the cluster.
Crossplane
A Kubernetes-based tool that creates and manages cloud resources from custom Kubernetes objects, letting platform teams offer simple self-service APIs.
CSI · Container Storage Interface
The standard plug-in interface that lets Kubernetes use storage systems such as EBS or EFS through a driver.
CVE · Common Vulnerabilities and Exposures
A public ID for a known vulnerability, such as CVE-2021-44228 (Log4Shell). Scanners report CVEs found in your images and dependencies.
CVSS · Common Vulnerability Scoring System
A 0–10 severity score for vulnerabilities. High scores mean potentially severe impact, not necessarily that anyone is exploiting it.

D

DaemonSet
A Kubernetes object that runs one copy of a pod on every node (or chosen nodes), used for agents such as log collectors and monitoring.
dashboard
A page of graphs and numbers showing the health of a system at a glance.
DAST · dynamic application security testing
Testing a running application from the outside by sending attack-like requests.
data drift
When production input data changes compared with the data a model was trained on, often making its predictions worse without any error.
Database
A program that stores data in an organised way and answers questions about it, such as "which orders were placed today?". PostgreSQL, MySQL and DynamoDB are examples.
DDoS · distributed denial of service
An attack that floods a service with traffic from many machines so real users can't get through.
declarative
Describing the end result you want ("three web servers") and letting a tool work out the steps, rather than listing the steps yourself (imperative).
default gateway
The router a machine sends traffic to when the destination is not on its own network, like the exit door of a building.
dependency
A library or package your code relies on, such as a JSON parser. Transitive dependencies are the libraries your libraries need.
Deployment
A Kubernetes object that keeps a set number of identical pods running and replaces them gradually (a rolling update) when you change the image or settings.
deployment strategy
How a new version replaces the old one: recreate, rolling update, blue/green or canary, each trading speed against risk.
desired state
What you declared you want (for example "3 copies of version 1.4"). Kubernetes and GitOps tools keep working until reality matches it.
DestinationRule
An Istio resource that sets what happens after routing to a service: subsets by version, load balancing, connection limits and outlier detection.
DevOps
A way of working where the people who build software and the people who run it share responsibility, automating the path from code change to production so releases are frequent, safe and repeatable.
DevSecOps
Building security into every step of the delivery pipeline and platform, with automated checks, instead of a separate review at the end.
DHCP · Dynamic Host Configuration Protocol
The service that automatically hands a device an IP address, gateway and DNS server when it joins a network.
digest
The SHA-256 fingerprint of an image's content, like sha256:4f3a.... It never changes, so pulling by digest always gives exactly the same image.
Direct Connect · AWS Direct Connect
A dedicated private network line between your data centre or office and AWS, with more consistent speed than the internet.
disaster recovery · disaster recovery
The plan and setup for restoring service after a major failure such as losing a whole region or a database.
disk snapshot
A point-in-time copy of a disk or database, used for backups and for creating new copies.
distroless
A minimal base image that contains only the application runtime — no shell or package manager — which means fewer vulnerabilities and less for attackers to use.
DKIM · DomainKeys Identified Mail
A digital signature added to outgoing email, with the public key published in DNS, so receivers can check the message wasn't forged or altered.
DLQ · dead-letter queue
A queue where messages or events go after failing too many times, so they can be inspected and retried instead of blocking or disappearing.
DMARC · Domain-based Message Authentication, Reporting and Conformance
A DNS policy telling email receivers what to do when SPF or DKIM checks fail (allow, quarantine or reject) and where to send reports.
DNAT · destination network address translation
NAT that rewrites the destination address, for example to forward traffic arriving at a public IP to a private server.
DNS · Domain Name System
The internet's phone book: it turns names like shop.example.com into IP addresses. It is a distributed system of servers that each know part of the answer.
DNS delegation
Handing responsibility for part of the DNS tree to other name servers by adding NS records, like a head office telling callers "for that department, ring this number."
DNS record
One entry in a DNS zone, such as "shop.example.com has IP 203.0.113.10". Each record has a name, type, TTL and value.
DNSSEC · DNS Security Extensions
Digital signatures on DNS records so resolvers can check that answers weren't forged or altered.
Docker
The tool that made containers popular: it builds images from a Dockerfile and runs them as containers on your machine.
Docker Compose
A tool that starts several related containers (for example an app, a database and a cache) from one YAML file with one command. Aimed at a single machine.
Docker Swarm
Docker's built-in tool for running containers across several machines as one cluster. Simpler than Kubernetes but with far fewer features.
Dockerfile
A text file of instructions (FROM, COPY, RUN, CMD...) describing how to build a container image, like a recipe.
domain name
A human-friendly name such as example.com. Names are read right to left in a tree: com is the top, example.com sits under it, shop.example.com under that.
DORA metrics · DevOps Research and Assessment
Four measures of delivery performance: how often you deploy, how long a change takes to reach production, how often changes fail, and how fast you recover.
drift
When real infrastructure no longer matches its code because someone changed it by hand. IaC and GitOps tools detect and correct it.
DynamoDB
AWS's serverless key-value and document database, with consistent single-digit-millisecond responses at any scale.

E

EBS · Elastic Block Store
Network-attached disks for EC2 instances (and Kubernetes pods). A volume lives in one Availability Zone and can be snapshotted to S3.
EC2 · Elastic Compute Cloud
AWS's virtual servers. You choose an instance type (CPU and memory size), an AMI (disk image) and a network, and pay while it runs.
EC2 Image Builder
AWS's managed service for building, testing and distributing machine images on a schedule, an alternative to running Packer.
ECR · Elastic Container Registry
AWS's private container image registry, with scanning, lifecycle rules and IAM-based access.
ECS · Elastic Container Service
AWS's own container orchestrator, simpler than Kubernetes and tightly integrated with AWS.
EFS · Elastic File System
AWS's managed shared network file system (NFS) that many instances or pods in different Availability Zones can mount at once.
egress
Traffic leaving a network or system. Cloud providers charge for data egress to the internet and between zones or regions.
EKS · Elastic Kubernetes Service
AWS's managed Kubernetes. AWS runs and patches the control plane; you run your workloads on nodes you choose (EC2, managed node groups, Karpenter or Fargate).
EKS Pod Identity
The newer way to give EKS pods IAM roles: you link a service account to a role through the EKS API, and an agent on each node hands out credentials. No per-cluster OIDC setup needed.
Elastic IP
A fixed public IPv4 address you own in your AWS account and can attach to an instance, NAT gateway or Network Load Balancer.
ElastiCache
AWS's managed in-memory cache service, running Valkey, Redis OSS or Memcached.
ELB · Elastic Load Balancing
AWS's family of managed load balancers: Application (ALB), Network (NLB) and Gateway (GWLB).
embedding
A list of numbers representing the meaning of a piece of text (or image), so similar meanings have nearby vectors.
encryption at rest
Encrypting data where it is stored (disks, databases, backups), so stolen storage is unreadable.
encryption in transit
Encrypting data as it travels over a network, usually with TLS.
ENI · Elastic Network Interface
A virtual network card in a VPC, with its own private IP addresses and security groups. EC2 instances (and EKS pods, through the VPC CNI) get their IPs from ENIs.
Entra ID
Microsoft's identity service (formerly Azure Active Directory), used for sign-in to Azure and many company apps.
envelope encryption
Encrypting data with a data key, then encrypting that data key with a master key (for example in KMS). Fast for large data, and the master key never leaves its vault.
Environment
One complete, separate copy of a system — such as dev, staging or production — with its own servers, data and settings.
environment (GitHub)
A named deployment target in GitHub, such as production, that can require approval and hold its own secrets.
Envoy
A high-performance proxy configured dynamically through APIs, used by service meshes such as Istio and many Gateway API implementations.
EPSS · Exploit Prediction Scoring System
A score estimating how likely a vulnerability is to be exploited in the next 30 days, used to prioritise fixes.
error budget
The amount of failure an SLO allows. A 99.9% target leaves a 0.1% budget — about 43 minutes of downtime in 30 days — that can be "spent" on risky releases or incidents.
ETag · entity tag
A version identifier for a response. Caches send it back to ask "has this changed?" and get a small 304 Not Modified if not.
etcd
The key-value database where Kubernetes keeps the whole cluster's state. If etcd is lost without a backup, the cluster forgets everything it was running.
event source mapping
The Lambda feature that polls a queue or stream (SQS, Kinesis, DynamoDB Streams, Kafka) and invokes a function with batches of records.
EventBridge
A managed event bus that routes events (from AWS services, your apps or SaaS tools) to targets using matching rules, and can also run things on a schedule.
execution environment
The isolated micro-VM where a Lambda function runs; it handles one request at a time and is reused for later requests while warm.
executor
A slot on a Jenkins agent that runs one build at a time. An agent with 4 executors runs up to 4 builds in parallel.
exit code
A number a program returns when it ends: 0 means success, anything else means failure. Pipelines and scripts use it to decide what to do next. 137 means the process was killed with SIGKILL (often out of memory).
expand and contract
Changing a schema in backward-compatible steps (add, backfill, switch, remove) so old and new app versions both work during a rollout.
exponential backoff
Waiting longer after each failed attempt — for example 1, 2, 4, 8 seconds — so retries don't overwhelm a struggling service.
exporter
A small program that exposes metrics in Prometheus format for something that doesn't do so itself, such as a Linux host (node_exporter) or a database.
External ID
A secret value a third-party vendor must supply when assuming a role in your account, to prevent the confused-deputy problem.
External Secrets Operator
A Kubernetes operator that copies secrets from an external store such as AWS Secrets Manager into Kubernetes Secrets, so only references live in Git.

F

Falco
A runtime security tool that watches system calls on nodes and alerts on suspicious behaviour, such as a shell starting inside a container.
Fargate
A serverless way to run containers on ECS or EKS: AWS provides the compute for each task or pod, so you don't manage servers.
Fault Injection Service
An AWS service for controlled chaos experiments, such as stopping instances or simulating an Availability Zone failure, with automatic stop conditions.
feature flag
A switch in the code that turns a feature on or off without redeploying, so code can be released hidden and enabled gradually or switched off instantly.
feature store
A system that computes and serves model input features the same way for training and production, preventing training/serving skew.
federation
Trusting an outside identity provider (such as Okta, Google or GitHub) to vouch for users, instead of creating separate accounts in every system.
file descriptor
A number a process uses to refer to an open file, socket or pipe. Each process has a limit; "too many open files" means it hit that limit.
file permissions
Linux rules saying who can read (r), write (w) and execute (x) a file: the owner, the group, and everyone else. Written as rwxr-x--- or the number 750.
FIN · finish
A TCP message meaning "I have finished sending." Each side sends one to close a connection politely.
FinOps
The practice of making engineering, finance and business share responsibility for cloud spending: see it, optimise it, and keep it under control.
firewall
A filter that allows or blocks traffic based on rules such as source, destination and port.
Flagger
A progressive-delivery tool for Kubernetes that automates canary and blue/green releases by shifting traffic step by step and checking metrics.
Flow Logs · VPC Flow Logs
Records of the network connections to and from interfaces in a VPC (who, where, which port, accepted or rejected), used for troubleshooting and security.
Fluent Bit
A lightweight log collector, usually run on every Kubernetes node, that reads container logs, adds details such as the pod name and ships them to a log store.
Flux
A CNCF GitOps tool made of Kubernetes controllers that pull manifests and Helm charts from Git or OCI registries and keep the cluster in sync.
Flyway
Tools that apply versioned database migration files in order and record which have run, so schema changes go through the pipeline.
FOCUS · FinOps Open Cost and Usage Specification
An open, common format for cloud and SaaS billing data, so costs from different providers can be compared in one place.
force push
Overwriting a branch on the server with your local version, discarding commits that only exist on the server. --force-with-lease refuses if someone else pushed in the meantime.
FQDN · fully qualified domain name
A complete domain name that spells out every level up to the root, such as api.shop.example.com. (the final dot stands for the root).
Free tier
The free allowance or starter credits a cloud provider gives new accounts. Limits apply, so set a budget alert and delete resources when you finish practising.

G

game day
A planned exercise where a team simulates a failure and practises responding to it.
Gatekeeper
The Kubernetes admission controller that runs OPA policies, for example blocking pods without resource limits.
Gateway API
The Kubernetes successor to Ingress: GatewayClass, Gateway and HTTPRoute objects that split ownership between platform and app teams and support weights, header matching and more in the standard.
gauge (metric)
A metric that can go up and down, such as current memory use or queue length.
Git
The version-control system that records every change to a project as a commit, so you can see history, work on branches in parallel and undo mistakes.
Git commit
A saved snapshot of the project in Git, with an author, a message and a unique ID (its SHA hash).
Git merge
Combining the changes from one branch into another. Git creates a merge commit when both sides have new work.
GitFlow
An older branching model with long-lived develop and release branches. Suits scheduled releases, but is heavy for continuous delivery.
GitHub Action
A reusable building block for GitHub Actions workflows, such as actions/checkout, referenced with uses:.
GitHub Actions
GitHub's built-in CI/CD. Workflows are YAML files in .github/workflows/ that run jobs on runners when events such as a push or pull request happen.
GitHub Actions workflow
A GitHub Actions automation defined in one YAML file: what triggers it and which jobs to run.
GITHUB_TOKEN
A short-lived token GitHub creates for each workflow run so it can call the GitHub API for that repository, with permissions you can restrict.
GitOps
Running deployments from Git: the desired state of each environment lives in a Git repository, and an agent in the cluster continuously pulls and applies it. A change is a commit; a rollback is a revert.
golden image
A pre-built, tested machine image with the OS, patches, agents and hardening every server needs, used to launch identical servers.
golden path
The supported, opinionated and easiest way to do a common task, such as creating a new service with everything already wired up.
golden signals
The four things Google's SRE book says to watch for any service: latency, traffic, errors and saturation.
gp3
General-purpose SSD EBS volume types. gp3 is cheaper than gp2 and lets you set speed (IOPS and throughput) separately from size.
GPU · graphics processing unit
A processor with thousands of small cores, used for the heavy maths in training and running machine-learning models; much more expensive than CPU capacity.
graceful shutdown
Graceful shutdown: finishing in-flight work before exiting. Graceful degradation: keeping core features working (perhaps with reduced quality) when a part of the system fails.
Gradle
A Java and Kotlin build tool configured with code (build.gradle.kts), known for speed thanks to incremental builds and caching.
Grafana
A popular dashboard tool that draws graphs from Prometheus, Loki, CloudWatch and many other data sources.
Graviton
AWS's own ARM-based processors, which usually give better price-performance than Intel or AMD instances if your software supports ARM (arm64).
Groovy
A programming language that runs on the Java platform. Jenkins pipelines and shared libraries are written in it.
GuardDuty
AWS's threat detection service. It watches logs for signs of attack, such as credentials used from an unusual country or crypto-mining.
GWLB · Gateway Load Balancer
An AWS load balancer that sends traffic through third-party security appliances such as firewalls.

H

HAProxy
A fast, mature load balancer and proxy for TCP and HTTP, known for active health checks and detailed statistics.
HCL · HashiCorp Configuration Language
The readable configuration language used by Terraform files.
health check
A regular test request (for example GET /healthz) that a load balancer, DNS service or orchestrator sends to decide whether a target is working.
Helm
A package manager for Kubernetes. A Helm chart is a reusable package of templated manifests; installing it with your settings (values) creates a release you can upgrade or roll back.
HelmRelease
Flux's custom resource describing a Helm chart to install, its version and its values; Flux's helm-controller keeps it applied.
high availability · high availability
Designing a system to keep working when parts fail, usually by running several copies in different places.
histogram
A metric that counts observations into buckets (for example request durations under 100 ms, 250 ms, 500 ms), so percentiles such as p99 can be calculated.
hit ratio
The share of requests a cache answers itself. A drop from 95% to 80% means four times as many requests reach the origin.
Home directory
Your personal folder on a Linux machine, usually /home/<your-name> (/root for the root user). ~ is a shortcut for it.
hosted zone
Route 53's container for a domain's DNS records. Public hosted zones answer the internet; private hosted zones answer only inside chosen VPCs.
HPA · Horizontal Pod Autoscaler
Kubernetes' built-in autoscaler that adds or removes pod copies based on metrics such as CPU usage or requests per second.
HSTS · HTTP Strict Transport Security
A response header telling browsers to use only HTTPS for a site for a set time, so they never connect over plain HTTP.
HTTP · Hypertext Transfer Protocol
The request-and-response language browsers and APIs use: a client asks with a method such as GET or POST, and the server answers with a status code such as 200 or 404 plus content.
HTTPS · HTTP Secure
HTTP sent inside an encrypted TLS connection, so nobody in between can read or change it. The padlock in the browser means HTTPS.
Hypervisor
Software that splits one physical computer into several virtual machines, giving each its own share of CPU, memory and disk. AWS uses its own hypervisor (Nitro) to run EC2 instances.

I

IaaS · Infrastructure as a Service
Renting the basic building blocks — virtual machines, disks, networks — while you manage the operating system and everything above it. Amazon EC2 is IaaS.
IAM · Identity and Access Management
AWS's system for deciding who (users, roles, services) can do what (actions) on which resources, using JSON policies.
IAM Identity Center
AWS's single sign-on service: people log in with the company identity provider and get time-limited access to the AWS accounts and roles assigned to their group.
IAM policy
A JSON document listing allowed or denied actions on resources, optionally with conditions. Attached to identities (what they can do) or to resources (who can use them).
IAM role
An AWS identity with permissions but no password or permanent keys. Trusted people or services "assume" it and get temporary credentials, like borrowing a visitor badge.
IAM user
A permanent AWS identity with a password and/or access keys. Modern setups avoid them in favour of single sign-on for people and roles for software.
ICMP · Internet Control Message Protocol
The network's messaging system for errors and diagnostics. ping and traceroute use it, and "packet too big" warnings travel over it.
idempotent
Safe to run more than once with the same result. Running an idempotent deploy or script twice doesn't create duplicates or break anything.
image
A read-only package containing an application and everything it needs to run (libraries, runtime, files), built from a Dockerfile and stored in a registry. A container is a running copy of an image, like a house built from a blueprint.
image automation
A GitOps feature that watches a registry for new image tags and commits the updated tag to Git automatically.
image layer
One slice of a container image, created by one Dockerfile instruction. Layers are cached and shared between images, so unchanged steps don't need rebuilding or re-downloading.
image tag
A human-readable label on an image version, such as orders:1.4.2 or orders:a1b2c3d. Tags can be moved to point at a different image, unlike digests.
ImagePullBackOff
A pod status meaning the node can't download the container image — usually a wrong name or tag, a missing permission or no network route to the registry.
IMDS · Instance Metadata Service
A local address (169.254.169.254) where an EC2 instance can read facts about itself and fetch its role's temporary credentials. Version 2 (IMDSv2) requires a session token, which blocks a common attack.
immutable
Never changed after creation. Instead of patching a running server or image, you build a new one and replace the old.
imperative (style)
Giving step-by-step commands ("create a server, then another"), as opposed to describing the desired end result (declarative).
incident
An unplanned event that hurts users or the business, such as an outage or a security breach, handled through a defined response process.
incident commander
The person who coordinates an incident: assigning roles, making decisions and keeping communication flowing, rather than debugging hands-on.
inference
Using a trained model to make predictions or generate output for new inputs, as opposed to training it.
Infracost
A tool that estimates the monthly cost change of Terraform changes and comments it on pull requests.
infrastructure as code · infrastructure as code
Describing servers, networks and other infrastructure in files that tools apply automatically, so environments are repeatable, reviewable and version-controlled.
Ingress
A Kubernetes object holding HTTP routing rules (hostnames, paths, TLS) for traffic coming into the cluster. An ingress controller, such as ingress-nginx or the AWS Load Balancer Controller, puts the rules into effect.
ingress (traffic)
Traffic coming into a network or system.
inode
The record a Linux filesystem keeps for each file: its owner, permissions, size and where its data lives (but not its name). A disk can run out of inodes even with free space left.
instance refresh
An Auto Scaling group operation that replaces instances in batches with ones from the current launch template, keeping a minimum healthy percentage.
instance type
The size and family of an EC2 server, such as m7g.large (general purpose, Graviton, 2 vCPU, 8 GiB). Letters signal the family: m general, c compute, r memory, t burstable, g Graviton.
internal developer platform
The tools, templates, self-service infrastructure and defaults a platform team provides so product teams can build and run software themselves.
internet gateway
The VPC component that connects public subnets to the internet in both directions.
IOPS · input/output operations per second
How many separate read or write operations a disk can handle each second.
IP · Internet Protocol
The basic rule set that moves packets of data from one address to another across networks. It only delivers; reliability is left to protocols like TCP.
IP address · Internet Protocol address
The number that identifies a device on a network, like a postal address for a house. IPv4 addresses look like 10.0.1.25; IPv6 addresses are longer, like 2001:db8::1.
iptables
The Linux kernel's built-in packet filtering and NAT system. Docker and kube-proxy write iptables rules to route container traffic.
IPv4 · Internet Protocol version 4
The original addressing scheme: 32-bit addresses written as four numbers from 0 to 255, such as 192.168.1.10. There are only about 4.3 billion of them, which is why private ranges and NAT exist.
IPv6 · Internet Protocol version 6
The newer addressing scheme with 128-bit addresses, written in hexadecimal groups like 2001:db8::1. It has so many addresses that NAT is rarely needed.
IRSA · IAM Roles for Service Accounts
The EKS feature that lets a pod's Kubernetes service account assume a specific IAM role through OIDC, so each app gets only its own AWS permissions instead of the node's.
ISP · internet service provider
The company that connects you to the internet, such as Airtel or Jio.

J

JCasC · Jenkins Configuration as Code
A Jenkins plugin that configures the whole server from YAML files instead of clicking through the UI.
JDK · Java Development Kit
The tools to compile and run Java. The JRE (runtime) is the smaller part needed only to run it.
Jenkins
A long-established, self-hosted open-source automation server for CI/CD, extended with plugins and driven by Jenkinsfiles.
Jenkins agent
A machine or container that runs build jobs on behalf of a CI server such as Jenkins.
Jenkins controller
The central Jenkins server that stores configuration, schedules jobs and shows the UI. It hands the actual build work to agents. (Formerly called "master".)
Jenkinsfile
The file, stored in a repository, that defines a Jenkins pipeline in Groovy-based syntax.
jitter
A small random delay added to retries or schedules so many clients don't all retry at exactly the same moment.
Job
A Kubernetes object that runs pods until a task finishes successfully, such as a database migration.
job (CI)
A unit of work in a CI system. In GitHub Actions, a group of steps that run together on one runner (jobs run in parallel unless one depends on another); in Jenkins, a configured build project.
Job DSL
A Jenkins plugin for defining Jenkins jobs in Groovy code instead of creating them by clicking through the web interface.
jq
A command-line tool for filtering and reshaping JSON, often used with aws and kubectl output.
JSON · JavaScript Object Notation
A widely used text format for structured data, built from { "key": "value" } objects and [ ] lists. IAM policies and most APIs use it.
JVM · Java Virtual Machine
The program that runs Java (and Kotlin, Groovy, Scala) code. Its memory settings matter in containers.
JWT · JSON Web Token
A signed token containing claims (such as who the user is and when it expires) that a service can verify without calling the issuer.

K

Karpenter
An AWS-built node autoscaler that launches right-sized EC2 instances directly for waiting pods and consolidates nodes to save money. Faster and more flexible than Cluster Autoscaler.
KEDA · Kubernetes Event-Driven Autoscaling
An add-on that scales pods based on external events such as queue length in SQS or Kafka, including scaling to zero.
kernel
The core of the operating system that controls the hardware, memory, processes and network, and that every program asks for help through system calls.
kernel image
The compressed file containing the Linux kernel, which the bootloader loads into memory when the machine starts.
KEV · Known Exploited Vulnerabilities catalogue
A list from the US agency CISA of vulnerabilities attackers are known to be actively exploiting — fix these first.
Key pair
Two matching keys used to prove identity. The public key can be shared and is placed on servers; the private key stays secret with you. Whatever the private key signs, the public key can verify.
KMS · Key Management Service
AWS's service for creating and controlling encryption keys. Keys never leave KMS; services ask it to encrypt or decrypt, and every use is logged.
KServe
A Kubernetes platform for serving machine-learning models through InferenceService resources, with autoscaling and canary rollouts.
kube-proxy
A component on each node that programs networking rules (iptables or IPVS) so traffic to a Service reaches one of its pods.
Kubecost
Tools that split Kubernetes cluster costs by namespace, label and workload, based on what pods request and use.
kubectl
The command-line tool for talking to a Kubernetes cluster: kubectl get pods, kubectl apply -f, kubectl logs and so on.
kubelet
The agent on each Kubernetes node that receives pod assignments, starts the containers through the runtime, runs health checks and reports status back.
Kubernetes
An open-source system that runs containers across a group of machines and keeps them in the state you describe: the right number of copies, healthy, connected and updated safely.
Kubernetes controller
A program that repeatedly compares what you asked for (desired state) with what exists (actual state) and makes changes to close the gap, like a thermostat.
Kubernetes operator
A controller that encodes the know-how to run a specific application, such as a database, using custom resources (for example creating backups or handling failover automatically).
Kubernetes scheduler
The Kubernetes component that picks which node each new pod should run on, based on free resources and rules such as affinity and taints.
Kubernetes Secret
A Kubernetes object for sensitive values such as passwords. By default it is only base64-encoded (not encrypted), so access must be restricted and encryption at rest turned on.
Kustomize
A tool built into kubectl that customises plain YAML without templates: you keep a base and add small overlays (patches) for each environment.
Kustomize overlay
In Kustomize, a folder of small changes (such as replica count or image tag) applied on top of a shared base for one environment.
KV cache · key-value cache
GPU memory an LLM uses to remember the conversation so far for each request; it grows with context length and the number of concurrent requests.
Kyverno
A Kubernetes policy engine where rules are written as YAML instead of a programming language. It can validate, change and generate resources and verify image signatures.

L

L4 · layer 4 (transport)
The transport layer: TCP and UDP, ports and connections. A layer-4 load balancer forwards connections without reading their content.
L7 · layer 7 (application)
The application layer, such as HTTP. A layer-7 load balancer can read URLs, headers and cookies to make routing decisions.
label
A tag attached to something to identify, filter or group it. In Prometheus and Kubernetes it is a key-value pair such as app="orders"; in Jenkins, a name that decides which agents may run a job.
Lambda
AWS's serverless functions: you upload code and it runs in response to events, billed per request and per millisecond, with no servers to manage.
landing zone
A ready-made, secure multi-account AWS setup — accounts, network, logging, guardrails and access — that new workloads are placed into.
latency
The delay before data arrives or a request is answered, usually in milliseconds. Low latency feels instant; high latency feels sluggish.
launch template
A saved recipe for EC2 instances (AMI, instance type, security groups, user data) used by Auto Scaling groups and node groups.
least privilege
Giving each person or program only the permissions it actually needs, nothing more, so a mistake or breach can do less damage.
lifecycle rule
An automatic rule that moves old data to cheaper storage or deletes it after a set time, for example "delete logs after 90 days" in S3 or "keep the last 30 images" in ECR.
LLM · large language model
A very large model trained on text that generates language, such as the models behind chat assistants.
load average
The average number of processes running or waiting to run over 1, 5 and 15 minutes. Compare it to the number of CPU cores: a load of 8 on a 4-core machine means work is queuing.
load balancer
A component that spreads incoming traffic across several servers and stops sending to unhealthy ones, like a receptionist directing visitors to free counters.
location block
An Nginx block that matches part of the URL path and says how to handle it: serve files, proxy to a backend or return a response.
lockfile
A file recording the exact version of every dependency that was resolved, so every future build installs exactly the same set.
log
A timestamped text record of something that happened, such as "payment failed for order 123". Detailed but bulky.
Loki
A log storage system from Grafana that indexes only labels, not full text, which makes it much cheaper to run than full-text search systems.

M

MAC address · media access control address
The hardware address burned into a network card, used to deliver frames on the local network. IP addresses route between networks; MAC addresses deliver within one.
main branch
The primary branch of a repository that represents the official, releasable code.
Make (build tool)
A classic build tool that runs commands defined as targets in a Makefile, such as make test or make deploy, giving a team one consistent set of commands.
managed node group
A set of EKS worker nodes created and updated by AWS, backed by an Auto Scaling group.
Managed service
A cloud service where the provider runs the hard parts — patching, backups, failover — for you, such as Amazon RDS for databases or EKS for the Kubernetes control plane.
manifest
A file declaring what something should contain. In Kubernetes, a YAML (or JSON) file describing an object you want, such as a Deployment with three replicas; for container images, the index of an image's layers.
matrix
A CI feature that runs the same job for each combination of settings, for example Java 17 and 21 on Linux and Windows.
Maven
A Java build tool driven by an XML file (pom.xml) that downloads dependencies, compiles, tests and packages code through a fixed series of phases.
merge conflict
What happens when two branches changed the same lines differently; Git stops and asks a person to choose the right result.
metric
A number measured over time, such as requests per second or memory used, cheap to store and easy to graph and alert on.
MFA · multi-factor authentication
Signing in with something you know (a password) plus something you have (a phone app or security key).
Microservice
A small service that does one job (orders, payments, search) and talks to others over the network, so each can be built, deployed and scaled separately.
MLflow
An open-source platform for tracking ML experiments, packaging models and managing them in a model registry.
MLOps
Applying DevOps practices — versioning, automation, CI/CD, monitoring — to machine-learning data, training, deployment and retraining.
model registry
A store of versioned, approved machine-learning models with their metrics and lineage, which deployments reference.
monitoring
Watching known signals (error rates, CPU, disk) with dashboards and alerts to spot problems you expect.
MSS · maximum segment size
The largest chunk of data TCP puts in one packet, normally the MTU minus 40 bytes of headers (1,460 bytes on standard Ethernet).
mTLS · mutual TLS
TLS where both sides show certificates, so the server checks the client too. Used between microservices so each service can prove its identity.
MTTD · mean time to detect
The average time between a problem starting and someone noticing it.
MTTR · mean time to restore
The average time it takes to get a service working again after a failure. Lower is better.
MTU · maximum transmission unit
The largest packet a network link carries in one piece, usually 1,500 bytes on Ethernet (9,001 inside some AWS networks). Bigger packets must be split or are dropped.
Multi-AZ
Running a standby copy of something (often a database) in a second Availability Zone so it can take over automatically if the first fails.
multi-stage build
A Dockerfile with several FROM stages: build tools are used in early stages and only the finished result is copied into a small final image.
MX · mail exchanger
A DNS record naming the servers that receive email for a domain, with priorities.

N

NACL · network access control list
A subnet-level firewall in a VPC with numbered allow and deny rules. It is stateless: return traffic must be allowed explicitly.
name server
A server that answers DNS questions. The NS records of a domain list which name servers are authoritative for it.
namespace
In Kubernetes, a named section of a cluster used to group and separate resources, for example one per team or application; names only need to be unique within a namespace. (Linux kernel namespaces, which isolate containers, are a different thing.)
namespaces (Linux)
A kernel feature that gives a process its own private view of things such as process IDs, network interfaces and mounted filesystems. Containers are built from namespaces plus cgroups.
nano
A simple text editor for the terminal. Ctrl-O saves, Ctrl-X exits; the shortcuts are listed at the bottom of the screen.
NAT · network address translation
Rewriting the addresses on packets as they pass a router. Typically lets many machines with private addresses share one public address to reach the internet, like a company switchboard.
NAT gateway
An AWS-managed service that lets resources in private subnets make outbound connections to the internet while staying unreachable from outside. Charged per hour and per GB processed.
ndots
A setting in a Linux machine's resolver configuration: names with fewer dots than this number are first tried with the search domains appended. Kubernetes sets ndots:5, which can multiply DNS lookups.
negative caching
Remembering "this name does not exist" for a while, so repeated lookups for a missing name don't hit the servers each time.
NetworkPolicy
A Kubernetes firewall rule for pods, saying which pods or addresses may talk to which, on which ports.
Nginx
A fast, widely used web server and reverse proxy that serves files, terminates TLS, balances load across backends and enforces limits, configured with nested blocks of directives.
NLB · Network Load Balancer
AWS's layer-4 load balancer for TCP and UDP. Very fast, keeps the client's IP and can have fixed IP addresses.
NODATA
A DNS answer meaning "this name exists, but it has no record of the type you asked for" (for example no AAAA record).
node
A machine (usually a VM such as an EC2 instance) that runs your containers in a cluster.
NodePort
A Service type that opens the same port (30000–32767) on every node and forwards it to the pods.
NS · name server record
A DNS record listing the authoritative name servers for a zone; it is how delegation works.
NXDOMAIN · non-existent domain
The DNS answer meaning "this name does not exist at all."

O

observability
How well you can understand what a system is doing from the data it produces — metrics, logs and traces — including problems nobody predicted.
OCI · Open Container Initiative
The industry standard for container image and registry formats. "OCI artifact" means any file (such as a Helm chart) stored in a container registry using that standard.
OIDC · OpenID Connect
A standard way for one system to issue signed identity tokens that another system trusts. GitHub Actions and EKS use it so workloads can get AWS credentials without stored keys.
on-call
Being the engineer responsible for responding to urgent alerts during a set period, usually on a rotation.
On-Demand
The standard EC2 pricing: pay by the second with no commitment and no interruption.
OOM · out of memory
When memory runs out, the kernel's OOM killer picks a process and kills it. In Kubernetes, a container that exceeds its memory limit is killed and shown as OOMKilled.
OPA · Open Policy Agent
A general-purpose policy engine. You write rules in a language called Rego, and OPA answers "allow or deny?" for Kubernetes, CI pipelines, APIs and more.
OpenSearch
A search and analytics engine often used to store and full-text search logs. OpenSearch is the open-source fork of Elasticsearch, available as an AWS service.
OpenTelemetry
The open standard and toolkit for producing traces, metrics and logs from applications and sending them to any backend, so you aren't tied to one vendor.
orchestrator
Software that decides where containers run across many machines, restarts failed ones, scales them and connects them. Kubernetes is the most widely used.
origin (CDN)
The server or bucket a CDN fetches content from when it doesn't have a fresh copy.
OSI model · Open Systems Interconnection model
A seven-layer way of describing networking, from physical cables (layer 1) up to applications (layer 7). Engineers use the layer numbers as shorthand.
OU · organisational unit
A folder of AWS accounts within AWS Organizations, such as "Prod" or "Sandbox", to which shared policies are applied.
outlier detection
Temporarily removing an endpoint from load balancing after it returns several errors in a row, so one bad pod stops receiving traffic.
overlay network
A virtual network built on top of the real one, usually by wrapping packets (for example with VXLAN), so containers on different machines can talk as if they were on the same network.

P

PaaS · Platform as a Service
A cloud service that runs your application for you: you hand over code or a container, and the provider handles servers, operating systems and scaling.
Packer
A tool that builds machine images (such as AMIs) from a template: it launches a temporary machine, runs setup steps and saves the result.
packet
A small chunk of data sent across a network with addressing information attached, like a letter in an envelope. Large data is split into many packets.
paging
Waking or interrupting the on-call engineer with an urgent alert through a tool such as PagerDuty, as opposed to a ticket that can wait.
Parameter Store
A Systems Manager feature for storing configuration values and simple secrets, optionally encrypted with KMS.
PCI DSS · Payment Card Industry Data Security Standard
The security standard any company handling credit card data must follow.
PDB · PodDisruptionBudget
A rule saying how many pods of an app may be down at once during voluntary disruptions such as node upgrades, so maintenance never takes everything offline.
Pending
A pod status meaning it hasn't been placed on a node yet, usually because no node has enough free resources or a scheduling rule can't be met.
percentile
The value below which a given share of measurements fall. p99 latency of 800 ms means 99% of requests were faster than 800 ms and 1% were slower.
permission set
In IAM Identity Center, a bundle of permissions that becomes a role in each AWS account it is assigned to.
permissions boundary
A ceiling on the permissions an IAM user or role can ever have, whatever other policies say. Used to let teams create roles safely.
PersistentVolume
A piece of storage (such as an EBS disk) made available to the cluster.
Pets vs cattle
A way to describe servers. Pets are hand-made and nursed back to health; cattle are identical, built by automation and simply replaced when they fail. Modern teams aim for cattle.
PgBouncer
A lightweight connection pooler for PostgreSQL that lets many clients share a small number of database connections.
PID · process ID
The number Linux gives each running process. PID 1 is special: it is the first process and must handle signals and reap finished children.
pipeline
An automated sequence of steps (build, test, scan, deploy) that runs for each change, defined as code in the repository.
pipeline as code
Defining your CI/CD pipeline in a file stored with the application code (such as a Jenkinsfile or a GitHub Actions workflow), so it is reviewed and versioned like code.
PITR · point-in-time recovery
Restoring a database to any chosen moment by replaying archived logs on top of a backup, for example one second before a bad delete.
Platform engineering
Building an internal "platform" — templates, self-service tools, shared clusters and pipelines — so other teams can ship software without becoming infrastructure experts.
plugin
An add-on that extends a tool with extra features, such as Jenkins plugins for Git, Kubernetes agents or credentials.
pod
The smallest unit Kubernetes runs: one or more containers that share a network address and storage and are always scheduled together on the same node.
Pod Security Standards
Kubernetes' built-in security levels for pods — privileged, baseline and restricted — enforced per namespace with labels.
policy as code
Writing rules (such as "no public buckets" or "containers must not run as root") as code that tools enforce automatically in CI and in clusters.
POM · Project Object Model
Maven's project file, describing the project's coordinates, dependencies and plugins.
POP · point of presence
A CDN data centre close to users where content is cached and served.
port
A number from 0 to 65535 that picks which program on a machine receives the traffic. If the IP address is the building, the port is the flat number: 443 for HTTPS, 22 for SSH, 5432 for PostgreSQL.
postmortem
A written review after an incident covering what happened, why, and what will change. "Blameless" means it focuses on fixing systems, not blaming people.
PR · pull request
A request to merge a branch into another, where teammates review the change and automated checks run before it is accepted.
prefix delegation
A VPC CNI mode that assigns blocks of 16 IP addresses to each network interface instead of single addresses, so nodes can run many more pods.
private IP range
Address blocks reserved for internal networks and never routed on the internet: 10.0.0.0/8, 172.16.0.0/12 and 192.168.0.0/16.
probe (Kubernetes)
A health check Kubernetes runs on a container. Liveness: "is it stuck? restart it." Readiness: "can it take traffic now?" Startup: "has it finished starting?"
process (Linux)
A running program, with its own memory and a process ID (PID). Opening the same app twice gives two processes.
Prometheus
The CNCF monitoring system that regularly collects (scrapes) metrics from your apps over HTTP, stores them as time series and evaluates alert rules.
prompt injection
An attack where text given to an LLM (by a user, a document or a web page) contains instructions that hijack the application.
PromQL · Prometheus Query Language
The query language for Prometheus, used in dashboards and alerts, for example rate(http_requests_total[5m]).
provenance
A signed record of how an artifact was built: which source commit, which build system, which steps.
provisioned concurrency
A Lambda setting that keeps a number of environments initialised in advance so those requests have no cold start, at an hourly cost.
provisioner (Packer)
A Packer step that configures the temporary machine, such as running shell scripts or copying files.
proxy
A middleman that makes requests on someone else's behalf. A forward proxy acts for clients (for example to control outbound access); a reverse proxy acts for servers.
PTR · pointer record
A DNS record used for reverse lookups: turning an IP address back into a name.
PVC · PersistentVolumeClaim
A pod's request for storage ("I need 20 GiB of fast disk"). Kubernetes finds or creates a matching PersistentVolume and attaches it.

Q

quantisation
Storing model weights with fewer bits (such as 8 or 4) to cut memory and cost, with a small loss in quality.
QUIC
A modern transport protocol built on UDP that gives TCP-like reliability plus built-in encryption and faster connection setup. HTTP/3 runs on it.

R

RAG · retrieval-augmented generation
Answering with an LLM after first retrieving relevant documents and adding them to the prompt, instead of retraining the model.
rate limiting
Capping how many requests a client may make in a period, rejecting the excess (usually with 429), to protect services from abuse and overload.
RBAC · role-based access control
Granting permissions through roles: a role lists allowed actions, and a binding gives that role to a user or service account. Kubernetes uses it to control who can do what.
RCP · resource control policy
An AWS Organizations policy that sets the maximum permissions on resources in member accounts, for example "nobody outside our organisation can read our S3 buckets".
RDS · Relational Database Service
AWS's managed SQL databases (PostgreSQL, MySQL and others): AWS handles installation, patching, backups and failover.
RDS Blue/Green Deployments
An RDS feature that creates a synchronised copy of a database (for example on a new version), lets you test it, then switches over in about a minute.
RDS Proxy
A managed connection pool in front of an RDS or Aurora database that shares a smaller number of database connections among many clients.
Reachability Analyzer
An AWS tool that checks whether traffic can get from one resource to another by examining routes, security groups and NACLs, and tells you which one blocks it.
read replica
A copy of a database that receives changes from the primary and serves read-only queries, spreading the load.
rebase
Replaying your commits on top of another branch's latest commit, producing a straight-line history. It rewrites commit IDs, so don't rebase commits others already use.
reconcile
The act of comparing desired state with actual state and fixing any difference. Controllers and GitOps tools reconcile continuously.
RED method · Rate, Errors, Duration
A simple checklist for monitoring request-based services: how many requests (rate), how many fail (errors), how long they take (duration).
Redis
An in-memory key-value store used as a cache, session store and message broker; ElastiCache and MemoryDB run it on AWS.
region
A geographic area where AWS has data centres, such as ap-south-1 (Mumbai) or ap-south-2 (Hyderabad). Most resources live in one region.
registrar
The company you buy a domain name from. It tells the TLD which name servers are authoritative for your domain.
registry
A server that stores and serves container images, such as Docker Hub, Amazon ECR or GitHub Container Registry.
Rego
The policy language used by Open Policy Agent. Rules describe conditions, such as "deny if a container image doesn't come from our registry".
Relative path
A file location written from the folder you are in now, such as logs/app.log. .. means the folder above.
ReplicaSet
The Kubernetes object that keeps an exact number of pod copies running. Deployments create and manage ReplicaSets for you.
replication lag
How far a replica is behind the primary. Reads from a lagging replica return old data, and an asynchronous failover can lose the lagging changes.
replication slot
A PostgreSQL record that keeps WAL until a replica or consumer has read it; an abandoned slot can fill the primary's disk.
repository
A storage place for versioned content. A Git repository is a project folder with its full change history (hosted on GitHub or GitLab); an image or package repository (in ECR, Nexus or Helm) holds container images or packages.
requests and limits
Requests are the CPU and memory a container is guaranteed and are used for scheduling; limits are the most it may use. Exceeding the memory limit gets the container killed; exceeding CPU limits slows it down (throttling).
reserved concurrency
A Lambda setting that guarantees a function part of the account's concurrency and also caps it at that number.
resolver
The DNS server that does the legwork for you: it asks the root, TLD and authoritative servers in turn, caches the answers and hands back the final IP address. Your ISP, 8.8.8.8 and the AWS VPC resolver are resolvers.
resource-based policy
A policy attached to a resource (such as an S3 bucket or KMS key) saying which identities, including ones in other accounts, may use it.
retry
Trying a failed request again. Safe only for operations that can be repeated, and should wait longer each time (exponential backoff) with some randomness (jitter).
reusable workflow
A GitHub Actions workflow designed to be called by other workflows, letting a platform team share one standard pipeline across many repositories.
reverse proxy
A server that sits in front of applications and forwards client requests to them, often adding TLS, caching or routing. NGINX and Envoy are common examples.
rightsizing
Matching instance or container sizes to what workloads actually use, to stop paying for idle capacity.
rollback
Returning to the previous working version after a bad release.
rolling update
Replacing an application a few instances at a time, so some copies keep serving while the new version comes up.
root server
The servers at the top of the DNS tree. They don't know final answers; they point resolvers to the right TLD servers.
root user
The all-powerful administrator account. On Linux it is user ID 0; in AWS it is the email login that created and owns the account.
Route 53
AWS's DNS service: it hosts your domain's records, can register domains, and routes traffic using health checks and policies such as weighted, latency and failover.
route table
A list of rules saying where traffic for each destination range should go, for example "0.0.0.0/0 → internet gateway". The most specific matching rule wins.
router
A device that forwards packets between networks by reading the destination IP address and choosing the next hop.
routing policy
In Route 53, the rule that decides which answer to return for a name: simple, weighted, latency, failover, geolocation, geoproximity, IP-based or multivalue.
RPO · recovery point objective
The most data you can afford to lose after a disaster, measured in time, such as "no more than 5 minutes of transactions".
RST · reset
A TCP message that aborts a connection immediately, for example when nothing is listening on the port. Clients usually report it as "connection refused" or "connection reset".
RTO · recovery time objective
The longest acceptable time to get a service back after a disaster, such as "within 1 hour".
RTT · round-trip time
How long a message takes to reach the other side and for the reply to come back. Chennai to Mumbai is roughly 25 ms; India to the US east coast is about 200 ms.
runbook
A short, step-by-step guide for handling a specific alert or task: what it means, what to check and how to fix it.
runner
The machine that executes a CI job. GitHub-hosted runners are fresh VMs GitHub provides; self-hosted runners are machines you run yourself.

S

S3 · Simple Storage Service
AWS's object storage: you store files (objects) in buckets and fetch them over HTTPS. Extremely durable and effectively unlimited in size.
SaaS · software as a service
Software you use over the internet on a subscription, such as GitHub or Slack, instead of installing and running it yourself.
SAM · Serverless Application Model
A CloudFormation extension and CLI for defining, testing and deploying serverless applications.
SAML · Security Assertion Markup Language
An older single sign-on standard where an identity provider sends signed XML statements about the user.
SAST · static application security testing
Scanning source code for security bugs without running it.
saturation
How "full" a resource is — work queuing because CPU, memory, disk or connections are maxed out. Usually the first sign of trouble before errors appear.
Savings Plans
A discount (up to about 70%) in exchange for committing to spend a fixed amount per hour on compute for one or three years.
SBOM · software bill of materials
A list of every component and version inside a piece of software or a container image, like an ingredients label. It lets you find quickly which services contain a newly vulnerable library.
SCA · software composition analysis
Checking the third-party libraries your code uses for known vulnerabilities and licence problems.
SCP · service control policy
An AWS Organizations policy that sets the maximum permissions for every identity in the accounts below it. It never grants anything; it only limits.
scrape
Prometheus fetching the current metric values from a target's /metrics web page at a regular interval.
SDK · software development kit
A library that makes it easy to call a service from a programming language, such as boto3 for AWS in Python.
Sealed Secrets
A Kubernetes tool where you encrypt a secret with the cluster's public key; only the controller in that cluster can decrypt it, so the encrypted file can live in Git.
search domain
Suffixes the resolver adds to short names, so orders can become orders.default.svc.cluster.local inside Kubernetes.
seccomp · secure computing mode
A Linux feature that limits which system calls a process may make. RuntimeDefault blocks risky ones for containers.
secret
Sensitive information such as passwords, API keys, tokens and private keys that must be kept out of code and logs.
secret rotation
Regularly replacing a password, key or certificate with a new one so a leaked old value stops working.
Secrets Manager · AWS Secrets Manager
An AWS service that stores secrets such as database passwords, controls who can read them and can rotate them automatically.
security group
A virtual firewall attached to an instance, load balancer or pod network interface. It only has allow rules and remembers connections, so replies are allowed back automatically (stateful).
Security Hub
A central AWS dashboard that collects security findings from many services and checks accounts against standards such as CIS.
semantic versioning
Version numbers in the form MAJOR.MINOR.PATCH: bump MAJOR for breaking changes, MINOR for new compatible features, PATCH for fixes.
serverless
Running code or services without managing servers yourself: the provider handles capacity and you pay only for use (for example Lambda, Fargate, DynamoDB).
SERVFAIL · server failure
A DNS answer meaning "I couldn't get a valid answer," often because authoritative servers are unreachable or DNSSEC validation failed.
Service
A Kubernetes object that gives a group of pods one stable name and virtual IP and spreads traffic across them, even as individual pods come and go.
service mesh
An extra layer (usually proxies next to every pod) that handles service-to-service traffic: encryption with mTLS, retries, traffic splitting and metrics, without changing app code.
ServiceAccount
An identity for programs running in pods, used to call the Kubernetes API and (with IRSA or Pod Identity) AWS.
Session Manager
A Systems Manager feature that opens a shell on an instance through AWS, with no open SSH port, controlled by IAM and fully logged.
SHA · Secure Hash Algorithm
A fingerprint calculated from data; any change to the data changes the fingerprint completely. Git commit IDs and image digests are SHA hashes.
shared library
Reusable Jenkins pipeline code kept in its own repository and loaded by many Jenkinsfiles, so teams don't copy and paste pipeline logic.
shell
The command-line program that reads what you type and runs it. Bash is the most common shell on Linux.
Shell prompt
The text the shell prints when it is ready for a command, such as deploy@web-1:~/app$. It shows your user, the machine and the current folder; # instead of $ means you are root.
ShellCheck
A tool that finds bugs and risky patterns in shell scripts, such as missing quotes around variables.
shift left
Moving testing and security checks earlier in the development process, where problems are cheaper and faster to fix.
showback
Showback reports each team's cloud costs to them; chargeback actually bills those costs to their budgets.
sidecar
A helper container running next to the main app in the same pod, such as a log shipper or a mesh proxy.
SIGKILL · kill signal
An immediate, uncatchable stop. The program gets no chance to clean up. Sent when a process ignores SIGTERM for too long or runs out of memory.
SIGTERM · terminate signal
A polite request to shut down: the program can finish its work, close connections and then exit. Kubernetes and Docker send it first when stopping a container.
SLA · service level agreement
A contract with customers promising a level of service, with penalties (such as refunds) if it is missed. It should be looser than the internal SLO.
SLI · service level indicator
A measurement of how well a service is doing for users, expressed as good events divided by all valid events — for example the share of requests that succeed within 400 ms.
SLO · service level objective
An internal target for an SLI over a time window, such as "99.9% of checkout requests succeed over 28 days".
SLSA · Supply-chain Levels for Software Artifacts
A framework (pronounced "salsa") of levels describing how trustworthy a build process is, centred on signed provenance.
SNAPSHOT
A Maven version ending in -SNAPSHOT that means "work in progress": each new build replaces it, so it is not reproducible.
SnapStart
A Lambda feature that resumes functions from a snapshot of an initialised environment, cutting cold starts for Java, Python and .NET.
SNAT · source network address translation
NAT that rewrites the sender's address, so replies come back to the translator. Used for outbound internet access.
SNI · Server Name Indication
A field in the TLS handshake where the client says which hostname it wants, so one IP address can serve many websites with different certificates.
SNS · Simple Notification Service
A managed publish/subscribe service that pushes each message to all subscribers, such as SQS queues, Lambda functions, email or SMS.
SOA · start of authority
The record at the top of every zone holding its admin details: primary server, contact, serial number and timers, including how long to cache "name does not exist" answers.
socket
One end of a network conversation, identified by an IP address plus a port. A TCP connection is a pair of sockets.
software catalogue
A central list of every service, API and resource with its owner, dependencies and links.
SOPS · Secrets OPerationS
A tool that encrypts the values in YAML or JSON files (with KMS or age keys) so secrets can be stored safely in Git and decrypted at deploy time.
span
One timed step inside a trace, such as "call payments API: 120 ms".
SPF · Sender Policy Framework
A TXT record listing which servers may send email for a domain, so receivers can spot forged senders.
SPIFFE · Secure Production Identity Framework for Everyone
A standard for workload identities, such as spiffe://cluster.local/ns/payments/sa/api, carried in short-lived certificates.
split-horizon DNS
Giving different answers for the same name depending on who asks, for example a private IP inside the company network and a public IP on the internet.
Spot
Spare EC2 capacity sold at a large discount (often 60–90%) that AWS can take back with a two-minute warning. Good for work that can be interrupted.
SQL injection
An attack where malicious input is inserted into a database query, for example to read or delete data, possible when apps build queries from raw user input.
SQS · Simple Queue Service
A managed message queue: producers drop messages in, consumers pick them up and delete them when done, so the two sides don't need to run at the same speed.
SRE · site reliability engineering
Google's approach to running production systems with software engineering: setting reliability targets, measuring them and automating operations work.
SRV · service record
A DNS record that gives the host and port for a named service, for example where to find a SIP or LDAP server.
SSH · Secure Shell
The encrypted protocol for logging into and running commands on remote Linux machines.
SSL · Secure Sockets Layer
The old name and predecessor of TLS. All SSL versions are now insecure and disabled, but people still say "SSL certificate" out of habit.
SSO · single sign-on
Signing in once with your company login and getting access to many systems. In AWS this is IAM Identity Center.
SSRF · server-side request forgery
An attack that tricks a server into making requests on the attacker's behalf, for example to the cloud metadata address to steal credentials.
stack trace
The list of function calls a program was in when an error happened, printed with the error; it shows where in the code the failure occurred.
StackSets
A CloudFormation feature that deploys the same stack to many AWS accounts and regions at once.
Staging
A copy of production used for the final checks before a release reaches real users. The closer it is to production, the more problems it catches.
StatefulSet
A Kubernetes object for stateful apps such as databases: pods get stable names (db-0, db-1), their own persistent disks, and start and stop in order.
Status code
The three-digit number a web server sends back: 2xx success (200 OK), 3xx redirect, 4xx a problem with the request (404 not found), 5xx a server failure (500, 503).
stdout · standard output / standard error
The default output streams of a program: stdout for normal results, stderr for errors and diagnostics. Containers should write logs to them so the platform can collect them.
Step Functions
AWS's workflow service: state machines that coordinate steps with retries, error handling, parallel branches and waits.
StorageClass
A Kubernetes template that says how to create disks on demand, for example "EBS gp3, encrypted, created in the zone where the pod runs".
STRIDE · Spoofing, Tampering, Repudiation, Information disclosure, Denial of service, Elevation of privilege
A checklist for threat modelling: six kinds of attack to consider for each part of a system.
structured logging
Writing logs as fields (usually JSON) instead of free text, so they can be searched and filtered reliably.
STS · Security Token Service
The AWS service that issues temporary credentials when someone assumes a role or signs in through federation.
subnet
A smaller slice of a network's address range. In AWS a subnet lives in exactly one Availability Zone and is public or private depending on its route table.
subnet mask
Another way of writing the network part of an address: /24 is the same as 255.255.255.0.
sudo
A command that lets an allowed user run a single command as root, with the action logged.
supply chain
Everything that goes into your software before it runs: your code, open-source dependencies, build tools, CI systems and registries. Attacks on any link can compromise the result.
swap
Disk space the kernel uses as overflow memory. It prevents crashes but is far slower than RAM; Kubernetes nodes usually run without it.
SYN · synchronise
The first message of a TCP handshake: "I want to start a connection, here is my starting sequence number."
systemd
The service manager on most Linux systems: it starts services at boot, restarts them if they crash and collects their logs (journald).
Systems Manager · AWS Systems Manager
A set of AWS tools for operating servers, including Session Manager (secure shell access without SSH), Parameter Store and patching.

T

taint
A taint on a node repels pods; only pods with a matching toleration may run there. Used to reserve nodes, for example for GPU workloads.
target (Make)
A named entry in a Makefile, such as build or test, with the commands that produce it.
target group
The list of destinations (instances, IP addresses or pods) a load balancer sends traffic to, together with the health check used to decide which are healthy.
TCP · Transmission Control Protocol
A protocol that gives a reliable, ordered conversation between two programs. It numbers every byte, resends anything lost and slows down when the network is busy. Web pages, SSH and databases use it.
Team Topologies
A model for organising teams (stream-aligned, platform, enabling, complicated-subsystem) and how they interact, to reduce cognitive load.
TechDocs
Backstage's docs-as-code feature: Markdown kept next to the code, built with MkDocs and shown in the portal.
Terminal
The window where you type commands and see their output. The program inside it that understands the commands is the shell.
Terraform
HashiCorp's infrastructure-as-code tool. You describe resources in HCL files; terraform plan shows what will change and terraform apply makes it so. Works across AWS, Azure, Kubernetes and more. OpenTofu is its open-source fork.
Terraform module
A reusable package of Terraform code with inputs and outputs, such as a standard VPC or EKS cluster.
terraform plan
Terraform's preview of what it would create, change or destroy to match your code, without doing anything yet.
Terraform provider
A Terraform plugin that knows how to talk to one platform's API, such as AWS or Kubernetes.
Terraform state
Terraform's record of which real resources it manages and their current settings. It is kept in a shared, locked backend such as S3 so a team doesn't overwrite each other.
Thanos
Systems that extend Prometheus for many clusters and long retention by storing metrics in object storage such as S3 and combining queries.
threat modelling
Systematically thinking through how a system could be attacked and what defences it needs, ideally before building it.
three-way handshake
How TCP opens a connection: the client sends SYN, the server replies SYN-ACK, and the client answers ACK. Only then can data flow.
throttling
Being slowed down because you hit a limit — for CPU, the container gets paused for part of each period; for APIs, requests are rejected or delayed.
throughput
How much data can move per second, for example 1 Gbps. A wide road (bandwidth) is different from a short road (latency).
time to first token
How long an LLM takes to start returning its answer; the main measure of how responsive it feels.
timeout
The maximum time to wait for something before giving up. Every network call should have one.
TLD · top-level domain
The last part of a domain name, such as .com, .in or .org. TLD servers know which name servers hold each domain under them.
TLS · Transport Layer Security
The encryption layer that protects data travelling over a network and proves the server is who it claims to be, using certificates. It replaced the older SSL.
toil
Manual, repetitive operations work that could be automated and adds no lasting value, such as resizing disks by hand every week.
token
A string that proves identity or permission, like a temporary pass. Short-lived tokens limit the damage if stolen.
token (LLM)
A small piece of text (about three-quarters of an English word) that LLMs read and write; cost, limits and speed are measured in tokens.
topology spread
A rule that spreads a set of pods evenly across zones or nodes so one failure doesn't take out every copy.
trace
The record of one request's journey through every service it touched, with timings for each step (span), showing where time was spent.
Transit Gateway
A central AWS router that connects many VPCs and on-premises networks in a hub-and-spoke layout, instead of a tangle of peering links.
Trivy
A popular open-source scanner for vulnerabilities and misconfigurations in container images, file systems, IaC and Kubernetes.
trunk-based development
A branching style where everyone merges small changes into the main branch at least daily, keeping branches short-lived; unfinished features hide behind feature flags.
trust policy
The part of an IAM role that says who may assume it, for example "the GitHub Actions workflows of repository acme/orders".
TTL · time to live
In DNS, how many seconds a resolver may keep a cached answer before asking again. A TTL of 300 means changes take up to five minutes to be seen. (In IP packets, TTL is a hop counter that stops packets looping forever.)
TXT · text record
A DNS record holding free text, mostly used to prove domain ownership and publish email security policies such as SPF and DMARC.

U

UDP · User Datagram Protocol
A lightweight protocol that sends individual messages with no connection setup, no ordering and no resending. It is fast and simple, so DNS lookups, video calls and QUIC use it.
unit cost
Cloud cost divided by a measure of business value, such as cost per order or per active user, to judge whether spending growth is healthy.
Unix signal
A short message the kernel delivers to a process, such as "please stop" (SIGTERM) or "stop now" (SIGKILL).
upstream (Nginx)
In Nginx, the backend servers a request is passed to, often grouped in an upstream block with a load-balancing method.
USE method · Utilisation, Saturation, Errors
A checklist for monitoring resources such as CPU, memory and disks: how busy they are, how much work is queued, and how many errors occur.
user data
A script you give an EC2 instance that runs on first boot, typically to install or configure software.

V

Valkey
An open-source fork of Redis, run by the Linux Foundation, created after Redis changed its licence in 2024; supported by ElastiCache.
Vault
HashiCorp's secrets manager, which can also generate short-lived credentials on demand (for example a database login that expires in an hour).
vector database
A database that stores embeddings and finds the nearest ones to a query quickly, such as pgvector or OpenSearch.
vim
A text editor found on almost every Linux server. Press i to type, Esc to stop, then :wq to save and quit or :q! to quit without saving.
VirtualService
An Istio resource that controls how requests to a host are routed: matching, traffic weights, timeouts, retries and fault injection.
VLAN · virtual LAN
A way to split one physical network switch into several separate logical networks.
vLLM
An open-source engine for serving large language models efficiently, with continuous batching and careful GPU memory management.
VM · virtual machine
A complete simulated computer, with its own operating system, running on shared physical hardware. EC2 instances are virtual machines.
volume
A unit of storage attached to something. For containers, storage that lives outside the container's own filesystem, so data survives when the container is replaced; in AWS, an EBS disk attached to an instance.
VPA · Vertical Pod Autoscaler
A Kubernetes add-on that recommends or adjusts the CPU and memory requests of pods based on actual use.
VPC · Virtual Private Cloud
Your own private, isolated network inside AWS, with an IP range you choose, split into subnets. Like a gated campus whose roads and gates you design.
VPC CNI
EKS's default pod networking plugin. It gives every pod a real IP address from the VPC, so pods talk to other AWS resources directly.
VPC endpoint
A private doorway from your VPC to an AWS service, so traffic to services like S3 or ECR doesn't go over the internet or through a NAT gateway.
VPC peering
A direct private connection between two VPCs. It is not transitive: if A peers with B and B with C, A still can't reach C.
VPN · virtual private network
An encrypted tunnel across the internet that makes a remote network behave as if it were directly connected.
vulnerability
A weakness in software or configuration that an attacker could exploit.
VXLAN · Virtual Extensible LAN
A way of wrapping network frames inside UDP packets so containers on different hosts can appear to share one network. Used by overlay networks such as Docker Swarm and Flannel.

W

WAF · web application firewall
A firewall that inspects HTTP requests and blocks common attacks such as SQL injection, bad bots and floods from one source. AWS WAF attaches to ALB, CloudFront and API Gateway.
WAL · write-ahead log
The database log where every change is written and flushed before a commit is confirmed; used for crash recovery, replication and point-in-time recovery.
waypoint proxy
In Istio ambient mode, an optional Envoy proxy added for a namespace or service when Layer 7 features (HTTP routing, retries, path rules) are needed.
Working directory
The folder your shell is "in" right now. pwd prints it, cd changes it, and relative paths start from it.

X

xDS
The family of APIs a control plane uses to send listeners, routes, clusters, endpoints and secrets to Envoy proxies while they run.

Y

YAML · YAML Ain't Markup Language
A human-readable data format based on indentation, used for Kubernetes manifests, GitHub Actions workflows, Helm values and much more.

Z

zone
A part of the DNS tree managed as one unit, such as example.com and everything under it unless delegated elsewhere.
ztunnel
Istio ambient mode's per-node proxy that handles mutual TLS, identity and Layer 4 traffic for every pod on the node, without sidecars.