AWS V: Data, Resilience & the Edge: glossary
203 terms, in plain words, with where each one is taught.
- AbortIncompleteMultipartUpload (incomplete multipart upload, incomplete multipart uploads)
- The lifecycle action that deletes the parts of multipart uploads that never finished: invisible in listings, billed until something aborts them. Lesson 38.19
- access pattern (access patterns)
- A question the application asks of its data, such as "all orders of a customer, newest first". In DynamoDB you list them first and shape the keys so that each one is a single cheap request. Lesson 38.10
- access point (EFS) (EFS access point, EFS access points)
- An application's own entry into an EFS file system: a root directory, created with an owner and mode on first use, and a forced POSIX user and group, whatever the client claims. Lesson 38.17
- ACM (AWS Certificate Manager, ACM certificate)
- AWS Certificate Manager: issues and renews TLS certificates for AWS services. CloudFront only reads certificates from us-east-1, whatever Region the origin is in. Lesson 38.31
- alternate domain name (alternate domain names, CNAME (CloudFront))
- A custom name such as cdn.shop.example served by a CloudFront distribution. It needs an ACM certificate in us-east-1 and a Route 53 alias record to the distribution. Lesson 38.31
- amazon-efs-utils (mount -t efs, _netdev, noresvport)
- The EFS mount helper: mount -t efs fs-id:/ /mnt, with TLS in transit (-o tls) and IAM authorization. A permanent mount in /etc/fstab gets _netdev,nofail so a boot does not hang without the network. Lesson 38.17
- Application Recovery Controller (Route 53 Application Recovery Controller, routing control)
- The Route 53 service for Region failover you switch on purpose: readiness checks and routing controls instead of health checks guessing. Lesson 38.28
- ASN (autonomous system number, BGP ASN)
- Autonomous system number: the number that identifies one side of a BGP session, such as your router and the AWS gateway. Lesson 38.41
- association (TGW) (TGW association, associate (TGW))
- Which TGW route table an attachment uses to decide where its own traffic goes. One per attachment. Lesson 38.37
- Aurora (Amazon Aurora, Aurora PostgreSQL, cluster volume)
- AWS's PostgreSQL- and MySQL-compatible database that keeps one storage volume (six copies across three zones) shared by a writer and up to 15 readers. A failover promotes a reader in about 30 s with no data copy. The reader with the lowest promotion tier is promoted first. Serverless v2 scales capacity (ACUs) with load. Compatible, not identical: it is not PostgreSQL itself. Lesson 38.3
- Aurora Global Database (Global Database)
- An Aurora cluster with read-only copies in other Regions, usually under a second behind, that can be promoted when a Region is lost. Lesson 38.3
- AUTH (ElastiCache) (AUTH, AUTH token, in-transit encryption)
- The token (or Valkey / Redis OSS users and ACLs) a cache client must present. With TLS in transit and encryption at rest it is the cache's security; for older engine versions all three are set at creation. Lesson 38.15
- automated backups (automated backup, BackupRetentionPeriod, backup retention period)
- RDS's own backups: a daily snapshot plus the transaction logs, kept 1-35 days, which allow a point-in-time restore. Retention 0 turns them off, and with them PITR and read replicas. They are deleted with the instance unless you choose to retain them; manual snapshots survive. Lesson 38.8
- AWS Backup (aws backup)
- The AWS service that puts every resource type's backups (RDS, DynamoDB, EBS, EFS, S3...) behind one policy: plans say when and how long, selections say what, vaults hold the recovery points, copies go to other Regions and accounts. Lesson 38.24
- AWS Client VPN (Client VPN)
- The managed VPN that connects users' laptops into a VPC, unlike Site-to-Site VPN which joins whole networks. Lesson 38.41
- AWS Network Firewall (Network Firewall, Suricata rules)
- A managed stateful firewall inside the VPC (domain lists, Suricata rules), for traffic that never passes CloudFront or an ALB. Lesson 38.35
- AWS Shield (Shield, Shield Standard)
- AWS's DDoS protection. Shield Standard is on for everyone at no cost and absorbs common network-layer attacks at CloudFront, Route 53 and Global Accelerator. Lesson 38.35
- AWS WAF (WAF (AWS), wafv2, aws wafv2)
- AWS's web application firewall: it inspects HTTP requests at CloudFront, an ALB, API Gateway and a few others, and allows, blocks, counts or challenges them by rules. Never on an NLB or an instance. Lesson 38.35
- AWS:SourceArn (aws:SourceArn)
- The condition key a bucket policy uses to allow cloudfront.amazonaws.com only for requests from one distribution, so nobody can read the bucket through another distribution or directly. Lesson 38.31
- backup job (backup jobs, start-backup-job)
- One run of AWS Backup, scheduled or on demand: CREATED, RUNNING, then COMPLETED, FAILED, or EXPIRED if it could not start inside the start window. Alarm on the failed and expired ones. Lesson 38.24
- backup plan (backup plans, backup rule, backup rules)
- An AWS Backup policy of rules: a schedule (cron, UTC), start and completion windows, a lifecycle (cold storage, delete after N days) and copy actions. Lesson 38.24
- backup selection (backup selections, resource assignment)
- What assigns resources to a backup plan, usually by tag (backup=daily), with the IAM role AWS Backup uses. An untagged resource is not protected, and nothing tells you. Lesson 38.24
- backup vault (backup vaults, vault (AWS Backup))
- The container that holds AWS Backup recovery points, with its own KMS key and access policy. Lesson 38.24
- backup window (PreferredBackupWindow)
- The daily time slot in which RDS takes its automated snapshot. On a Multi-AZ instance it is taken from the standby, so the primary does not pause. Lesson 38.8
- blackhole route (TGW) (TGW blackhole, static blackhole route)
- A Transit Gateway route that drops traffic for a CIDR: on purpose (to cut a VPC off), or by itself when the attachment it pointed at is deleted. Lesson 38.37
- block storage (file storage, object storage)
- The three storage shapes: block storage is a disk for one instance (EBS, instance store), file storage a shared file system many instances mount (EFS, FSx), object storage an HTTP API for whole objects (S3). Lesson 38.17
- cache behavior (cache behaviors, path pattern)
- A CloudFront rule per path pattern (/api/, the default /): which origin, which cache policy, which viewer protocol policy. Lesson 38.31
- cache key (cache key (CloudFront), cache hit ratio, CacheHitRate)
- What a CDN uses to decide that two requests are the same cached object: the path plus any headers, cookies and query strings the cache policy adds. Each extra part lowers the cache hit ratio. Lesson 38.31
- cache policy (cache policies, CachingOptimized, CachingDisabled)
- The CloudFront setting for how long objects stay cached (TTLs) and what goes into the cache key. Managed ones: CachingOptimized (a day by default) and CachingDisabled for APIs. An origin request policy decides separately what is forwarded to the origin without being part of the key. Lesson 38.31
- cache-aside (lazy loading)
- The common caching pattern: look in the cache, and on a miss ask the database and store the answer with a TTL. The TTL is the freshness contract. Lesson 38.15
- Cache-Control (Cache-Control header, no-cache, max-age)
- The HTTP header the origin uses to tell caches how long a response may be kept (max-age) or that it must be revalidated every time (no-cache). CloudFront honours it within the cache policy's limits. Lesson 38.31
- CLF-C02 (Cloud Practitioner, AWS Certified Cloud Practitioner)
- The code of the AWS Certified Cloud Practitioner exam: the entry-level one, what each AWS service is for and how AWS bills. Lesson 38.1
- CloudFront Functions
- Tiny, cheap JavaScript functions that run at CloudFront edges on each request or response: rewrites, redirects, headers. Lesson 38.31
- cluster endpoint (writer endpoint)
- The Aurora DNS name that always points at the current writer, whichever instance that is after a failover. Lesson 38.3
- cluster mode (cluster mode enabled, cluster mode disabled, hash slot)
- The Valkey / Redis OSS setting that splits the keys over several shards (16,384 hash slots) for more memory and write throughput than one node. Clients must be cluster-aware and use the configuration endpoint. Lesson 38.15
- compliance mode (Object Lock in compliance mode)
- The Object Lock mode whose retention nobody can shorten or remove before the date, not even the root user: what regulations require and what makes a backup ransomware-proof. A wrong date cannot be undone either. Lesson 38.22
- condition expression (--condition-expression, ConditionalCheckFailedException, attribute_not_exists)
- A rule a DynamoDB write must pass, such as "only if no item with this key exists yet". When it fails the write is refused with ConditionalCheckFailedException: someone else got there first.
attribute_not_exists(PK)is "create only". The basis of optimistic locking: a version attribute in the condition. Without a condition, put-item replaces an item with the same key. Lesson 38.10 - continuous backup (EnableContinuousBackup)
- An AWS Backup rule option that keeps point-in-time restore for RDS, Aurora and S3: an RPO of minutes instead of the time between two recovery points. Lesson 38.24
- Contributor Insights (CloudWatch Contributor Insights, update-contributor-insights, PKC)
- The CloudWatch feature that ranks the most used and the most throttled keys of a DynamoDB table or index: how you find a hot key. Lesson 38.13
- copy action (copy actions, CopyActions, cross-account copy)
- The part of a backup rule that copies each recovery point to a vault in another Region or another account, with its own retention. Lesson 38.24
- copy-db-snapshot (cross-Region snapshot copy, aws/rds)
- Copying an RDS snapshot to another Region or for another account. An encrypted snapshot needs a KMS key in the target Region, and only one encrypted with a customer managed key can be shared. Lesson 38.8
- Count action (WAF) (Count mode, count mode (WAF))
- Running a WAF rule so it only counts what it matches: how you check a new rule for false positives before letting it block. Lesson 38.35
- credential stuffing (credential-stuffing)
- Trying usernames and passwords leaked from other sites against your login page, at scale. Against it: rate-based rules on the login path, CAPTCHA or challenge actions, account lockouts and MFA. Lesson 38.35
- customer gateway (CGW, customer gateway device)
- AWS's record of your side of a VPN: the router's public IP address and BGP ASN. Lesson 38.41
- DataSync (AWS DataSync)
- The AWS service that moves large data sets from on-premises storage, or between storage services, into S3, EFS or FSx. Lesson 38.19
- DAX (DynamoDB Accelerator)
- DynamoDB Accelerator: an in-memory cache in front of DynamoDB with the same API, for microsecond reads. Lesson 38.10
- DB instance (DB instances, RDS instance, RDS DB instance)
- One managed database server in RDS, of a chosen class (db.t4g.micro, db.r7g.large...) that sets its CPU and memory. You never log in to its host: everything goes through the API and the database protocol. Lesson 38.1
- DB parameter group (DB parameter groups, parameter group (RDS), parameter group)
- The named set of server settings an RDS instance uses instead of a config file. The default group cannot be changed, so the first real change is creating your own and attaching it. Some values are formulas:
max_connections=DBInstanceClassMemory / 9531392, so a small instance class gets fewer than 100 connections. Lesson 38.1 - DB subnet group (DB subnet groups, DBSubnetGroupDoesNotCoverEnoughAZs)
- The list of subnets an RDS instance may be placed in. It must cover at least two Availability Zones, even for a single-AZ instance, so RDS has somewhere to put a standby or a replacement. Lesson 38.1
- DDoS (distributed denial of service, DDoS attack, SYN flood)
- Distributed denial of service: many machines flooding a target with traffic (SYN floods, UDP reflection, HTTP floods) until it cannot serve real users. Lesson 38.35
- dead peer detection (DPD)
- The IPsec check that notices a dead tunnel. With static routing, moving traffic to the other tunnel depends on it. Lesson 38.41
- default retention (default retention (Object Lock), RetainUntilDate, retain-until date)
- The Object Lock period applied to every new object version in a bucket. Deleting such a version fails; deleting the key only adds a delete marker. Lesson 38.22
- default security group (default security group (VPC))
- The security group every VPC comes with. A restored RDS instance gets it unless you pass --vpc-security-group-ids, so the application cannot reach it: the classic "the restore worked but the app times out". Lesson 38.8
- DeleteMarkerReplication (delete marker replication)
- The replication rule setting for whether delete markers are copied too. Off by default, so a delete in the source leaves the copy readable: what you want against mistakes and attackers. Lesson 38.22
- deletion protection (--deletion-protection, final snapshot, --skip-final-snapshot)
- An instance setting that makes delete-db-instance fail until someone turns it off. A delete must also either name a final snapshot or say --skip-final-snapshot on purpose. Lesson 38.1
- Direct Connect (AWS Direct Connect, DX, hosted connection)
- A dedicated private fibre connection from your network to AWS at a Direct Connect location: predictable bandwidth and latency, weeks to order, not encrypted by itself. A VPN is the usual cheaper backup. Dedicated ports of 1, 10, 100 or 400 Gbps, or smaller hosted connections from a partner. Lesson 38.41
- Direct Connect gateway (DX gateway, DXGW)
- A global object that lets one Direct Connect connection reach VPCs in many Regions through their virtual private gateways or Transit Gateways. Lesson 38.41
- distribution (CloudFront) (CloudFront distribution, CloudFront distributions)
- One CloudFront setup: its origins, cache behaviors, certificates and domain names, reached at a *.cloudfront.net name. Lesson 38.31
- DMS (AWS DMS, Database Migration Service)
- AWS Database Migration Service: copies a database into AWS or between engines and keeps the copy in sync until the switch, so the move needs little downtime. Lesson 38.43
- DynamoDB (Amazon DynamoDB, DynamoDB JSON)
- AWS's serverless key-value and document database: single-digit-millisecond reads at any scale, no servers or connections to manage, and the keys designed around the questions you will ask. Items are written in DynamoDB JSON, a type next to every value:
{"S": "shipped"},{"N": "42"}. Lesson 38.10 - DynamoDB Streams (DynamoDB stream, stream (DynamoDB))
- The ordered log of every change to a table's items (TTL deletions included), which another program can read, for example to archive deleted items to S3. Global tables need it. Lesson 38.13
- edge location (edge locations, point of presence, x-amz-cf-pop)
- One of the hundreds of AWS sites close to users where CloudFront ends the TLS connection and serves cached content. The x-amz-cf-pop header names it.
x-amz-cf-pop: FRA60-P5is an edge in Frankfurt. Lesson 38.31 - EFS lifecycle (Infrequent Access (EFS), Archive (EFS), Elastic throughput)
- EFS moving files not read for 30 days to Infrequent Access and after 90 to Archive, cheaper to keep and billed per GB read, and back to Standard on the next access. Elastic throughput, the default now, scales with the workload and bills per GB transferred. Lesson 38.17
- ElastiCache (Amazon ElastiCache, ElastiCache Serverless)
- AWS's managed in-memory cache inside your VPC, running Valkey, Redis OSS or Memcached. Serverless removes node sizing and bills per data stored and per request. Lesson 38.15
- Enhanced Monitoring
- The RDS option that adds operating-system metrics (processes, memory, disk) from an agent on the database host, which you cannot log in to. Lesson 38.1
- ENI (elastic network interface, network interface (AWS), ENIs)
- Elastic network interface: a virtual network card in a subnet, with a private address and security groups. Instances, mount targets, interface endpoints and load balancer nodes all have one. Lesson 38.17
- EvaluateTargetHealth
- The alias record setting that uses the AWS target's own health instead of a separate health check: an ALB with no healthy targets makes the record unhealthy. Lesson 38.28
- eventually consistent read (eventually consistent reads, strongly consistent read, strongly consistent reads)
- A DynamoDB read that may miss a write from the last moment, at half the price of a strongly consistent read, which always sees it. Strong reads exist only on the table and its LSIs. Lesson 38.10
- failover routing (failover records, failover record, DNS failover)
- A Route 53 primary and secondary pair: the primary is answered while its health check passes, the secondary otherwise. Never instant: detection plus the TTL plus clients that cache longer. Lesson 38.28
- file system policy (EFS file system policy, elasticfilesystem:ClientMount, elasticfilesystem:ClientWrite)
- The resource-based policy of an EFS file system, deciding who may mount and write (ClientMount, ClientWrite) and whether TLS is required (aws:SecureTransport). Lesson 38.17
- filter expression (--filter-expression, ScannedCount)
- A condition DynamoDB applies after reading the items. It saves bandwidth, not capacity: you pay for ScannedCount (what was read), not for Count (what matched). Lesson 38.10
- Firewall Manager (AWS Firewall Manager)
- The AWS service that applies WAF rules, Shield Advanced, security groups and Network Firewall policies across every account of an organization. Lesson 38.35
- FSx (Amazon FSx, FSx for Windows File Server, FSx for Lustre)
- AWS's managed file systems beyond EFS: FSx for Windows File Server (SMB and Active Directory), FSx for Lustre (high-performance computing, linked to S3), and NetApp and OpenZFS file systems. Lesson 38.17
- geolocation routing (geolocation records, CountryCode)
- A Route 53 policy that answers by where the user is (continent, country, US state), for compliance and language. It needs a default record (CountryCode "*") or users elsewhere get no answer. Lesson 38.28
- Glacier Deep Archive (Deep Archive, DEEP_ARCHIVE)
- The cheapest storage on AWS, for compliance archives and last copies: a restore takes 12 hours (Standard) to 48 hours (Bulk), and there is a 180-day minimum. Lesson 38.19
- Glacier Flexible Retrieval (S3 Glacier, GLACIER)
- The S3 archive class for backups whose objects cannot be read until a restore makes a temporary copy, in minutes to hours depending on the retrieval tier. A 90-day minimum. Lesson 38.19
- Glacier Instant Retrieval (GLACIER_IR)
- The S3 archive class read in milliseconds, for archives opened a few times a year: a 90-day minimum and a higher retrieval fee. Lesson 38.19
- Global Accelerator (AWS Global Accelerator)
- The other AWS edge service: two static anycast IP addresses that carry TCP and UDP traffic over AWS's network to healthy endpoints in one or more Regions, with fast failover and no caching. Lesson 38.31
- global tables (global table, --replica-updates, last writer wins)
- DynamoDB replicas of a table in several Regions, all writable, replicated asynchronously (usually under a second); concurrent writes to the same item resolve as last writer wins. Multi-Region active-active at a multiplied write cost. Lesson 38.13
- governance mode (s3:BypassGovernanceRetention)
- The Object Lock mode (and AWS Backup Vault Lock mode) that protects against mistakes but can be bypassed by a principal with a special permission (s3:BypassGovernanceRetention) who asks for it explicitly. Lesson 38.22
- GSI (global secondary index, global secondary indexes, GSIs)
- Global secondary index: a second copy of a DynamoDB table's data under other keys, for another access pattern, with its own capacity. Kept in sync asynchronously, so never strongly consistent. Its projection says which attributes are copied (KEYS_ONLY, INCLUDE, ALL). Every table write also costs capacity in each GSI that receives it; up to 20 per table. Lesson 38.10
- health check (Route 53) (Route 53 health check, Route 53 health checks, calculated health check)
- A Route 53 probe of an endpoint from many internet locations (HTTP, HTTPS, TCP, string match), a calculated combination of others, or one that follows a CloudWatch alarm for private endpoints. A record tied to a failing check is left out of answers. Lesson 38.28
- hot key (hot keys, hot partition, hot partitions)
- A partition key that gets a large share of the traffic (today's date, a global counter, a celebrity user), so its partition throttles even when the table has capacity to spare. Lesson 38.13
- HPC (high-performance computing)
- High-performance computing: large numeric jobs (simulations, model training) spread over many machines that read and write huge data sets fast. Lesson 38.17
- IA (Infrequent Access)
- Infrequent Access: the storage tier for data read rarely (EFS Infrequent Access, S3 Standard-IA and One Zone-IA): less per GB stored, a fee per GB read. Lesson 38.17
- IAM database authentication (IAM authentication (RDS), rds_iam, rds-db:connect)
- Logging in to RDS with a 15-minute token signed locally from IAM credentials instead of a password. The user needs the rds_iam role and the caller an IAM allow on rds-db:connect for the instance's resource ID. Lesson 38.3
- instance store (instance store volume, instance storage)
- Disks physically attached to the host an instance runs on: the fastest local storage, and lost when the instance stops. For caches, scratch space and buffers only. Lesson 38.17
- Intelligent-Tiering (S3 Intelligent-Tiering, INTELLIGENT_TIERING)
- The S3 storage class that moves each object between tiers by how often it is read, with no retrieval fees and a small monitoring fee per object: for unknown or changing access patterns. Lesson 38.19
- invalidation (CloudFront) (CloudFront invalidation, invalidations, create-invalidation)
- Removing paths (/* for everything) from every CloudFront edge cache so the next request fetches the new version. The first 1,000 paths a month are free. Lesson 38.31
- IP set (IP sets, geo match)
- A list of addresses and CIDRs a WAF rule can allow or block; a geo match statement does the same by country. Lesson 38.35
- Lambda@Edge
- Full AWS Lambda functions that run at CloudFront's regional edges, for logic too big or slow for CloudFront Functions. Lesson 38.31
- latency routing (latency-based routing, latency records)
- A Route 53 policy that answers with the Region that has the lowest measured latency from the user's resolver: the basis of active-active multi-Region designs. Lesson 38.28
- legal hold (legal holds, put-object-legal-hold)
- An Object Lock flag that blocks deleting an object version with no end date, until someone removes it: for litigation. Lesson 38.22
- load-bearing cache (load-bearing, cold cache)
- A cache the database can no longer run without: a restart (a cold, empty cache) becomes an outage. Size the database for a cold cache, or warm the cache first. Lesson 38.15
- LSI (local secondary index, local secondary indexes)
- Local secondary index: keeps the table's partition key and changes only the sort key. It can only be created with the table, and unlike a GSI it allows strongly consistent reads. Lesson 38.10
- MACsec
- Encryption on the Ethernet link itself, available on dedicated 10 and 100 Gbps Direct Connect ports; the other way to encrypt Direct Connect is a VPN over it. Lesson 38.41
- managed rule group (managed rule groups, AWSManagedRulesSQLiRuleSet, AWSManagedRulesKnownBadInputsRuleSet)
- A set of WAF rules maintained by AWS or a Marketplace seller: SQL injection, known bad inputs (exploits such as Log4Shell), IP reputation. OverrideAction None keeps the group's own blocks; Count turns them all into counts. Lesson 38.35
- manual snapshot (manual snapshots, DB snapshot, DB snapshots)
- An RDS snapshot you take and name yourself, kept until you delete it: the cheap insurance before a migration or an upgrade, and the way to keep a copy beyond 35 days. Lesson 38.8
- master user (master password, rds_superuser, rdsadmin)
- The admin login RDS creates with the database. It can create roles and databases and is a member of rds_superuser, but it is not a real superuser: rdsadmin, used only by AWS, is. So
ALTER SYSTEM, file access and anything that touches the host are closed; server settings go through the parameter group. Lesson 38.1 - maxmemory-policy (volatile-lru, eviction (cache))
- The Valkey / Redis OSS setting for what is thrown out when memory is full. ElastiCache's default, volatile-lru, evicts keys that have a TTL, least recently used first. Watch the Evictions metric. Lesson 38.15
- Memcached
- A simple, multi-threaded in-memory object cache: strings only, no replicas, no failover, no persistence. Scales by adding nodes, with the clients spreading the keys. Lesson 38.15
- MFA Delete
- An S3 versioning setting that requires a one-time code from the root user's MFA device to delete a version or change versioning. Lesson 38.22
- minimum storage duration (minimum billable size, minimum duration)
- What cheaper S3 classes bill at least: 30, 90 or 180 days per object, and for Standard-IA and One Zone-IA 128 KB per object. Moving millions of tiny objects to Standard-IA costs more than leaving them. Lesson 38.19
- mount target (mount targets)
- The NFS entry point of an EFS file system in one Availability Zone: a network interface with a private address and a security group, which must allow TCP 2049 from the clients or mounts hang. The file system's DNS name resolves to the mount target in the client's own zone. Lesson 38.17
- Multi-AZ DB cluster (Multi-AZ DB clusters)
- The RDS deployment with a writer and two readable standbys in three zones, replicated semi-synchronously: availability and some read scaling. The one case where an RDS standby can be read. Lesson 38.3
- Multi-AZ standby (standby (RDS), Multi-AZ instance, --force-failover)
- The hidden copy of an RDS database in another zone, written synchronously, never readable, that takes over by itself when the primary fails: the endpoint's CNAME moves to it and every open session is cut. A failover takes 60-120 s on AWS.
reboot-db-instance --force-failoveris how you test that clients survive one. Lesson 38.3 - multivalue answer routing (multivalue answer, multivalue records)
- A Route 53 policy that returns up to eight healthy values: simple spreading by the clients, not a load balancer. Lesson 38.28
- networkaddress.cache.ttl (Java DNS cache)
- The Java setting for how long a looked-up address is remembered: forever with a security manager, 30 s otherwise. Database clients should keep it to a few seconds, or they keep connecting to the old primary after a failover. Lesson 38.3
- non-transitive (not transitive, transitive routing, edge-to-edge routing)
- A peering carries traffic only between its own two VPCs: A-B and B-C do not make A-C, and a peer cannot use your internet gateway, NAT gateway, VPN or Direct Connect. Lesson 38.37
- OAI (origin access identity)
- Origin access identity: the older way CloudFront read private S3 buckets. OAC replaced it: OAC signs with SigV4, supports SSE-KMS buckets and every Region. Lesson 38.31
- on-demand (DynamoDB) (on-demand capacity, on-demand mode, PAY_PER_REQUEST)
- The DynamoDB mode with no capacity to plan: the table scales by itself (up to twice the previous peak at once) and bills per request, more per request than well-used provisioned capacity. Lesson 38.13
- One Zone (EFS) (EFS One Zone)
- An EFS file system that keeps its data in one Availability Zone for about half the price: for dev and data you can rebuild, not data that must survive a zone outage. Lesson 38.17
- One Zone-IA (ONEZONE_IA)
- Standard-IA kept in a single Availability Zone, cheaper and lost if the zone is lost: for copies you can re-create, such as thumbnails. Lesson 38.19
- origin (CloudFront) (CloudFront origin, Origin Shield)
- Where CloudFront fetches what it has not cached: an S3 bucket, a load balancer or any HTTP server. Origin Shield adds a regional cache layer in front of the origin to reduce its load. Lesson 38.31
- origin group (origin groups, origin failover)
- A CloudFront primary and secondary origin: on 5xx errors or timeouts it retries the secondary, for example a bucket replicated to another Region. Lesson 38.31
- overlapping CIDRs (overlapping CIDR, CIDR overlap)
- Two networks using the same addresses. They cannot be peered or routed to each other; the only fix afterwards is renumbering, which is why companies keep an address plan. Lesson 38.37
- partition (DynamoDB) (DynamoDB partition)
- One slice of a DynamoDB table's storage and throughput, chosen by hashing the partition key. Each handles at most 1,000 WCU and 3,000 RCU per second, whatever the table has. Lesson 38.13
- partition key (partition keys, hash key)
- The part of a DynamoDB key that is hashed to choose where an item lives. It should have many distinct values that spread the traffic; every request names it. Lesson 38.10
- PendingModifiedValues (--apply-immediately, maintenance window (RDS), maintenance window)
- The changes an RDS instance has accepted but not applied yet. Without --apply-immediately they wait for the weekly maintenance window, where RDS applies them with a short outage. On a Multi-AZ instance RDS changes the standby first and then fails over, so the outage is the failover time. Lesson 38.1
- Performance Insights (DBLoad, average active sessions)
- The RDS feature that shows database load as DBLoad, the average number of active sessions, broken down by wait event and by SQL statement. DBLoad above the vCPU count means sessions are waiting. Lesson 38.1
- point-in-time recovery (point-in-time recovery (DynamoDB), on-demand backup (DynamoDB), create-backup)
- DynamoDB's continuous backups: restore to any second of the last 1-35 days up to about 5 minutes ago, into a new table that does not copy auto scaling, IAM policies, alarms, TTL or stream settings. On-demand backups (
create-backup) are kept until deleted, for copies before a migration and long retention. Lesson 38.13 - point-in-time restore (point-in-time restore (RDS), PITR (RDS), PITR)
- Restoring an RDS database to any second inside its restore window, for example the second before a DELETE without a WHERE. Like every RDS restore it creates a new instance with a new endpoint. Lesson 38.8
- price class (price classes, PriceClass_100)
- A CloudFront setting that limits which edge locations serve a distribution, and so what it costs. Lesson 38.31
- primary endpoint (primary endpoint (ElastiCache))
- The DNS name of an ElastiCache replication group for writes. It follows the primary through a failover, so clients that use it and reconnect on errors carry on. Lesson 38.15
- private hosted zone (private hosted zones)
- A Route 53 zone answered only inside the VPCs associated with it, which wins over a public zone of the same name there. The VPCs need DNS support and DNS hostnames on. Lesson 38.28
- promote-read-replica (promote a replica, replica promotion (RDS))
- Turning an RDS read replica into a standalone, writable instance. It is manual and one-way, and the application must be pointed at the new endpoint, which is why a replica is no substitute for Multi-AZ. Lesson 38.3
- propagation (TGW) (TGW propagation, propagate (TGW))
- An attachment writing its CIDRs into a TGW route table, so traffic for them is sent to it. An attachment can propagate into several tables. Lesson 38.37
- provisioned capacity (provisioned mode, burst capacity, reserved capacity)
- The DynamoDB mode where you set RCU and WCU per second, optionally moved by auto scaling towards a target use (70% is typical) within minutes. Unused capacity of the last 5 minutes covers short bursts. Reserved capacity makes steady, predictable tables the cheapest per unit. Lesson 38.13
- Query (DynamoDB) (aws dynamodb query, key condition, key condition expression)
- The DynamoDB read of one partition: the key condition names the partition key with = and may narrow the sort key (<, >, BETWEEN, begins_with). Results come in sort-key order. Cheap and fast. Lesson 38.10
- ransomware
- Malware or an attacker that encrypts or deletes your data and demands payment. Backups are its first target, which is why the last copies live in another account and under a compliance-mode lock. Lesson 38.22
- rate-based rule (rate-based rules, rate-based statement, managed keys (WAF))
- A WAF rule that blocks a key (an IP, a header, a combination) sending more than N requests in a time window, until it slows down. Its managed keys are the addresses it is blocking right now. Lesson 38.35
- RCU (read capacity unit, read capacity units, RCUs)
- Read capacity unit: one strongly consistent read per second of up to 4 KB, or two eventually consistent ones. Lesson 38.13
- RDS endpoint (endpoint (RDS), DB endpoint, PubliclyAccessible)
- The DNS name clients use to reach an RDS database: a CNAME with a 5-second TTL that points at whichever host is the primary right now. Connect by this name, never by the address behind it. With
PubliclyAccessible: falseit resolves to a private address only, reachable from inside the VPC or a peered network. Lesson 38.1 - RDS events (RDS event, RDS event subscription, event subscription (RDS))
- The record of what happened to an RDS resource (failover started, failover completed, backup done), readable with describe-events and delivered to EventBridge, the AWS event bus, or to an event subscription for alarms. Lesson 38.3
- RDS Proxy (Amazon RDS Proxy, proxy pinning, pinning (RDS Proxy))
- A managed connection pool in front of an RDS database, with its own endpoint and security group. It multiplexes many client connections onto few database ones and keeps them open through a failover, so clients see a pause instead of an error. It reads the password from Secrets Manager. Session state (
SET, temporary tables, prepared statements) pins a client to one database connection and reduces the multiplexing. Lesson 38.3 - rds.force_ssl (force_ssl, sslmode=verify-full, RDS CA bundle)
- The RDS for PostgreSQL parameter that refuses connections without TLS. It defaults to 1 since PostgreSQL 15, so an old client configured without TLS breaks after an upgrade. Full protection is
sslmode=verify-fullwith the RDS CA bundle, which also proves the server is the one you meant. Lesson 38.1 - read replica (read replicas, RDS read replica, cross-Region read replica)
- A read-only copy of an RDS database, fed asynchronously, with its own endpoint, for reports and other read load. It may be behind by ReplicaLag seconds and never takes over by itself. It can live in another Region (a cheap DR copy) and needs automated backups on the source. Lesson 38.3
- reader endpoint (reader endpoint (Aurora))
- The Aurora DNS name that spreads read-only connections over the readers. Lesson 38.3
- reader endpoint (reader endpoint (ElastiCache))
- The ElastiCache DNS name that spreads reads over the replicas. A read just after a write may miss it, because replication is asynchronous. Lesson 38.15
- recovery point (recovery points, recovery point (AWS Backup))
- One backup in an AWS Backup vault: for RDS a DB snapshot, for DynamoDB a table backup, for EBS a snapshot. The time between two of them is the RPO the plan really gives. Lesson 38.24
- replication group (replication groups, test-failover, automatic failover)
- An ElastiCache Valkey or Redis OSS cache: per shard one primary (reads and writes) and up to five read-only replicas. With Multi-AZ and automatic failover a replica in another zone is promoted when the primary fails. A write to a replica gets
READONLY You can't write against a read only replica.Replication is asynchronous, so the last writes before a failover can be lost. Lesson 38.15 - replication status (ReplicationStatus, OperationsFailedReplication, ReplicationLatency)
- Per object: PENDING, COMPLETED or FAILED on the source and REPLICA on the copy. A FAILED object (role, permissions, KMS key) is not retried by itself, so alarm on OperationsFailedReplication. Lesson 38.22
- Replication Time Control (S3 Replication Time Control, RTC)
- The paid S3 replication option that promises 99.99% of objects copied within 15 minutes, with metrics to prove it. Lesson 38.22
- reserved word (DynamoDB) (reserved words (DynamoDB), --expression-attribute-names, expression attribute name)
- One of more than 500 names (status, name, date, count, total...) DynamoDB refuses in an expression. Use a placeholder such as #s and map it with --expression-attribute-names. Lesson 38.10
- restore job (restore jobs, start-restore-job, restore metadata)
- An AWS Backup restore of a recovery point, which always creates a new resource; the restore metadata lists what can be set for it. Lesson 38.24
- restore testing (restore test, restore tests, restore testing plan)
- Restoring backups on purpose to prove they work: the data is there, the time it took (your measured RTO), the steps written down. AWS Backup restore testing plans do it on a schedule and clean up. Lesson 38.24
- restore window (EarliestRestorableTime, LatestRestorableTime)
- The span a point-in-time restore can reach: from the oldest kept backup to LatestRestorableTime, which trails the present by about five minutes because RDS uploads the logs every five minutes. Lesson 38.8
- restore-object (aws s3api restore-object, InvalidObjectState, Restore header)
- The S3 call that asks for a temporary readable copy of a Glacier object for some days. Until then reads fail with InvalidObjectState; the Restore header shows ongoing-request and, afterwards, the copy's expiry date. Lesson 38.19
- retrieval tier (retrieval tiers, Expedited retrieval, Bulk retrieval)
- How fast a Glacier restore runs: Expedited (1-5 minutes, Flexible Retrieval only, expensive), Standard (3-5 hours, 12 for Deep Archive) or Bulk (5-12 hours, 48 for Deep Archive, cheapest). Lesson 38.19
- Route 53 (Amazon Route 53, Route53)
- AWS's DNS service: hosted zones, records, health checks and routing policies that can answer differently per query. Lesson 38.28
- Route 53 Resolver (Resolver endpoint, inbound endpoint, outbound endpoint)
- The VPC's DNS resolver. Inbound endpoints let on-premises servers resolve private zones; outbound endpoints plus forwarding rules send names like corp.example to on-premises servers. DNS Firewall blocks lookups of bad domains. Lesson 38.28
- route propagation (VGW) (route propagation, EnableVgwRoutePropagation, enable-vgw-route-propagation)
- A VPC route table option that lets the virtual private gateway add the VPN's routes (static or learned over BGP) by itself. Without it, "the VPN is up but nothing goes through". Lesson 38.41
- routing policy (routing policies, geoproximity routing, geoproximity)
- How Route 53 picks the answer from a set of records with the same name and type: weighted, latency, failover, geolocation, geoproximity (by distance, with a bias), multivalue answer or IP-based (by the resolver's CIDR). Lesson 38.28
- S3 Batch Replication (Batch Replication)
- A job that replicates objects that already existed before the replication rule, or that failed, which normal replication never copies. Lesson 38.22
- S3 Replication (S3 replication, replication configuration (S3), put-bucket-replication)
- S3 copying new object versions to another bucket, in another Region (CRR) or the same one (SRR), possibly another account. It needs versioning on both buckets and an IAM role S3 assumes, and keeps the version ID. Rules have a filter, a priority, a destination and optionally another storage class. Overwrites are replicated; version deletes never are. Lesson 38.22
- S3 Transfer Acceleration (Transfer Acceleration, s3-accelerate)
- Uploads that enter AWS at the nearest CloudFront edge location and cross AWS's network to the bucket, through the <bucket>.s3-accelerate.amazonaws.com endpoint: faster from far away. Lesson 38.19
- SAA-C03 (Solutions Architect Associate, AWS Certified Solutions Architect - Associate)
- The code of the AWS Certified Solutions Architect - Associate exam: designing on AWS, choosing services for resilience, performance, security and cost. Lesson 38.1
- Scan (DynamoDB) (aws dynamodb scan, table scan)
- The DynamoDB read of the whole table, item by item. Slow and expensive on a large table, and it competes with production traffic for capacity: model the access pattern into the keys instead. Lesson 38.10
- Schema Conversion Tool (SCT, AWS SCT)
- The AWS tool that converts a database schema and code from one engine to another (Oracle to PostgreSQL, for example), used with DMS when the engine changes. Lesson 38.43
- scope (WAF) (CLOUDFRONT, REGIONAL, WAF scope)
- Where a web ACL can be used: CLOUDFRONT web ACLs are created in us-east-1, REGIONAL ones (ALB, API Gateway) in the resource's own Region. Lesson 38.35
- SetIdentifier (set identifier)
- The name that tells apart the records of one Route 53 routing set (same name and type). Every record in a set needs one. Lesson 38.28
- The split of duties between AWS and you. For a managed database AWS owns the hosts, the operating system, patching, the backup machinery and failover; you own the schema, queries, users, parameters, network path, alarms and restore tests. Lesson 38.1
- Shield Advanced (AWS Shield Advanced, Shield Response Team, cost protection)
- The paid DDoS protection for load balancers, CloudFront, Global Accelerator, Route 53 and Elastic IPs: the Shield Response Team, WAF included for protected resources, and credits for scaling caused by an attack. Lesson 38.35
- signed URL (CloudFront) (CloudFront signed URL, signed URLs, signed cookies)
- CloudFront URLs or cookies signed with a key from a key group, for paid or private content; geo restriction blocks or allows whole countries. Lesson 38.31
- SigV4 (Signature Version 4)
- The way AWS requests are signed with your credentials, so AWS knows who sent them and that nobody changed them. An RDS auth token is a SigV4-signed URL; OAC signs CloudFront's requests to S3 the same way. Lesson 38.3
- single-table design
- Keeping several item types (a customer profile, its orders) in one DynamoDB table under generic key names, so related items share a partition and one query returns them together. The keys are usually called
PKandSK:CUSTOMER#101+PROFILE,CUSTOMER#101+ORDER#<date>#<id>. Lesson 38.10 - Site-to-Site VPN (AWS Site-to-Site VPN, VPN connection (AWS), Site-to-Site VPN connection)
- Encrypted IPsec tunnels over the internet between your router and a virtual private gateway or a Transit Gateway. Ready in minutes, as good as the internet underneath. Lesson 38.41
- SOA-C03 (SysOps Administrator Associate, AWS Certified SysOps Administrator - Associate)
- The code of the AWS Certified SysOps Administrator - Associate exam: running workloads on AWS, monitoring, backups, networking and troubleshooting. Lesson 38.1
- sort key (sort keys, range key)
- The optional second part of a DynamoDB key, which orders items inside a partition and allows range conditions such as a date range or begins_with. Lesson 38.10
- sparse index (sparse indexes)
- A DynamoDB index that only holds the items that have its key attribute; the others are simply left out. A cheap way to index only the items you care about. Lesson 38.10
- SQL injection (SQLi)
- An attack that hides SQL in user input (a query string, a form field) so that a careless application runs it against its database. Lesson 38.35
- stale data (stale, cache invalidation, invalidate (cache))
- A cached copy that no longer matches the source, for example after a write that did not invalidate (delete or update) the cached entry. Lesson 38.15
- Standard-IA (S3 Standard-IA, Standard - Infrequent Access, retrieval fee)
- The S3 storage class for data read rarely but needed in milliseconds: cheaper per GB-month, with a fee per GB read, a 30-day minimum and a 128 KB minimum billable size. Lesson 38.19
- static parameter (static parameters, dynamic parameter, dynamic parameters)
- A database setting that only takes effect after a reboot (shared_buffers, max_connections); a dynamic one applies at once. Until the reboot RDS shows the change as pending-reboot. Lesson 38.1
- storage autoscaling (--max-allocated-storage, storage-full, FreeStorageSpace)
- RDS growing an instance's storage by itself when free space stays under 10% for five minutes, at most every six hours, up to a ceiling, never down. A full instance goes to storage-full and stops writing, so alarm on FreeStorageSpace anyway. Lesson 38.1
- Storage Gateway (AWS Storage Gateway)
- A box (a VM or appliance) in your data centre that gives on-premises servers file, volume or tape interfaces backed by S3. Lesson 38.19
- system of record
- The store that holds the true copy of some data. A cache never is: sessions and carts kept only in a cache need a fallback or a durable store. Lesson 38.15
- test-dns-answer (aws route53 test-dns-answer)
- The Route 53 call that shows what Route 53 itself would answer right now, with no resolver cache in the way: the first check when DNS failover "did not work". Lesson 38.28
- TGW attachment (transit gateway attachment, TGW attachments, VPC attachment)
- The link of one VPC, VPN or Direct Connect gateway to a Transit Gateway. Lesson 38.37
- TGW route table (transit gateway route table, TGW route tables)
- A Transit Gateway's routing table. Each attachment is associated with one; separate tables give segmentation (prod and dev cannot reach each other, shared services see both). Lesson 38.37
- throttling (DynamoDB) (ProvisionedThroughputExceededException, ThrottledRequests, WriteThrottleEvents)
- DynamoDB refusing a request because the table, an index or one partition used up its capacity for that second. The SDKs retry; when they give up the caller gets ProvisionedThroughputExceededException. Alarm on ThrottledRequests and Read/WriteThrottleEvents, and compare ConsumedWriteCapacityUnits against what is provisioned. Lesson 38.13
- thundering herd (cache stampede)
- A hot cache entry expires and a thousand requests miss at once and all hit the database. Staggered TTLs, or one request refreshing while the others get the old value, prevent it. Lesson 38.15
- transact-write-items (TransactWriteItems, DynamoDB transaction, TransactionCanceledException)
- The DynamoDB call that applies up to 100 writes across tables all or nothing, cancelled with a reason per item, at twice the capacity cost. Lesson 38.10
- Transit Gateway (TGW, transit gateway, Transit Gateways)
- A regional router that many VPCs, VPNs and Direct Connect attach to, and that routes between them transitively. Billed per attachment-hour and per GB. Lesson 38.37
- TTL (DynamoDB) (DynamoDB TTL, time to live (DynamoDB), epoch second)
- DynamoDB deleting items whose TTL attribute holds an epoch second in the past, for free. It is not exact: deletion can take days, so queries should filter on the attribute when it matters. Lesson 38.13
- TunnelState (TunnelIpAddress)
- The CloudWatch metric of a VPN tunnel: 1 up, 0 down. Per connection it averages the tunnels, so 0.5 means one of two is down: alarm below 1. Lesson 38.41
- Valkey (Redis OSS, Redis)
- The open-source fork of Redis that AWS now recommends: an in-memory key-value store with strings, hashes, lists, sets, sorted sets and streams, replicas and failover, on port 6379. Lesson 38.15
- Vault Lock (AWS Backup Vault Lock, --changeable-for-days)
- An AWS Backup lock that enforces a minimum and maximum retention on a vault's recovery points. Governance mode can be removed with the right IAM permissions; with --changeable-for-days it becomes compliance mode, which nobody can remove. Lesson 38.24
- versioned file names (versioned object names, cache busting)
- Putting a content hash in each asset's name (app.3f9a2c.js) so a new build is a new object: cached for a year, never invalidated; only index.html, which points at them, gets a short TTL. Lesson 38.31
- viewer protocol policy (redirect-to-https)
- What a CloudFront behavior does with plain HTTP from users: allow it, redirect it to HTTPS (redirect-to-https), or refuse it. Lesson 38.31
- virtual interface (virtual interfaces, VIF, private VIF)
- A logical connection on a Direct Connect link: private (to a VGW or a Direct Connect gateway), transit (to a Transit Gateway) or public (to public AWS endpoints such as S3). Lesson 38.41
- virtual private gateway (VGW, VPN gateway)
- The AWS end of a VPN or Direct Connect for one VPC. A Transit Gateway takes its place when many VPCs share the connection. Lesson 38.41
- VPC endpoint service (endpoint service, AWS PrivateLink, PrivateLink (own service))
- PrivateLink for your own service: put it behind an NLB, create an endpoint service, allow the consumer accounts. Consumers reach only that service through an interface endpoint, even with overlapping CIDRs and no routes between the networks. Lesson 38.41
- VPC peering (peering connection, VPC peering connection, pending-acceptance)
- A private link between exactly two VPCs, across accounts and Regions. The other side must accept it, the CIDRs must not overlap, and both VPCs need routes to each other through it. Free per hour (data across zones and Regions is billed per GB). Enable DNS resolution on the peering to resolve the peer's private names. Lesson 38.37
- VPN tunnel (VPN tunnels, two tunnels)
- One of the two IPsec tunnels every AWS VPN connection has, to two AWS endpoints. AWS replaces them one at a time during maintenance, so a router that uses only one has announced outages. Lesson 38.41
- WAF logs (aws-waf-logs-, sampled requests, get-sampled-requests)
- The record of every request a web ACL saw and which rule decided it, sent to CloudWatch Logs (group names must start with aws-waf-logs-), S3 or Firehose. Sampled requests show a recent sample per rule in the console. Lesson 38.35
- WCU (write capacity unit, write capacity units, WCUs)
- Write capacity unit: one write per second of an item up to 1 KB, rounded up per item, so a 1.1 KB item costs 2. Transactions cost double. Lesson 38.13
- web ACL (web ACLs, WebACL, default action (WAF))
- The AWS WAF object attached to a resource: rules evaluated by priority, then a default action for requests no rule decided. To CloudFront it is attached through the distribution's own config. Lesson 38.35
- weighted routing (weighted records, weighted record)
- A Route 53 policy that answers each record in proportion to its weight: 90/10 for a canary. Weight 0 takes an endpoint out without deleting it. Lesson 38.28
- write sharding (key sharding)
- Spreading one busy logical key over several by adding a suffix (2026-10-16#0 ... #9), so writes land in several partitions; readers query every shard. Lesson 38.13
- write-through (write-through cache)
- The pattern where every write to the database also updates the cache, so reads never miss after a change, at the price of caching data nobody may read. Lesson 38.15
- x-cache (Hit from cloudfront, Miss from cloudfront, RefreshHit from cloudfront)
- The CloudFront response header that says what happened: Miss (fetched from the origin), Hit (served from the edge), RefreshHit (expired, revalidated, served) or Error.
ageis how long this copy has been cached;x-amz-cf-idis the request ID to give AWS Support. Lesson 38.31 - XSS (cross-site scripting)
- Cross-site scripting: an attack that gets a site to send the attacker's JavaScript to other users' browsers. Lesson 38.35