Skip to Content
Cloud Engineering Academy

Microsoft / Azure → AWS mapping

The full translation table from the Microsoft and Azure services a .NET banking developer knows to their AWS equivalents — with what is different and what to watch out for.

Assumes: AWS service names and availability as of 2026-11-17. Verified against official sources on 2026-11-17.

You already know how to build and run banking systems. This page translates that knowledge. For each Microsoft or Azure service you get the closest AWS service and — more importantly — what is different. "Equivalent" almost never means "identical": the concept maps, but the defaults, limits, pricing and security model usually don't.

Analogy: moving from Azure to AWS is like moving to a bank in another country. There are still current accounts, cards and transfers, but the forms have different names, the opening hours differ, and some things you took for granted (like a free service) cost money here.

Availability changes you must know (checked 2026-11-17)

  • AWS App Runner closed to new customers on 2026-04-30. AWS now recommends Amazon ECS Express Mode for the same "give me a container, get a URL" experience.
  • AWS CodeCommit returned to general availability for new customers on 2025-11-24, after being closed in 2024.
  • AWS Cloud9 is closed to new customers (since mid-2024). Use your local IDE with the AWS Toolkit, or AWS CloudShell.
  • Porting Assistant for .NET, App2Container, Toolkit for .NET Refactoring and Microservice Extractor closed to new customers on 2025-11-07 → use AWS Transform (its .NET capability).
  • AWS Migration Hub closed to new customers on 2025-11-07 → AWS Transform.
  • Amazon Fraud Detector closed to new customers on 2025-11-07.
  • Amazon Q Developer: new sign-ups blocked from 2026-05-15; IDE plugins reach end of support 2027-04-30. The successor for coding help is Kiro.
  • AWS X-Ray SDKs and daemon are in maintenance mode (since 2026-02-25, end of support 2027-02-25). Instrument new .NET code with OpenTelemetry (ADOT); the X-Ray service itself still receives traces.

Details and official links: Resources → service availability.

The big picture: an on-premises banking app and its AWS shape

A typical in-house banking application: an ASP.NET site on IIS, a Windows Service that processes payments from an MSMQ queue, SQL Server with an Always On availability group, Active Directory for staff log-in, and secrets in config files or a vault.

The same application, rehosted and lightly modernised on AWS:

What changed, in one sentence each:

  • IIS web farm → ALB + ECS Fargate. The load balancer and the servers are managed; you ship a container image instead of a web-deploy package.
  • MSMQ → SQS. No queue server to patch; messages are stored across several data centres; poison messages go to a dead-letter queue instead of a "retry" sub-queue.
  • Windows Service → Worker Service on Fargate (or Lambda). The BackgroundService code you already write runs almost unchanged in a container.
  • SQL Server AG → RDS for SQL Server Multi-AZ. AWS runs the standby and failover; you lose sysadmin and OS access.
  • AD → federated identity. Staff keep signing in with the corporate directory; AWS trusts it through SAML/OIDC.
  • Config-file secrets → Secrets Manager + KMS. The app fetches secrets at runtime using its IAM role — no password in appsettings.json.

The full walkthrough is in Module 13 — Microsoft-to-AWS migration.

Mindset differences

These five ideas trip up almost every Azure engineer in their first month.

IdeaIn Azure / MicrosoftIn AWSWhy it matters
Accounts vs subscriptionsA tenant (Entra ID) holds many subscriptions; resource groups organise resourcesThe AWS account is the unit of isolation, billing and security. Accounts are grouped into OUs under AWS Organizations. There is no "resource group" boundary for permissions — use separate accounts and tagsBanks run dozens to hundreds of accounts (one per workload per environment). "Put prod in its own subscription" becomes "put prod in its own account".
IAM vs RBACAzure RBAC assigns built-in roles at a scope (management group → subscription → RG → resource)IAM policies are JSON documents listing allowed/denied actions on resources, attached to roles. Many resources (S3 buckets, KMS keys, SQS queues) also have their own resource policy. Every API call is checked, and an explicit deny always winsYou will write and read policy JSON. "Contributor on the resource group" has no direct equivalent — you grant exactly the actions needed.
Regions and AZsRegions, with availability zones in mostEach Region has several Availability Zones (separate data centres; three or more in most Regions). Most services are Regional; IAM, Organizations, Route 53 and CloudFront are globalYou design for "survive an AZ failure" by default. Data stays in the Region you choose — important for UK GDPR / residency.
Everything is an APIARM is also an API, but many teams click in the portalEvery console click is a signed API call (SigV4), recorded in CloudTrail. The CLI, SDK, CDK and Terraform all call the same APIsAutomation and audit are first-class: "who changed the security group?" is answerable from CloudTrail.
Pay-per-use, including the networkSimilar, but many PaaS tiers are flat-pricedYou pay per hour, request, GB stored and GB moved. NAT Gateways, load balancers, idle databases, cross-AZ traffic and public IPv4 addresses all cost money even when idleBudgets, alarms and teardown are part of engineering, not finance. See Module 15 — FinOps.

Mapping tables

Compute and hosting

Microsoft / AzureAWS equivalentWhat's different / watch out for
IIS on Windows Server (VM or physical)Amazon EC2 Windows instances (licence included), optionally behind an ALBClosest lift-and-shift. You still patch the OS (use Systems Manager Patch Manager). Windows licence cost is in the hourly price.
IIS website running modern .NETECS on Fargate + Application Load BalancerContainers on Linux are cheaper and start faster than Windows. Kestrel serves traffic; the ALB does TLS and routing that IIS used to do.
Azure App Service (Web Apps)Elastic Beanstalk (.NET on Linux or Windows/IIS), Amazon ECS Express Mode, or ECS on FargateBeanstalk is the nearest "upload code, get a site" PaaS and is still current. App Runner is closed to new customers — AWS points new users to ECS Express Mode.
App Service deployment slotsECS blue/green deployments, Lambda aliases with weighted traffic, Beanstalk "swap environment URLs"No single "slot" concept; each hosting service has its own blue/green mechanism.
Windows ServiceWorker Service (BackgroundService) in an ECS service on Fargate; or Lambda for event-driven work; or EC2Long-running loops belong in containers. Lambda has a hard 15-minute limit per invocation.
Windows Task Scheduler / SQL Agent jobs (non-SQL)EventBridge Scheduler triggering Lambda, an ECS task or Step FunctionsSchedules are cloud resources with IAM roles, retries and DLQs — not entries on one server.
Azure Virtual MachinesAmazon EC2Instance families (m, c, r, t…) replace VM sizes. Burstable "t" instances throttle CPU when credits run out.
VM Scale SetsEC2 Auto Scaling groupsScaling policies (target tracking, step, scheduled) attach to the group.
Azure BatchAWS BatchRuns on EC2 or Fargate; jobs defined as container images.
Azure Virtual DesktopAmazon WorkSpacesDifferent licensing model; check Microsoft licence mobility rules.
Azure Static Web AppsAWS Amplify Hosting or S3 + CloudFrontS3 + CloudFront with Origin Access Control is the common enterprise pattern.

Serverless

Microsoft / AzureAWS equivalentWhat's different / watch out for
Azure FunctionsAWS Lambda (managed dotnet10 runtime)One handler per function (or ASP.NET Core hosted in Lambda). Max 15 min, up to 10 GB memory. Cold starts: use SnapStart (.NET 8+) or Native AOT. Triggers are "event source mappings" and service integrations, not binding attributes — the Lambda Annotations library gives you a similar attribute style.
Functions bindings (input/output)SDK calls inside the handler; EventBridge Pipes for source → target without codeNo declarative output bindings; you write the SDK call (or use Pipes/Step Functions service integrations).
Durable FunctionsAWS Step Functions (Standard or Express workflows)Workflows are defined in JSON/YAML (Amazon States Language) or CDK, not C# orchestrator code. Standard workflows can run up to a year with exactly-once step execution.
Logic AppsStep Functions, EventBridge Pipes, Amazon AppFlow (SaaS connectors)Fewer drag-and-drop SaaS connectors than Logic Apps; more code-first.
Azure Functions Premium (pre-warmed)Lambda provisioned concurrency or SnapStartProvisioned concurrency is billed while idle.

Containers

Microsoft / AzureAWS equivalentWhat's different / watch out for
Azure Container AppsECS Express Mode (ECS + Fargate + ALB in one step) or ECS on FargateECS Express Mode creates normal ECS/ALB resources in your account that you can later manage directly. No extra charge beyond those resources.
Azure Container InstancesECS RunTask on FargateFargate tasks need a VPC, subnets and security groups.
Azure Kubernetes Service (AKS)Amazon EKSChoose EKS only if your organisation already runs Kubernetes; ECS is simpler for most .NET teams. EKS control plane is billed per hour.
Azure Container RegistryAmazon ECRTurn on image scanning and lifecycle policies (delete untagged images) from day one.
Dapr / service discovery in Container AppsECS Service Connect or AWS Cloud MapService Connect gives ECS services short names and traffic metrics.

Databases

Microsoft / AzureAWS equivalentWhat's different / watch out for
SQL Server (on-premises)Amazon RDS for SQL Server (managed) or SQL Server on EC2 (full control)RDS: no sysadmin, no OS access, some features limited (check the RDS SQL Server feature list). Licence-included pricing for Standard/Enterprise is significant — model it early. EC2 allows bring-your-own-licence under Microsoft's rules (licensing review needed).
Azure SQL DatabaseRDS for SQL Server, or Aurora PostgreSQL (with Babelfish to accept T-SQL)No serverless "vCore auto-pause" SQL Server option. Moving to PostgreSQL removes SQL Server licence cost but needs testing; Babelfish covers much, not all, of T-SQL.
Azure SQL Managed InstanceRDS for SQL Server or RDS Custom for SQL Server (OS access for special needs)RDS Custom trades some automation for control.
SQL Server Always On AGRDS Multi-AZ deploymentAWS manages the standby and automatic failover; the standby is not readable unless you use read replicas.
Azure Database for PostgreSQL / MySQLRDS for PostgreSQL/MySQL or Amazon AuroraAurora stores six copies across three AZs and fails over faster; costs more per hour, often less for I/O-heavy loads.
Cosmos DBAmazon DynamoDB (key-value/document); DocumentDB (MongoDB API); Keyspaces (Cassandra API); Neptune (graph)DynamoDB has no SQL joins and you design tables around access patterns. Capacity is on-demand or provisioned. Global tables give multi-Region writes.
Azure Cache for RedisAmazon ElastiCache (Valkey, Redis OSS or Memcached); Amazon MemoryDB for a durable Redis-compatible storeElastiCache is a cache — never the system of record for balances.
Synapse dedicated SQL poolAmazon Redshift (provisioned or Serverless)Columnar warehouse; distribution and sort keys replace distribution/index choices.
SSIS packagesAWS Glue jobs, Step Functions; or keep SSIS on SQL Server on EC2 / RDS (SSIS is supported on RDS for SQL Server)Rewriting SSIS to Glue is a refactor, not a lift-and-shift.
SSRS / SSASQuick Sight for dashboards; SSRS and SSAS are supported on RDS for SQL ServerCheck edition and feature limits before committing.
Azure Database Migration Service / Data Migration AssistantAWS DMS (data, CDC) + DMS Schema Conversion (managed) or AWS SCT (desktop)DMS Schema Conversion supports fewer sources than SCT; SCT is still available for cases it does not cover.

Storage

Microsoft / AzureAWS equivalentWhat's different / watch out for
Azure Blob StorageAmazon S3Bucket names are globally unique. Block Public Access is on by default — keep it on. Storage classes (Standard, Intelligent-Tiering, Glacier…) replace hot/cool/archive tiers.
Immutable blob storage (WORM)S3 Object Lock (governance or compliance mode)Compliance mode cannot be shortened — not even by the root user. Get records-retention sign-off first.
Data Lake Storage Gen2S3 + Glue Data Catalog + Lake FormationS3 has no real directories; "folders" are key prefixes.
Azure Files / Windows file sharesAmazon FSx for Windows File Server (SMB, AD-joined); Amazon EFS (NFS, Linux)FSx needs a directory (AWS Managed Microsoft AD or your own AD).
Managed DisksAmazon EBS volumesEBS volumes live in one AZ; snapshots go to S3 behind the scenes.
Azure BackupAWS BackupUse Backup Vault Lock and cross-account copies so a compromised account cannot delete backups.
Azure Site RecoveryAWS Elastic Disaster RecoveryBlock-level replication of servers into a staging area in AWS.
SFTP on a Windows server / Azure Blob SFTPAWS Transfer Family (SFTP, FTPS, AS2)Files land directly in S3 or EFS; identity via service-managed users, AD or a custom IdP.

Messaging and events

Microsoft / AzureAWS equivalentWhat's different / watch out for
MSMQAmazon SQSNo transactions spanning the queue and your database (no DTC). Use the outbox pattern and idempotent consumers. Messages up to 1 MiB (since 2025); larger payloads go to S3.
Service Bus queuesSQS standard or SQS FIFOService Bus "sessions" ≈ SQS FIFO message groups. Peek-lock ≈ visibility timeout. Dead-lettering is configured with a redrive policy. No scheduled messages beyond 15-minute delay — use EventBridge Scheduler.
Service Bus topics + subscriptionsSNS topic → SQS queues (fan-out), or EventBridge rulesSNS subscription filter policies ≈ subscription rules. SNS does not store messages — put an SQS queue behind each subscriber for durability.
Event GridAmazon EventBridgeEventBridge receives events from most AWS services automatically; content-based rules, archive and replay, schema registry.
Event HubsKinesis Data Streams, or Amazon MSK (Kafka)Event Hubs partitions ≈ Kinesis shards. MSK is the choice when you want the Kafka API (Event Hubs' Kafka endpoint).
Event Hubs CaptureAmazon Data FirehoseFirehose batches stream data into S3/Redshift/OpenSearch, optionally converting to Parquet.
IBM MQ / RabbitMQ / ActiveMQ / BizTalk adaptersAmazon MQ (ActiveMQ or RabbitMQ)Lift-and-shift for JMS/AMQP apps and core-banking adapters. It runs broker instances — not serverless.
Azure Storage QueuesSQS standardSimilar simplicity; SQS adds DLQs and long polling.
Azure SignalR ServiceAPI Gateway WebSocket APIs or AWS AppSyncDifferent programming model; there is no drop-in SignalR backplane service.

Identity

Microsoft / AzureAWS equivalentWhat's different / watch out for
Entra ID (workforce sign-in to Azure)IAM Identity Center, federated to Entra ID (SAML + SCIM)Staff sign in with their Entra account and get permission sets in each AWS account. No long-lived IAM users for humans.
Entra External ID / Azure AD B2CAmazon Cognito user poolsCustomer sign-up/sign-in, MFA, OAuth 2.0/OIDC tokens. Check FAPI and strong-customer-authentication needs separately.
On-premises Active DirectoryAWS Managed Microsoft AD, or AD Connector (proxy to your AD)Needed for Windows domain joins, FSx for Windows, and Kerberos to RDS SQL Server.
Azure RBAC role assignmentsIAM policies attached to roles / permission setsDefault is deny; you list actions. Explicit deny anywhere wins.
Managed IdentityIAM roles for EC2 (instance profile), ECS (task role), Lambda (execution role)The SDK picks credentials up automatically via the default credential chain — no secret in code.
Service principal + client secretIAM role assumed via OIDC federation (e.g. GitHub Actions) or IAM Roles Anywhere (X.509, on-prem)Avoid IAM user access keys for machines.
Management group policies limiting what admins can doService control policies (SCPs) and resource control policies (RCPs) in OrganizationsSCPs cap what principals in an account can do; RCPs cap what can be done to resources. They grant nothing by themselves.
Privileged Identity Management (PIM)Identity Center permission sets with short sessions, plus temporary elevated access tooling ⚠ VERIFY current native optionNo exact built-in equivalent to PIM's approval workflow in every setup; banks often add a ticket-driven broker.
Conditional AccessIdP-side conditional access (keep it in Entra) + IAM condition keys (source IP, VPC endpoint, MFA present, tags)Keep device/risk policies in Entra; enforce network and data perimeters in IAM.

Secrets and keys

Microsoft / AzureAWS equivalentWhat's different / watch out for
Key Vault — secretsAWS Secrets Manager (rotation, replication) or Parameter Store SecureStringTwo services: Secrets Manager for credentials that rotate (≈ $0.40 per secret per month — check pricing), Parameter Store for config and simple secrets (standard tier is free).
Key Vault — keysAWS KMS (customer managed keys)KMS keys never leave the service; you call Encrypt/Decrypt/GenerateDataKey. Every key has a key policy — if the policy doesn't allow you, IAM alone won't help.
Managed HSM / Dedicated HSMAWS CloudHSM, or KMS with a CloudHSM custom key storeSingle-tenant FIPS-validated HSMs; you manage users and clustering.
Key Vault — certificatesAWS Certificate Manager (public, free for ALB/CloudFront) and AWS Private CAStandard ACM public certificates work only on integrated services (ALB, CloudFront, API Gateway); exportable public certificates are a separate paid option ⚠ VERIFY. Private CA has a monthly fee per CA.
Azure App Configuration (+ feature flags)AWS AppConfig (feature flags, validated gradual rollout) + Parameter StoreAppConfig can roll back automatically when a CloudWatch alarm fires.

Networking

Microsoft / AzureAWS equivalentWhat's different / watch out for
Virtual Network (VNet)Amazon VPCA VPC lives in one Region; subnets live in one AZ. You create the internet gateway, NAT and route tables yourself.
Network Security Group (NSG)Security groups (stateful, on each network interface) + network ACLs (stateless, per subnet)Security groups are allow-only and can reference other security groups instead of IPs — use that.
Default outbound internet / NATNAT Gateway in a public subnetBilled per hour and per GB — often the surprise line on the bill. Use VPC endpoints for AWS services to avoid it.
Private Endpoint / Private LinkInterface VPC endpoints (AWS PrivateLink)One endpoint per service per VPC (or centralised); hourly cost per AZ.
Service endpointsGateway endpoints (S3, DynamoDB only)Free; added to route tables.
VNet peeringVPC peering (not transitive) or Transit Gateway (hub)Overlapping CIDRs cannot be peered — plan IP ranges across all accounts up front.
Virtual WAN / hub-spokeAWS Transit Gateway, AWS Cloud WANUsually owned by a central network account.
ExpressRouteAWS Direct ConnectDedicated private circuit; add Site-to-Site VPN as backup.
VPN GatewayAWS Site-to-Site VPN (on a virtual private gateway or Transit Gateway)Two tunnels per connection; configure both.
Azure Load BalancerNetwork Load Balancer (layer 4)Static IP per AZ; preserves client IP.
Application GatewayApplication Load Balancer (layer 7) + AWS WAFPath/host routing, OIDC authentication action, Lambda targets.
Front Door / Azure CDNAmazon CloudFront (+ WAF), AWS Global Accelerator for non-HTTPCloudFront caches at edge locations; Global Accelerator gives static anycast IPs.
Azure DNS / Traffic ManagerAmazon Route 53 (hosted zones + routing policies + health checks)Weighted, latency, failover and geolocation routing are record settings.
Azure FirewallAWS Network FirewallOften deployed in an inspection VPC behind Transit Gateway.
Azure DDoS ProtectionAWS Shield Standard (free, automatic) / Shield Advanced (paid, response team)
Azure BastionSystems Manager Session Manager, EC2 Instance Connect EndpointSession Manager needs no open inbound ports and logs every session.

API management

Microsoft / AzureAWS equivalentWhat's different / watch out for
Azure API Management (gateway)Amazon API Gateway (REST, HTTP or WebSocket APIs)No single XML policy pipeline. Throttling, API keys and usage plans exist on REST APIs; JWT validation is built into HTTP APIs; custom logic goes into Lambda authorizers. See decision trees.
APIM policies (rate-limit, validate, transform)Usage plans + throttling, request validators and models, mapping templates (REST APIs), AWS WAF rate-based rulesComplex transformations are usually moved into code.
APIM products and subscription keysAPI Gateway API keys + usage plansAPI keys identify callers for quotas — they are not an authentication mechanism.
APIM developer portalAPI Gateway developer portal capability ⚠ VERIFY current feature name, or a self-hosted portal from the OpenAPI export
APIM VNet integration / internal modePrivate REST APIs (interface VPC endpoint) and VPC Link to private ALB/NLB
Mutual TLS in APIMAPI Gateway mTLS on custom domains (truststore in S3), or ALB mTLS

Monitoring and observability

Microsoft / AzureAWS equivalentWhat's different / watch out for
Azure Monitor metrics + alertsAmazon CloudWatch metrics + alarms (→ SNS)Alarms evaluate metric data; notifications go through SNS topics.
Log Analytics + KQLCloudWatch Logs + Logs Insights query languageSet log-group retention — the default is "never expire", which costs money.
Application InsightsCloudWatch Application Signals + AWS X-Ray traces, instrumented with OpenTelemetry (ADOT)Use OpenTelemetry for new .NET code; the X-Ray SDK for .NET is in maintenance mode.
Application Insights availability testsCloudWatch Synthetics canariesCanaries are scripted (Node.js/Python) and billed per run.
Azure Activity LogAWS CloudTrailCreate an organisation trail to a locked-down log-archive account; enable log file integrity validation.
Workbooks / dashboardsCloudWatch dashboards, Amazon Managed Grafana
Azure Service HealthAWS Health Dashboard
Azure AdvisorAWS Trusted Advisor, Compute OptimizerThe full set of Trusted Advisor checks needs a paid support plan ⚠ VERIFY current plan names.
SCOM / Event Log on serversCloudWatch agent + Systems ManagerShip Windows event logs and performance counters through the agent.

DevOps and infrastructure as code

Microsoft / AzureAWS equivalentWhat's different / watch out for
Azure ReposGitHub (most common) or AWS CodeCommit (back in GA since 2025-11-24)Most AWS shops use GitHub, GitLab or Bitbucket.
Azure Pipelines (build)AWS CodeBuild or GitHub ActionsCodeBuild runs each build in a fresh container with an IAM role.
Azure Pipelines (release, approvals)AWS CodePipeline (+ manual approval actions), CodeDeployCodeDeploy handles blue/green and canary for ECS, Lambda and EC2.
Service connectionsIAM roles assumed via OIDC from GitHub Actions / CodeBuild's own roleNo stored AWS keys in the CI system.
Azure ArtifactsAWS CodeArtifact (NuGet, npm, Maven, PyPI)
ARM templatesAWS CloudFormation (JSON/YAML)CloudFormation keeps state for you; use change sets before updating production.
BicepAWS CDK in C# (synthesises CloudFormation)Real C# with loops, classes and unit tests. Add cdk-nag for compliance rules.
Terraform (azurerm)Terraform (aws provider)You manage the state file — store it in S3 with locking, encrypted, versioned.
Template specs / BlueprintsAWS Service Catalog, CloudFormation StackSets, Control Tower Account Factory
Azure Load TestingDistributed Load Testing on AWS (solution) ⚠ VERIFY, or k6/JMeter on Fargate
Azure Chaos StudioAWS Fault Injection Service (FIS)
GitHub Advanced Security / Defender for DevOpsAmazon Inspector (ECR images, Lambda functions; code-repository scanning ⚠ VERIFY), plus your existing SAST/SCA toolsMost banks keep their approved scanners and add Inspector for running workloads.

Data and analytics

Microsoft / AzureAWS equivalentWhat's different / watch out for
Azure Data Factory (pipelines)AWS Glue (ETL jobs, crawlers), orchestrated by Step Functions or Amazon MWAA (Airflow)Glue jobs are Spark (PySpark/Scala) or Python shell — no visual "copy activity" with 100 connectors; Glue Studio gives a visual editor.
ADF mapping data flowsGlue Studio visual jobsGenerated PySpark you can read and version-control.
Synapse AnalyticsAmazon Redshift + Amazon Athena (serverless SQL on S3)Athena bills per TB scanned — Parquet + partitions cut cost dramatically.
Microsoft Fabric / OneLakeS3 data lake + Glue Data Catalog + Athena/Redshift + Quick Sight; SageMaker Unified Studio as the integrated workspace ⚠ VERIFY scopeNo single SaaS equivalent; it's assembled from services.
Azure Databricks / HDInsightDatabricks on AWS (partner) / Amazon EMR
Stream AnalyticsAmazon Managed Service for Apache Flink (formerly Kinesis Data Analytics)
Microsoft Purview (catalogue, lineage)Glue Data Catalog + Lake Formation (permissions) + Amazon DataZone / SageMaker CatalogLake Formation gives row-, column- and tag-based access control on lake tables.
Power BIAmazon Quick Sight (QuickSight is now part of the Amazon Quick suite — ⚠ VERIFY current branding)Many banks keep Power BI and connect it to Redshift/Athena.

Security and governance

Microsoft / AzureAWS equivalentWhat's different / watch out for
Management groupsAWS Organizations OUsPolicies (SCPs, RCPs, tag policies, backup policies) attach to OUs.
SubscriptionsAWS accountsSeparate accounts per environment and per workload is normal.
Azure Landing ZonesAWS Control Tower (+ Landing Zone Accelerator)Creates log-archive and audit accounts, guardrails and Account Factory.
Azure Policy (deny)SCPs / RCPs, declarative policies, Control Tower preventive controlsPreventive: stops the API call.
Azure Policy (audit)AWS Config rules and conformance packsDetective: records configuration history and flags non-compliance.
Microsoft Defender for CloudAWS Security Hub (the unified service, GA 2025-12) + Security Hub CSPM (the original posture checks, renamed) + GuardDuty + InspectorSecurity Hub aggregates findings; GuardDuty detects threats; Inspector scans for CVEs.
Microsoft Sentinel (SIEM)Amazon Security Lake (OCSF data lake) + your SIEM, or OpenSearchAWS has no first-party full SIEM; banks usually feed Splunk/Sentinel/etc.
Defender for Storage malware scanningGuardDuty Malware Protection for S3
Purview Information Protection / DLP discoveryAmazon Macie (PII discovery in S3)Macie only scans S3.
Defender vulnerability managementAmazon InspectorEC2, ECR images and Lambda functions.
Compliance ManagerAWS Audit Manager + AWS Artifact (AWS's own audit reports)Evidence collection helps — it does not make you compliant.
Cost Management + BudgetsCost Explorer, AWS Budgets, Cost Anomaly Detection, Data Exports (CUR 2.0)
Azure LighthouseOrganizations delegated administrator + cross-account roles
Resource locksSCP deny statements, CloudFormation stack policies and termination protection, S3 Object LockNo generic "CanNotDelete" lock on any resource.

AI

Microsoft / AzureAWS equivalentWhat's different / watch out for
Azure OpenAI / Azure AI FoundryAmazon Bedrock (many model providers behind one API)Model availability differs per Region — check before you promise data residency.
Azure AI Search (RAG)Bedrock Knowledge Bases (with OpenSearch Serverless, Aurora pgvector or S3 Vectors) ⚠ VERIFY current store list
Azure AI Content SafetyBedrock Guardrails
Azure AI Agent Service / Copilot StudioAmazon Bedrock AgentCoreBedrock Agents (now Agents Classic) closed to new customers on 2026-07-30.
Azure Machine LearningAmazon SageMaker AI
Document Intelligence (Form Recognizer)Amazon Textract
AI Language / Speech / TranslatorAmazon Comprehend / Transcribe & Polly / Translate
GitHub CopilotKiro (successor), Amazon Q Developer (no new sign-ups since 2026-05-15)Your bank's AI-tool approval policy applies to both.
Fraud detection servicesSageMaker AI models, rules on Kinesis/Lambda, AWS WAF Fraud Control (account takeover/creation)Amazon Fraud Detector is closed to new customers (2025-11-07).

Migration strategies in one table (the 7 Rs)

StrategyMeaningTypical banking example
RehostMove as-is ("lift and shift")IIS VM → EC2 Windows with AWS Application Migration Service
ReplatformSmall changes to use managed servicesSQL Server on VM → RDS for SQL Server
RefactorRe-architect for cloud-native.NET Framework monolith → .NET 10 services on Fargate + SQS
RepurchaseReplace with SaaSOn-prem CRM → SaaS CRM
RetainKeep on-premises for nowMainframe core banking, connected over Direct Connect
RetireSwitch offDuplicate reporting app nobody uses
RelocateMove a whole virtualised estate without changesVMware estate → Amazon Elastic VMware Service ⚠ VERIFY current offering

Where to go next

Independent learning resource — not affiliated with, endorsed or sponsored by Amazon Web Services. AWS and related marks are trademarks of Amazon.com, Inc. or its affiliates. All banking data in examples is fictional. Sample architectures are for learning and are not automatically compliant with GDPR, PCI DSS or any banking regulation.

On this page