Design infrastructure solutions

A retail company is designing a new order-processing backend on Azure. The workload consists of short, stateless functions triggered by messages arriving on an Azure Service Bus queue. Message volume is highly variable — near zero overnight and spiking to thousands per minute during flash sales. The development team wants the least operational overhead, automatic scaling driven directly by queue depth, and the ability to scale to zero to minimize cost during idle periods. They do not want to manage or configure any container orchestration or virtual machines. Which compute service should you recommend?

A logistics company runs an on-demand image-processing job several times a day. Each invocation spins up a container that runs for 3 to 8 minutes, then exits. Volume is unpredictable — sometimes zero jobs for hours, sometimes 40 concurrent jobs. The team wants to run these containerized workloads with per-second billing, no infrastructure or node pool to manage, and rapid startup, while avoiding paying for idle capacity. Which compute option best fits these requirements?

A retail company is migrating a monolithic .NET web application packaged as a single Linux container to Azure. The operations team is small, has no Kubernetes experience, and wants to minimize infrastructure management. The application requires deployment slots for zero-downtime releases, integrated authentication with Microsoft Entra ID, and the ability to scale automatically based on HTTP traffic. There is no requirement for complex service-to-service orchestration or multiple microservices. Which compute service should you recommend?

A retail company exposes several internal microservices through Azure API Management (Standard tier) to third-party partner developers. During flash sales, a few partners generate bursts of traffic that overwhelm the backend microservices, causing timeouts for all consumers. The architecture team wants to protect the backends by capping how many calls each partner subscription can make per time window, and reduce load for repeated identical read requests, without modifying the microservice code. Which combination of API Management capabilities should you recommend?

A retail company runs several microservices across Azure App Service and Azure Kubernetes Service. Development teams currently store configuration settings and feature toggles in each application's local files, requiring redeployments to change any value. The architect must recommend a solution that centralizes configuration, allows dynamic feature-flag changes without redeployment, supports point-in-time configuration snapshots for rollback, and integrates securely with Key Vault for secrets. Which Azure service should the architect recommend?

A retail company hosts a single public web application at www.contoso.com. Behind it are three backend pools of VMs: one serving the main storefront, one serving the /api/* endpoints, and one serving the /images/* static content. The architect must route incoming HTTPS traffic to the correct backend pool based on the URL path, terminate TLS at the edge, and protect the application against common web exploits such as SQL injection. Which Azure load-balancing solution should be recommended?

A company runs a single-region web application on Azure VMs behind a public endpoint in the East US region. Security requires that all inbound HTTPS traffic be inspected by a Layer 7 web application firewall, that TLS be terminated and re-encrypted to the backend pool, and that requests be routed based on URL path to different VM pools. The solution must stay within one region and must not introduce a global anycast entry point. Which service should you recommend?

A retail company runs a stateless web API on Azure Virtual Machine Scale Sets deployed in two regions: East US and West Europe. Customers are global, and the company wants all inbound API traffic to be automatically directed to the nearest healthy regional deployment based on network latency, with automatic failover if one region becomes unavailable. The solution must operate at the TCP/UDP layer because the API uses a custom binary protocol over a non-HTTP port, and it must expose a single static anycast IP address to clients. Which Azure load-balancing service should you recommend?

A retail company is migrating a portfolio of 200 on-premises servers to Azure following the Cloud Adoption Framework. During the current phase, the migration team needs to build a business case that quantifies expected Azure costs, identifies dependencies between application servers, and assesses each server's readiness for migration before any workloads are moved. Which Cloud Adoption Framework methodology and Azure tooling should the architect recommend for this phase?

A company runs a stateful ASP.NET web application across a horizontally scaled Azure App Service plan with 8 instances behind a load balancer. Users report being unexpectedly logged out and losing shopping cart contents when their requests hit different instances. The architects want a solution that externalizes session state so any instance can serve any request, provides sub-millisecond read latency, and scales independently of the web tier without requiring code that pins users to a single instance. Which approach should they recommend?

A pharmaceutical research firm needs to run a molecular simulation workload that consists of thousands of independent, short-lived compute tasks submitted in bursts throughout the day. The workload is embarrassingly parallel, requires no direct communication between tasks, and each task reads input from Blob storage and writes results back. Compute demand varies dramatically depending on how many researchers submit jobs. Management wants to minimize idle compute cost while providing a job scheduling and task distribution service so the team does not have to build orchestration logic themselves. Which Azure compute solution best meets these requirements?

A media company needs to process nightly video rendering jobs. Each job is a self-contained executable that reads input files from Azure Blob Storage, produces output files, and requires no coordination with other jobs. The workload is highly parallel, runs for 3-4 hours each night, and has no work during the day. The company wants to minimize compute cost and eliminate the need to manage any VM infrastructure or scheduler themselves. Which Azure compute solution should the architect recommend?

A retail company runs an e-commerce API on Azure App Service backed by Azure SQL Database. Product pricing data is read very frequently but updated only a few times per day by a back-office system. The team adds Azure Cache for Redis to reduce database load. However, when prices are updated, customers occasionally see stale prices for extended periods because cached entries only expire after a long TTL. The architect must ensure customers see updated prices immediately after a back-office change while still minimizing read load on the database. Which caching approach should be implemented?

A company is planning a large datacenter migration to Azure following the Microsoft Cloud Adoption Framework (CAF). They have hundreds of physical and virtual servers with poorly documented interdependencies. Before committing to a migration wave plan, the architecture team must produce an inventory of servers, assess their readiness and right-sized Azure targets, and understand which applications communicate with each other so that tightly coupled workloads are migrated together. Which approach should you recommend to gather this information with minimal manual effort?

A healthcare company is migrating a workload that processes highly sensitive patient records to Azure IaaS virtual machines. Compliance requires that data must be protected not only at rest and in transit, but also while it is being processed in memory, and the company wants hardware-based isolation that prevents the hypervisor and Azure operators from accessing the workload's memory. Which compute option should you recommend?

A startup is deploying several stateless HTTP microservices that experience sporadic, unpredictable traffic — often idle for hours, then bursting during marketing campaigns. The team wants to minimize cost during idle periods, avoid managing any Kubernetes control plane or node infrastructure, and support event-driven autoscaling based on concurrent HTTP requests, including scaling down to zero replicas. Which compute service should you recommend?

A retail company runs an order-processing system on Azure. The write workload (placing and updating orders) is moderate, but the read workload (order status dashboards, reporting queries, and customer lookups) is extremely heavy and involves complex joins that slow down the transactional database. The architecture team wants to scale reads independently, keep the transactional write model simple, and serve pre-computed query results with low latency. Which application architecture pattern should the architect recommend?

A retail company runs a mission-critical SQL Server 2016 database on-premises supporting a 24/7 e-commerce platform. Leadership wants to migrate the database to Azure SQL Managed Instance with the least possible downtime, and the application cannot tolerate an extended outage during business hours. The team needs an approach that keeps the source database available and continuously synchronized until a brief final cutover. Which migration approach should you recommend?

A company runs a stateless web API on a Virtual Machine Scale Set (VMSS) behind Azure Load Balancer. The deployment team wants to release new application versions with zero downtime and the ability to instantly roll back if error rates spike after a release. They also want to avoid maintaining double the compute capacity permanently. Which deployment strategy should the architect recommend?

A financial services company is building a serverless order-processing workflow on Azure. Each order must run through a sequence of steps: credit check, inventory reservation, payment capture, and shipping notification. Some steps call external APIs that can take several minutes, and one step requires waiting for a human approval that may arrive hours later. The team wants to keep code in a serverless model, maintain state across each step automatically, and support compensation logic if a later step fails. Which compute approach best meets these requirements?