Technology: Compute, Storage, Networking, and Managed Services
What to know
This domain is the widest. Build a mental map of four areas. Compute: virtual machines for full control, containers for portable packaged workloads, serverless functions for event-driven short tasks, and managed platform runtimes for web apps. Match the workload to the model - steady predictable load suits reserved virtual machines, spiky event traffic suits serverless.
Storage: object storage for unstructured data and static assets, block storage for disks attached to a single instance, file storage for shared POSIX access, and archival tiers for rarely read data with retrieval latency. Know durability versus availability and lifecycle policies that transition or expire data.
Networking: virtual networks with subnets, route tables, gateways, load balancers, DNS, and content delivery networks. Understand public versus private subnets, how a load balancer distributes traffic, and how a CDN caches content close to users.
Managed services: databases with replication and automated backups, message queues for decoupling, and monitoring services. Know that managed does not mean configuration-free - you still choose redundancy, backup retention, and access rules.
Traps
Questions often hide a constraint such as "shared across many instances" (rules out block storage) or "milliseconds latency for random reads" (rules out archival tiers). Watch for decoupling scenarios: if two components must not fail together, insert a queue. Distinguish vertical scaling (bigger instance) from horizontal (more instances).
Worked rule of thumb
Read the constraint words first - shared, durable, low-latency, event-driven, global - and eliminate options that violate any single one before comparing cost.
Study approach
Sketch the request path from user to data store and name the service at each hop.