Study Azure AZ-305 Storage Choices: key concepts, common traps, and exam decision cues.
On this page
This lesson is about data shape and durability. AZ-305 wants you to recognize when the answer is a globally distributed document model, durable object storage, a data-lake shape, or shared file semantics, then match the durability and cost choices to the real requirement.
Start with the data model
Requirement
Strongest first fit
Why
globally distributed document or key-value workload
Cosmos DB
distribution, scale, and consistency trade-offs are first-class
durable object storage and lifecycle management
Blob Storage
object storage path with lifecycle and tiering controls
analytics-oriented hierarchical object data
ADLS Gen2
data-lake shape on top of Azure Storage
SMB or NFS shared file access
Azure Files
shared file semantics instead of object semantics
Durability and protection
Question
Strongest first reasoning move
what storage model fits the workload?
choose the data shape first
what durability level is needed?
then decide the replication or protection posture
what cost or performance balance is acceptable?
finally choose the cost-performance trade-off that still meets the requirement
Common traps
Trap
Better rule
choosing Cosmos DB just because the workload is modern
use it when global distribution, low latency, or document-model needs actually matter
choosing Azure Files for object workloads
file shares and object storage solve different access patterns
choosing the most expensive durability option by default
align durability to the business requirement, not to architecture anxiety
What strong answers usually do
classify document, object, data-lake, and file patterns cleanly
separate primary data model from durability level
weigh performance, features, and cost together instead of optimizing one in isolation
keep file semantics and object semantics distinct
Decision order that usually wins
Choose the data model first: document, object, lake, or shared file.
Decide durability and replication after the storage model fits.
Keep file semantics and object semantics separate.
Match durability to business need rather than to maximum anxiety.
Balance scale, latency, and cost only after the workload shape is correct.