Pass Your Amazon AWS Certified Solutions Architect - Associate SAA-C03 Exam Easy!

Autoscaling is bounded by maximum storage settings and supported scaling requirements; reaching the cap blocks further automatic allocation.

The maximum allocation is an independent explicit limit. Raising a minimum or baseline does not guarantee the database will exceed the configured ceiling when storage pressure grows.

Metrics can reveal the pressure but cannot change the maximum allocated-storage setting; revise the allowed maximum after validating limits.

gp3 provides separately configured IOPS/throughput within limits, so insufficient provisioning can cause stalls despite ample GiB.

Faster application clients cannot provision more RDS storage performance and may add queueing at the still-bottlenecked volume. Measure actual gp3 IOPS/throughput against the supported provisioned settings.

Backup configuration controls recovery behavior, not the storage performance provisioning that is saturated under the write workload.

Long-running heavy RPU consumption can be more expensive than a well-utilized provisioned warehouse.

Serverless measures processing usage; continuous substantial RPU consumption can make a well-utilized provisioned design a better comparison.

There is no universal free-rate threshold for indefinite heavy compute; estimate RPUs, reservations and query tuning under measured demand.

RA3 decouples managed storage and compute sizing, enabling storage growth without automatically adding nodes.

A schema is a database namespace, not a storage-tier capacity control. RA3 managed storage scales independently of compute based on underlying data footprint.

RA3 nodes supply the query compute tier, so needless nodes add costs when the workload does not require more CPU/concurrency.

Row filters on unpartitioned CSV often cannot skip whole objects; partition pruning and columnar formats reduce scanned bytes.

Timeout changes query failure behavior, not file partition pruning or physical columnar reads; scan costs remain dominated by the data actually accessed.

Query predicates do not rewrite physical S3 file layout; a Parquet conversion and partition strategy are separate data-engineering operations.

GSIs are physically maintained indexes; writes and stored projected attributes create separate billing dimensions.

A GSI can maintain projected attributes and indexed entry changes when base items are modified. Ignoring the index footprint and write activity can understate actual billing.

Changes that affect the indexed item can trigger additional GSI writes; projection size and update frequency influence usage charges.

Manual snapshots are independent persistent resources that can keep accruing storage charges until explicitly removed.

Automated backup retention and manual snapshot lifecycle are separate; retained manual copies must be inventoried and explicitly deleted only when no recovery or compliance dependency remains.

Manual snapshots are durable stored backup artifacts that can remain billable after the original database has been removed.

PITR requires an enabled supported rolling recovery window; disabled or out-of-window periods cannot be assumed recoverable.

AWS documents that disabling and re-enabling PITR resets the recoverable start time. It cannot reconstruct the missing prior continuous restore window; use valid independent backup recovery if available.

Streams capture change records for limited retention, not a managed continuous PITR backup history for restoring a table to an arbitrary time.

License portability depends on engine/edition, AWS offering and contract terms; some deployments use license-included pricing.

Prior on-premises license ownership does not automatically establish eligibility for a particular managed RDS SQL Server edition, deployment or mobility model. Verify AWS and Microsoft contractual requirements.

Availability deployment alone cannot waive SQL Server licensing rules. Validate supported BYOL models and qualified license-mobility/Software Assurance conditions.

A supported S3 gateway endpoint can route eligible VPC subnet traffic without traversing a metered NAT gateway.

A different service endpoint does not replace the S3 gateway route in this network.

Public paths do not meet the private-access goal or eliminate relevant Internet/NAT charges.

Interface endpoints can add fixed hourly per-AZ and usage costs that may exceed alternative permitted connectivity for sparse traffic.

Endpoint ENIs can accumulate fixed hourly charges even if routing/DNS directs no requests through them. Use real endpoint utilization and connectivity requirements before duplicating endpoints.

Endpoint classes have different pricing and service eligibility. Where a free S3/DynamoDB gateway endpoint meets access needs, compare it before paying for interface endpoints.