AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 11
Question 100
A standard Aurora PostgreSQL cluster has a stable writer and existing readers. Its customer analytics queries fluctuate from low traffic at night to high read-only demand during the day. The team wants the number of read replicas to change automatically from a measured target metric, without resizing the writer for every spike. Which AWS capability fits?
- Increase the writer instance class and turn off reader endpoints, since Aurora reader scaling policies can only change writer CPU capacity.
- Use RDS Proxy to add physical read replicas automatically when pooled client connections reach the chosen target utilization.
- Configure Aurora Auto Scaling for reader instances with suitable minimum/maximum capacities and a target-tracking metric, routing eligible reads through the reader endpoint.
- Increase Aurora cluster storage capacity on a schedule and expect the storage layer to create and remove reader compute instances based on usage.
Explanation
Correct Answer
C:Aurora Auto Scaling adjusts the number of eligible Aurora Replicas using scaling policies and CloudWatch metrics. It does not automatically scale writer compute; new reader instances can receive connections through the reader endpoint.
Incorrect Answers
A:Writer scaling can be appropriate for writer saturation but does not match the request to adjust reader count automatically based on daytime analytics load.
B:RDS Proxy manages and reuses database connections; it does not provision new Aurora reader instances or implement target-tracking reader-fleet scaling.
D:Aurora storage grows independently of compute in its cluster architecture. Increasing storage does not create or remove read replica DB instances to handle query demand.
Question 101
An organization transfers large predictable traffic volumes between its data center and AWS each day. It requires a dedicated private connectivity path with predictable contracted bandwidth rather than an internet-routed VPN-only solution. Which AWS service should it evaluate?
- Use CloudFront Functions to turn public HTTPS edges into dedicated data-center network circuits.
- Use an S3 gateway endpoint outside any VPC as the company's physical Direct Connect cross-connect.
- Use only Route 53 latency routing because DNS responses allocate dedicated network ports.
- Evaluate Direct Connect with the correct circuit, virtual interface, location and redundancy design for the hybrid network.
Explanation
Correct Answer
D:Direct Connect offers dedicated connectivity into AWS via supported locations and virtual interfaces. The architecture must separately design redundant links, route advertisement and encryption needs.
Incorrect Answers
A:CloudFront Functions are edge request-processing logic, not dedicated hybrid network transport or private leased capacity.
B:An S3 gateway endpoint is a VPC route-table construct and does not provide a physical cross-connect from a customer site.
C:DNS routing chooses IP answers and does not reserve private bandwidth or create a dedicated network attachment.
Question 102
A high-traffic CloudFront distribution caches public images. Its cache policy includes every request cookie and many irrelevant headers in the cache key, producing very low hit rates even for identical objects. What change best improves caching efficiency?
- Increase S3 object size so CloudFront can recognize duplicate URLs without a cache key.
- Tune the CloudFront cache policy to forward and key on only the headers, cookies and query parameters that actually change the response.
- Increase CloudFront maximum TTL to a week while still placing every irrelevant request cookie and header into the cache key.
- Enable CloudFront Origin Shield but preserve the current cache policy that varies each image on all viewer cookies and headers.
Explanation
Correct Answer
B:A minimal correct cache key improves reuse of cached variants while preserving differences required for content correctness. Validate TTL, personalization and origin request policy separately.
Incorrect Answers
A:Larger objects do not remove unnecessary cache-key variations; the same URL will still be fragmented by irrelevant request attributes.
C:Longer TTLs can help reuse an individual cached variant, but they do not eliminate the excessive number of variants created by irrelevant cookies and headers. Key normalization addresses the identified fragmentation.
D:Origin Shield adds another cache layer and may consolidate some misses, but identical images still become separate variants under an over-specific cache key. Fix the cache-key policy rather than paying for another layer first.
Question 103
A SaaS platform receives JSON records from hundreds of tenants and delivers them continuously to S3. Analysts query only one tenant at a time with Athena, and the team wants tenant-specific S3 object prefixes without a separate reshuffling job. A new Firehose delivery stream is being designed. Which feature fits?
- Enable Firehose dynamic partitioning at delivery-stream creation, extract tenant_id from records, and configure S3 prefix/error-output expressions.
- Add Kinesis enhanced fan-out consumers to an S3 bucket so objects are partitioned by tenant_id automatically.
- Choose S3 Intelligent-Tiering because a storage-class transition creates tenant_id key-value folder partitions.
- Use CloudFront path-based behaviors to rewrite incoming Firehose record payloads into partitioned S3 objects.
Explanation
Correct Answer
A:Dynamic partitioning extracts configured keys and groups Firehose-delivered S3 records by prefix; it must be enabled at stream creation and needs suitable error handling.
Incorrect Answers
B:Enhanced fan-out gives Kinesis stream consumers dedicated read throughput, not per-record S3 prefix routing from Firehose.
C:Intelligent-Tiering changes storage economics based on access patterns, not the S3 key layout of ingested data.
D:CloudFront routes HTTP edge requests; it does not perform Firehose stream-record extraction and S3 partitioned delivery.
Question 104
A data-platform team already operates an Amazon EKS cluster with spare schedulable capacity and namespace-level controls. Its analysts need a managed Apache Spark runtime for periodic transformations of S3 datasets, while the platform team wants Spark executors scheduled as Kubernetes pods on that existing cluster rather than managing a second fleet of EMR EC2 cluster nodes. What should the architect choose?
- Create a conventional EMR on EC2 cluster, because it is the EMR option that registers an existing EKS namespace and schedules its Spark executors as pods there.
- Register an EKS namespace with Amazon EMR on EKS as a virtual cluster, grant the necessary job execution role and submit Spark jobs that the EKS scheduler runs as pods.
- Run EMR Serverless and attach the job to the existing EKS namespace, because EMR Serverless allocates its compute solely as pods on customer-managed EKS clusters.
- Configure Amazon Athena SQL workgroups as Kubernetes node groups and submit the Spark application through the Athena query API.
Explanation
Correct Answer
B:Amazon EMR on EKS creates a virtual cluster by registering an existing EKS namespace, then runs supported Spark jobs as pods scheduled by the Kubernetes control plane. A scoped EMR job execution role provides permitted AWS resource access.
Incorrect Answers
A:EMR on EC2 provisions and manages its own EC2-backed EMR cluster infrastructure. It does not map an existing EKS namespace into a virtual cluster for execution as EKS pods.
C:EMR Serverless provisions compute managed by the EMR Serverless service; it does not register a user-managed EKS namespace for Spark pod scheduling. It fits a different infrastructure-ownership requirement.
D:Athena is a managed SQL query service, not an EKS capacity manager for arbitrary Spark applications. An Athena workgroup is not a Kubernetes node group and cannot host Spark pods in this manner.
Question 105
A research institute makes a large S3 dataset available to authorized external partner accounts. Partners generate substantial download traffic. The institute wants eligible request and data-transfer charges attributed to requesters, with access patterns that support this model. What should it evaluate?
- Enable S3 Requester Pays on the bucket and ensure partner clients explicitly indicate requester payment and have the required permissions.
- Move the bucket to CloudWatch Logs because its billing always follows the person downloading data.
- Purchase RDS Reserved Instances to transfer S3 data egress charges to partners.
- Attach an EC2 security group allowing TCP 443 and assume that reassigns S3 bills automatically.
Explanation
Correct Answer
A:Requester Pays shifts eligible request and transfer charges for supported cross-account use cases to requesters, subject to service limitations and explicit client request settings. Bucket ownership and permissions remain important.
Incorrect Answers
B:CloudWatch Logs is not a drop-in public object dataset sharing platform with automatic requester billing for S3 objects.
C:RDS commitments apply to qualifying relational database instance costs and do not change S3 request billing responsibility.
D:Security groups filter ENI network traffic and cannot alter the S3 bucket's Requester Pays billing semantics.
Question 106
An archive owner compares low S3 storage prices but frequently needs large collections restored within minutes. The cheapest deep archive option is being recommended solely from its price per GB. What additional tradeoff is essential before deciding?
- Assume every S3 Glacier class provides the same instantaneous object retrieval with zero restore fees.
- Evaluate retrieval time, restore request charges, minimum duration, object-size requirements and actual restore frequency for candidate archival classes.
- Use an ALB to convert slow archive object restores into immediate local disk access.
- Use an internet gateway default route to waive S3 storage-class minimum-duration charges.
Explanation
Correct Answer
B:Full lifecycle economics includes how quickly data must be available, restore request and data retrieval pricing, minimum storage-duration accounting and actual retrieval frequency. Looking only at monthly dollars per gigabyte can select a class that fails the needed restore window.
Incorrect Answers
A:Archived S3 classes have materially different restore mechanisms, retrieval time and fee structures. Deep Archive cannot be assumed to return objects instantaneously at no retrieval charge simply because its monthly storage rate is low.
C:An Application Load Balancer distributes HTTP requests to supported targets. It does not modify S3 Glacier restoration time, archival retrieval fees or minimum storage-duration requirements.
D:An internet gateway provides routing for eligible VPC traffic. It cannot waive the storage duration, restore request or data retrieval charges assigned to an S3 archival class.
Question 107
A mobile app invokes a Lambda function millions of times each month to convert small JSON documents. The function currently reserves 10,240 MB of /tmp ephemeral storage because an old deployment template used the maximum setting. Detailed testing shows temporary files never exceed 180 MB per execution environment; CPU, memory, duration and concurrency are already appropriately tuned. The team must reduce avoidable Lambda charges while preserving the function’s current logic. What should the architect change?
- Decrease the function memory allocation dramatically and leave its /tmp ephemeral storage at 10,240 MB, assuming memory size controls all ephemeral-storage billing.
- Mount Amazon EFS for every temporary document but retain the existing 10,240-MB Lambda /tmp allocation, assuming EFS mounting cancels ephemeral-storage charges.
- Reduce the function’s configured /tmp ephemeral storage to the included 512 MB after validating peak working-set headroom, without changing its independently tuned memory and timeout.
- Enable provisioned concurrency for every function version while leaving /tmp storage at 10,240 MB, assuming initialized environments are exempt from excess temporary-storage pricing.
Explanation
Correct Answer
C:Each Lambda execution environment includes 512 MB of ephemeral /tmp storage at no additional ephemeral-storage charge. AWS charges for allocated /tmp beyond that allowance based on configured excess storage and execution duration; measured 180-MB temporary files fit with adequate headroom at 512 MB.
Incorrect Answers
A:Lambda memory and ephemeral /tmp storage are separately configured and priced. Lowering memory may reduce CPU availability or lengthen execution, but does not remove charges for a still oversized /tmp allocation.
B:EFS is shared persistent storage with its own pricing and does not cancel the additional Lambda ephemeral-storage allocation. The data already fits in included local /tmp space and does not need a shared filesystem.
D:Provisioned concurrency addresses initialization-sensitive latency for chosen versions or aliases and introduces its own charges. It does not waive the pricing of excess /tmp capacity while the oversized allocation remains configured.
Question 108
An organization executes nightly simulations that can checkpoint to S3 and restart after instance interruption. Completion is flexible, and the goal is to reduce compute cost compared with always-on On-Demand capacity. Which purchase approach is appropriate?
- Run every simulation solely on three-year dedicated hosts even when the jobs are intermittent.
- Use EC2 Spot capacity for the interruptible workers with checkpointing, retries and appropriate capacity diversification.
- Use only RDS database reserved pricing to purchase EC2 worker hours at Spot prices.
- Disable checkpointing and assume a Spot instance cannot be interrupted once it begins.
Explanation
Correct Answer
B:Spot provides discounted spare EC2 capacity that can be interrupted. Retryable, checkpointed simulations are a good fit when capacity pools and termination handling are designed accordingly.
Incorrect Answers
A:Long commitments for intermittently used dedicated hardware may waste capacity and do not exploit the workload's interruption tolerance.
C:RDS pricing instruments are for relational database instance usage, not EC2 Spot compute purchasing.
D:Spot can be reclaimed, and removing checkpointing would create avoidable lost progress and inconsistent completion times.
Question 109
A DynamoDB table retains years of rarely read audit entries. Storage costs dominate while request volume is very low; the team wants to retain the DynamoDB API rather than migrate to an archive object format. Which cost option should it evaluate?
- Set a table-level S3 Glacier Deep Archive class and assume DynamoDB API reads become free.
- Buy provisioned RDS IOPS for the existing DynamoDB table.
- Move all items to a single hot partition key to reduce billed storage bytes.
- Consider DynamoDB Standard-IA table class and compare storage savings against its higher data-access economics.
Explanation
Correct Answer
D:Standard-IA offers different storage/request tradeoffs suited to large rarely accessed tables. The decision requires measuring actual read and write volume and index-related costs.
Incorrect Answers
A:S3 archive storage classes are for S3 objects; they cannot be attached directly to a DynamoDB table while retaining its normal request interface.
B:RDS IOPS are not a capacity or storage-class knob for DynamoDB tables.
C:Concentrating keys can create throttling and hot partitions; it does not inherently reduce the storage size of retained items.
Question 110
A popular catalog endpoint issues the same relational queries millions of times daily. Adding a dedicated cache cluster would save significant RDS read capacity, and the company has measured cache-hit potential and TTL requirements. Which cost optimization should it assess?
- Add an Amazon ElastiCache tier, compare cache fleet cost with reduced database load, and define invalidation and TTL behavior.
- Purchase larger RDS instances without testing caching because every repeated read needs a fresh database query.
- Move the SQL service into S3 Requester Pays so clients automatically pay the RDS compute bill.
- Replace the catalog tables with an SQS dead-letter queue to make reads free.
Explanation
Correct Answer
A:An in-memory cache can avoid repeated database reads when hit rate and tolerated staleness support it. The added cache and replication charges must be weighed against reduced DB resource needs.
Incorrect Answers
B:A larger database might increase read capacity but can cost more than caching repeated values, and it ignores the measured opportunity to eliminate redundant queries.
C:Requester Pays is an S3 request/data transfer billing feature and does not shift RDS relational query compute costs by changing the database API.
D:An SQS DLQ is a failure buffer for messages, not a queryable, low-latency authoritative cache for SQL catalog entries.
