Week 4 review — storage, analytics, edge, compute, integration

30 timed scenario questions (~48 min, ~96 sec each). Reinforces night-29-week-4-review.md and Nights 23–28. No new lab deploy. Answer without peeking at night-29-quiz-answers.json first.

1. 1.The Iceland Fargate pipeline writes processed GeoJSON that must survive task stop, Spot interruption, and AZ failure. Temporary sort files on the worker can be discarded after upload. Where should durable output live?

   A.S3 Standard in the pipeline bucket — durable object store is the system of recordB.EBS gp3 data volume attached to the Fargate task for the full pipeline lifetimeC.EFS file system mounted by every Fargate task as shared durable pipeline outputD.Instance Store on the underlying host claimed as durable sequential write storage

2. 2.Night 23 Lambda implements cache-aside for resort_stats rows. On a cache miss, what is the correct sequence?

   A.Redis read-through proxy automatically runs SELECT against Aurora without Lambda logicB.Lambda GET Redis key → miss → SELECT Aurora → SETEX Redis with TTL → return JSONC.Lambda writes Aurora first on every request, then CloudFront invalidates the edge cacheD.Lambda publishes a miss event to SNS so Aurora pushes the row into Redis asynchronously

3. 3.A mobile app needs server-side session state with automatic Multi-AZ failover and sorted-set leaderboards. Which ElastiCache engine?

   A.Memcached cluster — multithreaded horizontal scale without Redis data structuresB.Memcached with DAX in front for session persistence across node restartsC.Redis replication group — structures, optional persistence, automatic failoverD.CloudFront edge cache with session cookies included in the viewer cache key

4. 4.GSA serves static wiki HTML globally and caches resort_stats JSON inside the Night 9 VPC. Which pairing matches each layer?

   A.ElastiCache at CloudFront edge POPs; Aurora placed in public subnets for faster readsB.CloudFront caches Aurora rows; ElastiCache replaces Route 53 DNS answer selectionC.S3 Intelligent-Tiering caches HTTP GET responses at the edge for api/wiki* pathsD.CloudFront caches HTTP at the edge; ElastiCache Redis holds app objects near Lambda in VPC

5. 5.Executives want a one-time report: count of resorts grouped by country_code from last month's snapshot. Live wiki bookings must stay on Aurora with millisecond UPSERTs. Where do you run the report SQL?

   A.Athena on the Iceberg/Parquet snapshot in S3 via Glue catalog — not Aurora OLTPB.Aurora reader endpoint with a heavy GROUP BY during peak booking traffic hoursC.DynamoDB Scan with FilterExpression across the entire wiki counter table partitionD.CloudFront origin request that aggregates JSON at the edge on every page view hit

6. 6.Auditors ask: 'What did resort_snapshot look like before yesterday's bad ETL?' The table is Apache Iceberg on S3. Which feature answers without restoring a separate backup bucket?

   A.DynamoDB PITR restore to a new table at yesterday's timestamp for the snapshotB.Iceberg snapshot / time travel — query an earlier table snapshot in AthenaC.Aurora backtrack to rewind the writer instance in place to yesterday's stateD.EBS snapshot rollback mounted read-only on a new EC2 instance for the table

7. 7.An Athena table spans 8 TB in S3 with a partition column country_code. Analysts filter WHERE country_code = 'IS' on almost every query. What reduces cost most?

   A.Move the table to Aurora Serverless v2 and run the same SQL thereB.Enable S3 Transfer Acceleration on the bucket for faster scansC.Partition pruning on country_code so Athena skips irrelevant S3 prefixesD.Replace Athena with Kinesis Data Streams so analysts consume shards directly

8. 8.A retailer runs petabyte-scale nightly BI with hundreds of concurrent complex JOINs across curated marts. Unpredictable ad hoc SQL on raw S3 is a separate team. Best primary warehouse?

   A.Athena engine v3 with Iceberg only — no cluster to manage for all BI workloadsB.Aurora Global Database with read replicas for every concurrent analyst connectionC.ElastiCache Redis storing pre-joined rows from every warehouse fact tableD.Amazon Redshift for the heavy warehouse; Athena remains for ad hoc lake queries

9. 9.globalskiatlas.com uses Route 53 failover: PRIMARY alias to us-east-1 ALB, SECONDARY to S3 static site. PRIMARY ALB fails health checks. What must be configured for automatic DNS failover?

   A.Health check associated with the PRIMARY failover record — not optional for auto flipB.Weighted routing 50/50 between PRIMARY and SECONDARY without attaching any health checkC.Geolocation policy sending US users to PRIMARY and EU users to SECONDARY alwaysD.CNAME at the zone apex pointing to the ALB DNS name for faster DNS propagation

10. 10.You operate identical API stacks in us-east-1 and eu-west-1. Each user should hit the Region with lowest network latency among those two. Which Route 53 policy?

    A.Geolocation routing based on the user's country compliance and residency rulesB.Latency routing — Route 53 picks lowest latency among defined Regional recordsC.Simple routing with one A record listing both Regional endpoint IP addressesD.Failover routing with PRIMARY in us-east-1 and passive SECONDARY in eu-west-1

11. 11.You must point the apex domain witcoskitech.com directly to a CloudFront distribution. Which DNS record type is valid at the zone apex?

    A.CNAME from apex to d111111abcdef8.cloudfront.net — treated as a standard aliasB.TXT record with the CloudFront domain as the RDATA value for HTTPS validationC.A alias record to the CloudFront distribution — CNAME is forbidden at apexD.NS record delegating the apex hostname to CloudFront's hosted zone automatically

12. 12.A gaming company needs static anycast IP addresses for UDP game traffic with automatic steering to healthy Regional endpoint groups. HTTP caching is irrelevant. Best service?

    A.CloudFront with CachingOptimized policy on the default distribution behavior pathB.Route 53 latency routing to EC2 Elastic IPs registered in each game Region endpointC.API Gateway WebSocket API with stage-level throttling configured per player sessionD.AWS Global Accelerator with endpoint groups — anycast IPs for TCP/UDP workloads

13. 13.Production wiki API at globalskiatlas.com/api/wiki* uses Cognito JWT on writes. Anonymous GET /api/wiki/pages must always reach origin today. Current CloudFront behavior?

    A.CachingDisabled on the api/wiki* behavior — every request hits API GatewayB.CachingOptimized on api/wiki* with Authorization in the cache key for all usersC.Origin Shield in eu-west-1 replacing the need for Lambda entirelyD.Lambda@Edge rewriting POST responses into cacheable GET bodies at the edge

14. 14.Three external partners each have a contracted API rate limit. Exceeding one partner's quota must not throttle the others. Which API Gateway feature?

    A.AWS WAF rate-based rule on the CloudFront distribution viewer requestsB.Usage plan with per-partner API keys and individual throttle quotasC.Account-level throttle raised to the sum of all partner limits globallyD.ElastiCache token bucket shared across all partners behind one Lambda

15. 15.You enable CloudFront caching on anonymous GET /api/wiki/pages with a 60-second TTL. Which cache-key design avoids serving one user's private draft to another?

    A.Omit Authorization from the cache key and cache all variants under one shared URLB.Cache POST and PUT in CachedMethods so writes invalidate GET responses automaticallyC.Cache only anonymous public reads; exclude Authorization from key OR include it per safe variantD.Use Origin Shield as the only cache layer and disable API Gateway stage throttling rules

16. 16.A general-purpose Linux web server EC2 instance needs a boot volume with moderate random IOPS. Default cost-effective SSD choice?

    A.st1 throughput-optimized HDD for the root volume on the standard AMIB.sc1 cold HDD for the boot volume with provisioned IOPS at 16,000C.io2 Block Express for every t3.medium web tier instance by defaultD.gp3 — decouple size, provisioned IOPS, and throughput independently

17. 17.MySQL runs on EC2 (not RDS) with sustained 80,000 random IOPS on the data volume. Which EBS type?

    A.io2 or io2 Block Express — highest random IOPS for latency-sensitive DB on EC2B.st1 — throughput-optimized HDD for large sequential database log write streamsC.sc1 — lowest-cost cold HDD for infrequently accessed database data filesD.gp2 legacy with IOPS tied to volume size only — no independent provisioned IOPS knob

18. 18.A log aggregation EC2 instance writes one 400 GB sequential file per night. Cost-sensitive; random IOPS irrelevant. Best EBS data volume?

    A.gp3 with 16,000 provisioned IOPS dedicated to the nightly sequential append jobB.st1 throughput-optimized HDD — cheap sequential throughput, not random IOPSC.io2 Multi-Attach shared across six instances inside a single Availability ZoneD.EFS One Zone-IA mounted for block-level random database page read workloads

19. 19.Twenty AWS Batch EC2 instances must read and write the same PMTiles build configuration files concurrently during a job. Which storage?

    A.One gp3 EBS volume attached to all instances via io2 Multi-Attach in one AZB.S3 Standard with POSIX mount via Instance Store on each workerC.Amazon EFS — regional NFS shared across all Batch instances in the VPCD.ElastiCache Redis storing multi-megabyte binary tile files per key

20. 20.A stateless image-encoding fleet runs 6 hours nightly. Two-minute interruption notice is acceptable; checkpoint to S3 every 15 minutes. Best EC2 purchasing model?

    A.Dedicated Hosts for socket-level license compliance on every encoder instance typeB.On-Demand only — Spot is never suitable for any batch processing workload patternC.Reserved Instances one-year all upfront for a job that runs only 6 hours per dayD.Spot Instances with capacity-optimized allocation — stateless with S3 checkpoints

21. 21.GSA Iceland pipeline must run at 06:00 UTC on the first day of every month with no long-running queue buffer. Which service triggers the job?

    A.EventBridge schedule rule with cron expression targeting ECS RunTaskB.SQS standard queue with DelaySeconds set to 2,592,000 for monthly deliveryC.SNS topic with email subscription reminding an operator to click RunTaskD.Kinesis Data Streams shard iterator advancing once per calendar month

22. 22.A producer Lambda emits 5,000 messages per minute. Downstream workers process 500 per minute and must not lose messages during spikes. Pull-based buffer?

    A.EventBridge custom bus with no targets — events wait until a rule is addedB.SQS standard queue — durable buffer; consumers poll at their own paceC.SNS topic without subscribers — messages buffer until email is confirmedD.Step Functions Express workflow invoked synchronously for every message

23. 23.One order-placed event must reach billing, shipping, and analytics teams. Each team processes at its own pace with isolated retry/DLQ. Best pattern?

    A.One shared SQS queue — all three workers compete for the same order messagesB.EventBridge schedule publishes directly to three Lambda functions synchronouslyC.SNS topic fan-out to three separate SQS queues — one durable queue per teamD.Kinesis Data Firehose delivering the same record to only one S3 prefix destination

24. 24.Clickstream events must land in S3 within minutes for Athena queries. No custom consumer code or shard management. Best ingest path?

    A.Kinesis Data Streams with a custom Lambda consumer writing batches of events to S3B.SQS FIFO queue with long-polling workers uploading each message body into S3 objectsC.AppSync GraphQL subscriptions pushing events directly into the S3 bucket PutObject APID.Kinesis Data Firehose — managed delivery and batching to S3 without custom consumers

25. 25.A fraud team needs the same live transaction stream consumed by three independent apps with replay from the last 24 hours. Custom aggregation logic per app. Best service?

    A.Kinesis Data Streams — multiple custom consumers with configurable retentionB.Kinesis Data Firehose — three delivery streams cannot share one source streamC.SNS FIFO topic with three email subscriptions for each transaction eventD.EventBridge archive replay only — no ongoing multi-consumer stream

26. 26.Millions of payment authorization flows per day; each workflow completes in under 30 seconds. Full execution history and audit trail are not required. Step Functions type?

    A.Standard — always required when more than one Lambda function is in the workflowB.Express — high volume, short duration, optional discard of execution historyC.Standard with Wait states lasting up to one year for each authorization callbackD.EventBridge Scheduler replacing Step Functions for any multi-step payment workflow

27. 27.After Iceland ECS task exits 0 and reaches STOPPED, a stats uploader should pull work at its own pace. GSA Night 13 pattern. Success signal path?

    A.CloudWatch Logs subscription filter → SNS → email the uploader operatorB.ECS task directly invokes Lambda synchronously before the container exitsC.EventBridge rule on ECS Task State Change → SQS completion queue → workerD.SNS topic push to Lambda — no queue buffer between event and uploader

28. 28.CloudWatch shows FailedInvocations > 0 on the EventBridge rule but Invocations = 0. The ECS task never appears in the cluster. Most likely failure mode?

    A.Container exit code 1 — the task ran and the application returned an errorB.Successful RunTask — SQS redrive policy moved the message to the DLQC.Aurora failover — the schedule rule paused until the writer endpoint returnsD.Rule could not invoke target — IAM PassRole, target ARN, or network config

29. 29.Seven years of monthly Iceland pipeline GeoJSON prefixes are accessed once a year for compliance audits. Retrieval can take hours. Lowest-cost S3 tier?

    A.S3 Glacier Flexible Retrieval or Deep Archive with lifecycle from StandardB.S3 Standard with CloudFront caching disabled on the long-term compliance prefixC.EFS Infrequent Access mounted by auditors' EC2 instances inside the study VPCD.EBS sc1 snapshots exported monthly to S3 Standard for long-term object retention

30. 30.Wiki viewCount reads hit DynamoDB hot keys at microsecond latency inside the same Region. Which caching service is purpose-built for DynamoDB?

    A.ElastiCache Memcached cluster in the Night 9 VPC private subnetsB.CloudFront cache behavior on GET /api/wiki/pages with zero TTLC.DynamoDB Accelerator (DAX) — in-memory cache for DynamoDB API callsD.Athena CTAS materialized view refreshed every minute from DynamoDB Streams