EC2 + EBS + Spot

15 timed scenario questions (~24 min, ~96 sec each). Reinforces night-27-ec2-ebs-spot.md and iceland-pipeline-compute.md. Answer without peeking at night-27-quiz-answers.json first.

1. 1.You launch a new Linux application server on EC2 and need a single-AZ boot volume with balanced price and performance. You want to tune IOPS independently of disk size. Which EBS volume type is the default exam answer?

   A.sc1 — cold HDD for infrequent sequential reads at lowest $/GBB.gp3 — general-purpose SSD with provisioned IOPS and throughputC.st1 — throughput-optimized HDD for large sequential workloadsD.io2 Block Express — maximum IOPS for latency-sensitive databases

2. 2.A team runs Oracle on EC2 and needs 80,000+ sustained random IOPS with low latency. Cost is secondary. Which storage meets the requirement?

   A.io2 or io2 Block Express provisioned IOPS SSDB.gp3 with baseline 3,000 IOPS only — no further provisioningC.st1 throughput HDD — IOPS scale with volume size per TiBD.S3 Standard with byte-range GET as a block device substitute

3. 3.A log aggregation EC2 instance performs large sequential writes and reads. Lowest-cost HDD with high throughput matters more than IOPS. Which EBS type fits?

   A.gp2 — legacy general SSD tied to volume size for IOPSB.EFS One Zone-IA — NFS shared file system for many instancesC.st1 — throughput-optimized HDD for sequential workloadsD.Instance Store — ephemeral disks bundled with the instance

4. 4.Infrequently accessed bulk data on EC2 must use the lowest-cost magnetic-style HDD. Throughput requirements are minimal. Which volume type?

   A.gp3 with minimum 125 MiB/s throughput provisionedB.io2 Multi-Attach shared across three Availability ZonesC.EFS Standard — elastic NFS storage for shared accessD.sc1 — cold HDD optimized for infrequent access at low cost

5. 5.Global Ski Atlas runs the Iceland OSM pipeline monthly via EventBridge → Fargate. Jobs take about two hours and can be retried next month if interrupted. Which compute purchase model best reduces per-run cost?

   A.Fargate Spot capacity provider or EC2 Spot for the batch workerB.Dedicated Hosts — physical server isolation for the whole monthC.On-Demand EC2 reserved 24/7 with cron on the instanceD.Lambda with 15-minute maximum timeout for the entire ETL

6. 6.Production Aurora PostgreSQL holds resort_stats as the authoritative relational store. Which compute/storage pairing is inappropriate for the database layer?

   A.Aurora Multi-AZ writer/reader cluster with automated backups enabledB.RDS on gp3 volumes with Multi-AZ automatic failover enabledC.Spot EC2 with Instance Store as the only copy of database filesD.Aurora Serverless v2 with RDS Proxy in front of the Lambda tier

7. 7.You must capture a consistent point-in-time copy of an EBS data volume for disaster recovery and later restore to a new volume in another Region. Which mechanism?

   A.Enable S3 Transfer Acceleration on the attached bucketB.EBS snapshot stored in S3 — copy snapshot cross-Region if neededC.Detach the live volume and attach it read-only to a second instanceD.Convert the gp3 volume to sc1 in place without downtime

8. 8.A video transcoding worker caches temporary frames on local disk. Data loss on instance stop is acceptable and you want the lowest latency local disk. What should you use?

   A.gp3 root volume with maximum provisioned throughputB.EFS Elastic throughput mode mounted on all workersC.io2 volume with Multi-Attach in three subnetsD.Instance Store volumes included with the instance family

9. 9.Ten EC2 instances in one Region must read and write the same application configuration files concurrently without a custom sync layer. Which storage service?

   A.Amazon EFS — shared NFS file system across instancesB.Single gp3 EBS volume attached to all ten instances by defaultC.S3 objects mounted as a native POSIX block device on each hostD.ElastiCache Redis strings holding entire configuration files

10. 10.Your Spot Fleet runs stateless render workers. AWS may reclaim capacity with about two minutes of warning. Which allocation strategy tends to lower interruption rates at similar discount?

    A.Lowest-price allocation across all instance pools regardless of capacityB.Capacity-optimized allocation — pick pools with available Spot capacityC.Dedicated Host affinity — bind jobs to a specific physical serverD.On-Demand base capacity only — no Spot instances in the fleet

11. 11.A gp3 data volume attached to a running EC2 instance needs higher provisioned IOPS for a nightly batch window. What is usually required?

    A.Stop the instance, detach the volume, and create a snapshot firstB.Migrate the workload to Instance Store before changing IOPSC.Use modify-volume to raise IOPS/throughput — typically no detachD.Convert the volume to S3 Intelligent-Tiering for automatic scaling

12. 12.WORLD_SCALE PMTiles generation shards thousands of tile builds that write results to S3. Work is parallel and fault-tolerant. Which AWS service commonly orchestrates Spot compute for this pattern?

    A.Amazon Cognito user pools — authenticate each tile workerB.AWS DataSync — scheduled sync from on-premises tile storageC.Amazon ElastiCache — cache finished PMTiles in RedisD.AWS Batch with a Spot-based compute environment

13. 13.You want to enable EC2 hibernation so a dev instance resumes in-memory state after stop. Which root volume constraint applies?

    A.Encrypted root volume up to 150 GiB on supported instance typesB.sc1 cold HDD root volume with Multi-AZ attachmentC.Unencrypted Instance Store root with T2 unlimited enabledD.EFS root mount with Provisioned Throughput mode only

14. 14.An exam question asks which EBS volume types are generally suitable only as additional data volumes, not as root boot volumes on standard AMIs. Which pair?

    A.gp3 and io2 — both block general-purpose and IOPS SSDB.st1 and sc1 — throughput and cold HDD typesC.EFS and FSx — shared file systems instead of blockD.Instance Store and S3 — object and ephemeral stores

15. 15.After the Iceland Fargate task finishes, durable GeoParquet output must survive task stop and instance termination. Where should the pipeline write final artifacts?

    A.Fargate ephemeral storage only — fastest and durable by defaultB.EBS gp3 scratch volume without snapshot — primary system of recordC.S3 bucket prefix such as iceland/YYYY-MM/ — durable object storeD.ElastiCache Redis with AOF persistence for Parquet blobs