Done reading? Take the timed quiz for Night 39.

Open quiz →

Night 39 — Keywords: compute, networking, integration

~2 hr. $0 AWS spend. Force words about time, path, and change feed decide these questions.


Plain language first

within seconds  → Lambda (often + API Gateway)
minutes / Auto Scaling group   → EC2 Auto Scaling family
predictable 9–5 → scheduled scaling
no NAT gateway + AWS APIs private → VPC endpoints
on-prem dedicated link → Direct Connect (not “private S3 endpoint”)
IPv6 outbound only → egress-only internet gateway
CloudFront auth near users + cheap → Lambda@Edge
many Regions for latency (expensive) → only if stem allows cost

Block 1 (~30 min) — Maps

Compute / scale

Force wordsPickKill
bursts within seconds; API frontendAPI Gateway + LambdaEC2/ECS/Beanstalk Auto Scaling
protect API backends from spikesAPI Gateway throttling (+ cache sometimes)“do nothing”; Multi-AZ API Gateway
predictable office-hours load; morning coldscheduled scalingwait for dynamic CPU
mixed instance types/sizes; known schedulescheduled over predictivepredictive (assumes same instance family)
need N after an Availability Zone diesAuto Scaling group min ≥ 2N across Availability Zonesmin = N with 1 per Availability Zone
default scale-in terminate preferenceoldest launch template (after Availability Zone balance)“oldest instance” / random first
memory / disk space on EC2CloudWatch agentdefault CloudWatch metrics
edge auth; don’t redeploy whole stack multi-RegionLambda@Edgemulti-Region + latency Route 53 for cost stems
CloudFront 504sorigin failover / origin groupignore high availability

Network / hybrid path

Force wordsPickKill
private AWS service access, no internet gateway / NAT gatewayVPC endpointDirect Connect, VPN CloudHub, encryption
dedicated on-prem ↔ AWS bandwidthDirect Connect“Direct Connect as S3 private endpoint” inside the VPC
interconnect many VPCs/accounts, share resourcesOrganizations + Resource Access Manager (when sharing)ParallelCluster

Integration / streaming

Force wordsPickKill
DynamoDB item change → notifyStreams + Lambda (+ SNS)DynamoDB Accelerator; Adapter without enable Streams
anonymize before durable NoSQLKinesis → Lambda → DynamoDBland in S3/Dynamo first
S3 event → many team queuesSNS fan-out → SQS (S3 → one destination only)two SQS queues directly on S3
S3 PUT → run Fargate/ECS task least effortEventBridge → ECS taskLambda wrapper unless needed

Block 2 (~60 min) — Stem drills

  1. Private subnet; S3+DynamoDB; no NAT gateway; no public Internet.
  2. Portal expects global spike; protect backends beyond network ACLs.
  3. Auto Scaling group slow every morning 9am; known schedule; mixed instance sizes.
  4. Fault tolerant: always ≥2 after Availability Zone loss; peak 6.
  5. Memory + disk utilization on Linux and Windows EC2.
  6. Long CloudFront logins + occasional 504; cost-effective.
  7. Real-time personal data; anonymize before NoSQL store.
  8. DynamoDB write capacity hot partitions / uneven keys.
  9. Default Auto Scaling scale-in; which instance preference after Availability Zone balance?
  10. SFTP least operational work was Night 36 — tonight: API bursts in seconds.
<details> <summary>Key</summary>
  1. VPC endpoints
  2. API Gateway throttling
  3. Scheduled scaling
  4. Min 4 / max 6
  5. CloudWatch agent
  6. Lambda@Edge + origin failover
  7. Kinesis → Lambda → DynamoDB
  8. High-cardinality keys
  9. Oldest launch template
  10. API Gateway + Lambda
</details>

Block 3 (~30 min) — Quiz

night-39-quiz.json (20 questions).


Tomorrow (Night 40)

Timed Practice Exam 2 (target ≥75%). For every miss: write the force words you missed — not a new service chapter.