Skip to main content

Compute and Storage Workshop

Retrieval Prompts​

  1. State the abstraction ladder from IaaS to SaaS and the main cost/control tradeoff at each step.
  2. State typical cold-start ranges for Lambda (small Node/Python vs Java/.NET vs VPC-attached).
  3. State the three hard limits of AWS Lambda you are most likely to hit.
  4. Describe the AMI + instance type + autoscaling-group triad for EC2.
  5. State when EFS is the right answer versus EBS versus S3.

Compare and Distinguish​

Separate these pairs:

  • ECS vs Fargate (and why "ECS vs Fargate" is a confused phrasing)
  • Lambda vs Cloud Run (limits and shape)
  • EBS gp3 vs io2 (for databases)
  • S3 Standard vs Standard-IA vs Glacier Deep Archive (access cost vs storage cost)
  • RDS Multi-AZ vs cross-region read replica
  • DynamoDB eventual vs strongly consistent reads

Common Mistake Check​

For each statement, identify the error:

  1. "We chose a t3.small because it's cheap."
  2. "Lambda scales infinitely, so we don't need a concurrency limit."
  3. "Serverless is always the cheapest option."
  4. "We keep millions of user uploads on an EBS volume so they stay close to the app."
  5. "Aurora is just RDS Postgres with extra cost."

Mini Application​

For each scenario, pick a compute primitive and a storage primitive; justify both choices with one sentence each; name one limit that would push you up or down a rung:

  1. A cron job that reconciles daily invoices; runs 1x/day for ~3 minutes; reads/writes ~200 MB of CSVs.
  2. A request/response REST API with steady 500 RPS 24x7, p99 latency budget 120 ms.
  3. A video-transcoding worker that takes 2-20 minutes per job; bursts of 500 jobs per hour; reads from and writes to ~5 GB per job.
  4. A team's internal wiki used by 30 people; trivial traffic, moderate availability SLO.

Evidence Check​

Complete only if you can defend each choice against at least one reasonable alternative and name a specific limit or cost line that rules the alternative out.