AWS Compute Purchasing Strategy: Curriculum Overview
Identifying when to use various compute purchasing options
AWS Compute Purchasing Strategy: Curriculum Overview
This curriculum provides a comprehensive path to mastering the selection and application of AWS compute purchasing models. It is designed to equip Cloud Practitioners with the skills to optimize infrastructure costs based on workload characteristics, duration, and business requirements.
Prerequisites
To successfully navigate this curriculum, students should possess the following foundational knowledge:
- Cloud Fundamentals: Understanding of the AWS Shared Responsibility Model and Cloud Value Proposition.
- Compute Basics: Familiarity with Amazon EC2 instance types (e.g., General Purpose, Compute Optimized) and what an Amazon Machine Image (AMI) is.
- Billing Concepts: Basic understanding of how AWS billing works (pay-as-you-go vs. upfront payments).
Module Breakdown
| Module | Topic | Difficulty | Primary Focus |
|---|---|---|---|
| 1 | On-Demand Fundamentals | Beginner | Flexibility and zero commitment |
| 2 | Steady-State: RI & Savings Plans | Intermediate | Long-term cost optimization |
| 3 | Cost-Efficiency: Spot Instances | Intermediate | Handling fault-tolerant workloads |
| 4 | Compliance & Specialized Needs | Advanced | Dedicated Hosts and Instances |
| 5 | Decision Strategy & Cost Management | Advanced | Tooling (Cost Explorer, Budgets) |
Module Objectives
1. Flexible Computing (On-Demand)
- Identify use cases for short-term, unpredictable workloads.
- Understand the pay-by-the-second billing model for specific AMIs.
2. Committed Usage (RI & Savings Plans)
- Compare Reserved Instances (RI) (Standard vs. Convertible) for 1-year or 3-year terms.
- Evaluate Savings Plans for flexibility across EC2, Lambda, and Fargate.
3. Maximum Savings (Spot)
- Analyze how to bid on unused AWS capacity for up to 90% savings.
- Design strategies for workloads that can handle a 2-minute interruption notice.
4. Physical Isolation (Dedicated Options)
- Distinguish between Dedicated Instances (hardware isolation) and Dedicated Hosts (physical server control for BYOL).
Visual Anchors
Compute Selection Decision Tree
Cost vs. Commitment Plot
Success Metrics
To demonstrate mastery of this topic, the learner must be able to:
- Scenario Matching: Correctly identify the most cost-effective purchasing option for 5 distinct business scenarios.
- Cost Calculation: Estimate the percentage savings when switching from On-Demand to a 3-year No-Upfront RI.
- Tool Proficiency: Use AWS Cost Explorer to identify which instances would benefit most from a Savings Plan.
[!IMPORTANT] Mastery is not just knowing the definitions, but knowing when to trade flexibility for cost (e.g., choosing an RI for a 24/7 database vs. Spot for a batch processing job).
Real-World Application
In a professional setting, identifying the correct compute purchasing option directly impacts the company's bottom line.
- Startup Growth: Using On-Demand during the MVP phase to avoid being locked into hardware that may change.
- Enterprise Efficiency: Moving legacy steady-state applications to Reserved Instances to slash operational expenditure (OpEx) by up to 72%.
- Big Data: Leveraging Spot Instances for massive data crunching tasks that are stateless, allowing for 10x more processing for the same budget.
Examples Section
| Scenario | Recommended Option | Why? |
|---|---|---|
| E-commerce Site | Reserved Instances | High-traffic, 24/7 steady-state requirement. |
| Development/Test Lab | On-Demand | Used only during business hours; needs to be started/stopped easily. |
| Image Rendering Job | Spot Instances | Stateless, fault-tolerant, and requires massive compute at low cost. |
| Regulatory Requirement | Dedicated Host | Requires physical isolation and ability to use existing per-socket software licenses. |
| Multi-Region Cluster | Savings Plans | Provides flexibility if the workload moves between regions or instance families. |
▶Click to expand: The 2-Minute Warning Strategy
When using Spot Instances, AWS provides a 2-minute interruption notice via Amazon EventBridge or the Instance Metadata Service. To succeed in the real world, your application must be architected to save its state (check-pointing) to EBS or S3 immediately upon receiving this signal.