Reserved Instances, Savings Plans, and Spot

AWS offers multiple EC2 purchasing models that trade flexibility for cost savings. On-Demand charges by the second with no commitment — the most flexible but most expensive. Reserved Instances provide up to 72% savings for 1 or 3-year commitments. Savings Plans provide up to 66% savings with a flexible hourly spend commitment. Spot Instances use spare capacity for up to 90% savings but can be interrupted. Dedicated Hosts provide physical servers for licensing compliance at premium cost. Understanding each model is critical for SAA-C03 cost scenarios.

Reserved Instances (RIs) are a billing construct, not a specific instance — they provide a discount applied to matching On-Demand instances in your account. RI types: Standard RIs — highest discount (up to 72%) but cannot change instance family. Convertible RIs — lower discount (~54%) but can exchange for different instance family, OS, or tenancy. Payment options: No Upfront (monthly payments, lowest commitment), Partial Upfront (lower effective rate), All Upfront (lowest effective hourly rate). Choose Standard RIs for predictable, unchanging workloads.

Reserved Instances have two scope options that affect flexibility. Regional scope provides the flexibility benefit of instance size flexibility within the same family — a regional m5 RI covers any size (m5.small, m5.large, etc.) in any AZ within the region. It does NOT reserve capacity. Zonal scope reserves capacity in a specific AZ (capacity reservation), ensuring instances are available during high-demand events, but has no instance size flexibility. Use regional scope unless you need guaranteed capacity (e.g., DR or strict SLA).

Savings Plans are a flexible pricing model where you commit to a minimum hourly spend (in dollars) for 1 or 3 years. AWS automatically applies the discount to eligible usage. Two types: Compute Savings Plans (up to 66% savings) — apply to EC2, Lambda, and Fargate across any family, region, OS, or tenancy. EC2 Instance Savings Plans (up to 72% savings) — apply within a specific instance family in a specific region, offering higher discount for less flexibility. Savings Plans are recommended over RIs for most new commitments due to their flexibility.

Key differences to guide your choice: RIs are better when you need capacity reservations (zonal RI), when you have very predictable workloads that won't change instance family, or when your older account has many existing RIs. Savings Plans are better for most new commitments because they automatically apply to new instance types (e.g., Graviton) without needing to exchange RIs. AWS recommends Savings Plans as the modern replacement for Compute RIs. For RDS, ElastiCache, and Redshift, Reserved Instances are still the mechanism (no Savings Plans for these services).

Spot Instances use spare AWS EC2 capacity and are priced dynamically based on supply and demand — typically 60-90% below On-Demand. AWS can reclaim Spot Instances with a 2-minute warning when it needs the capacity. Best use cases: big data processing (EMR), CI/CD build agents, stateless web applications with ELB, machine learning training (checkpoint model to S3 on interruption), and batch processing jobs that checkpoint progress. Design applications to handle 2-minute termination warnings gracefully.

To maximise Spot availability and minimise interruptions, use diversification: request multiple instance types across multiple AZs. A single Spot pool (specific instance type + AZ) can be interrupted simultaneously. Using mixed instance ASGs with many instance types (m5.large, m5a.large, m4.large, c5.large) dramatically reduces interruption risk — when one pool is reclaimed, the ASG replaces with another type. Use capacity-optimized allocation strategy to automatically pick the deepest pool (lowest interruption risk).

Batch workloads are ideal for Spot because they can be interrupted and resumed. AWS Batch natively manages Spot capacity, retrying failed jobs automatically and targeting optimal Spot pools. Amazon EMR lets you mix On-Demand master nodes (for reliability) with Spot core/task nodes (for savings). Managed Node Groups in EKS can use Spot instances, and Kubernetes' node-level graceful termination handles the 2-minute warning. These managed services abstract the complexity of Spot interruption handling.

The most cost-optimised architectures combine all three purchasing models: Savings Plans or RIs cover the steady-state baseline (the minimum number of instances always running), On-Demand covers predictable but variable overflow capacity, and Spot handles unpredictable burst traffic or batch workloads. In an ASG, set the On-Demand base capacity to your minimum reliable instances (covered by SP/RI) and configure the scaling increment to use Spot. This hybrid approach maximises savings while maintaining reliability.

After purchasing RIs or Savings Plans, monitor their utilisation to ensure you are using what you paid for. Cost Explorer RI/SP utilisation reports show the percentage of your committed capacity being used. A utilisation below 80% means you are wasting money on unused commitments. Investigate whether workloads have been terminated or resized. For RIs, you can sell unused RIs on the AWS Marketplace to recover some cost. For Savings Plans, there is no secondary market — ensure you are confident in the commitment before purchasing.

Dedicated Hosts allocate a physical EC2 server dedicated to your account, allowing you to use your existing per-core or per-socket software licenses (Windows Server, SQL Server, Oracle). Dedicated Hosts can be significantly expensive but are required for certain software licensing agreements. Dedicated Instances are a lighter version — instances run on hardware dedicated to your account but you do not control the specific host. Neither is recommended purely for security isolation (use VPC isolation and SGs instead); they are licensing compliance tools.

Test your understanding of AWS Solutions Architect (SAA-C03) concepts from this lesson.

In this lesson you learned: Savings Plans are more flexible than Reserved Instances and apply to EC2, Lambda, and Fargate, Spot Instances provide up to 90% savings for interruption-tolerant workloads, and combining Savings Plans for baseline with Spot for burst achieves maximum cost optimisation. Monitor utilisation of your commitments to ensure you are using what you paid for. Next up we explore Cost Explorer, Budgets, and cost allocation tags.