Mixed Instance Policies
What you'll learn: How to configure Auto Scaling groups that use multiple instance types and purchase options to optimize cost and availability.
What Are Mixed Instance Policies?
A mixed instance policy lets your Auto Scaling group launch instances from a pool of different instance types (like t3.medium, t3.large, m5.large) and purchase options (on-demand vs spot). Instead of being locked into one instance type, you give Auto Scaling flexibility to choose what's available and cost-effective.
Think of it like shopping with a grocery list that says "buy apples, oranges, or pears—whatever's freshest and cheapest today." Your Auto Scaling group gets similar flexibility.
Why Use Mixed Instance Policies?
- Cost savings: Spot instances can be up to 90% cheaper than on-demand, but availability varies. Mixing both keeps you running even when spot capacity is low.
- Better availability: If one instance type is unavailable in an Availability Zone, Auto Scaling can try another from your list.
- Performance flexibility: Combine similar instance types (same vCPU/memory ratios) to maintain consistent performance.
How It Works
You define:
- Instance type pool: A list of compatible instance types (e.g.,
t3.medium,t3a.medium,t2.medium) - Purchase options: Percentage split between on-demand and spot (e.g., 20% on-demand, 80% spot)
- Allocation strategy: How to pick from the pool (e.g., lowest price, capacity-optimized)
Auto Scaling automatically distributes launches across your configuration, balancing cost and reliability based on your rules.
Real-World Pattern
You set minimum capacity with on-demand instances for guaranteed baseline, then scale additional capacity using spot instances for cost efficiency. This hybrid approach protects critical workload while optimizing spend.
Key Takeaway: Mixed instance policies give Auto Scaling groups flexibility to use multiple instance types and purchase options, improving cost efficiency and availability without manual intervention.