What You'll Learn
You'll discover how T-series (burstable) instances work like a rechargeable battery for CPU performance, allowing you to save money while handling occasional workload spikes.
Understanding Burstable Instances
Most applications don't need full CPU power all the time. A small web server might sit idle 90% of the day, then suddenly handle a traffic spike. Burstable instances (like T2, T3, or T3a types) are designed exactly for this pattern—they cost less because they're optimized for baseline performance with the ability to burst when needed.
How CPU Credits Work
Think of CPU credits like a rechargeable phone battery for your instance:
- Earning Credits: When your instance uses less than its baseline CPU (say, 20% for a t3.small), it earns credits automatically—one credit per minute of idle capacity
- Spending Credits: When you need more power (like during a traffic spike), the instance "spends" accumulated credits to burst above baseline—up to 100% CPU
- The Bank: Credits accumulate in a credit balance (up to a maximum cap) while your workload is light
Real-World Example
A t3.medium has a 20% baseline. If your application typically uses 10% CPU, you're banking credits. When a user uploads a large file requiring 80% CPU for five minutes, those saved credits let you handle it smoothly—without paying for a larger, more expensive instance type 24/7.
When Credits Run Out
If you burst frequently and drain your credit balance to zero, the instance throttles back to baseline performance until credits replenish. This makes burstable instances perfect for development servers, small websites, and batch processing—but unsuitable for sustained high-CPU workloads.
Key Takeaway: Burstable instances use a CPU credit system to provide cost-effective performance for workloads with variable demand—earning credits during idle periods and spending them during bursts.