Free — no signup required

Amazon AppStream 2.0: The Ephemeral Stream

3 min read

Non-Persistence is the Core Design Principle

Imagine a library computer. You sit down, use it for an hour, log out, and the next person gets a completely clean machine — your browsing history, downloads, and open files are gone. AppStream 2.0 works exactly like this, but for any Windows application, streamed to a browser. It is designed for task workers, students, contractors, or kiosk scenarios where users need a specific tool, not a full computer.

  • Stateless: When a user logs out, the underlying EC2 instance is terminated. Any file saved to the local C: drive is permanently lost. This is not a bug — it is the design. It guarantees every user starts with a clean, known-good environment.
  • Fleets, Not Users: You don't provision individual users; you provision a Fleet — essentially an Auto Scaling Group of identical instances built from a golden image. If 50 students join a class simultaneously, the fleet scales out to 50 instances. When they leave, it scales back to zero. You pay only for active session time.
  • Storage Solution: Because the local drive is ephemeral, users must save files to persistent storage. AppStream integrates with: Home Folders (backed by Amazon S3), Google Drive, and OneDrive. When Home Folders are enabled, the user's Documents folder is transparently mapped to S3 — the user saves normally, but the data persists across sessions.

Fleet Types

AppStream offers two fleet types with different cost and latency trade-offs:

  • Always-On Fleet: Instances run continuously, even when no users are connected. Sessions start instantly (no cold start). Higher cost because you pay even during idle periods. Best for: business-critical applications where users cannot tolerate startup delays.
  • On-Demand Fleet: Instances are started when a user requests a session. Cold start takes 1–2 minutes. Lower cost — you pay only when sessions are active. Best for: scheduled use cases like university labs or training sessions where startup delay is acceptable.

The "SaaS-ification" Pattern

One of the most powerful AppStream use cases is turning legacy desktop applications into browser-delivered SaaS without rewriting them.

Scenario: You are an Independent Software Vendor (ISV) selling a heavy Windows application ("CAD-Pro 98") that requires a 5 GB installer and a specific GPU.

  • Old Way: Customer downloads the installer, struggles with dependencies, calls support.
  • New Way: Customer logs into your website. You embed an AppStream session in an iframe. They run "CAD-Pro 98" in Chrome instantly — no installation, no hardware requirements, no IT involvement.

The ISV controls the golden image (with the app pre-installed), manages the fleet, and bills customers per session. The customer sees a web app. This pattern is used by companies like Siemens (industrial software) and Autodesk (CAD tools) to modernize legacy desktop software delivery.

This is one of 18 chapters

Get every chapter — Kubernetes, Terraform, SRE, distributed systems, and more — with fast daily review built in.

See pricing