Server Survival: a tower-defense cloud architecture game

What is Server Survival?

As a Cloud Architect, players in Server Survival, an interactive 3D simulation game, are tasked with connecting services such as a Load Balancer, NoSQL DB, and CDN on a canvas controlled by a toolbar to ensure continuous traffic flow. It runs in the browser via Three.js, with three modes — Survival, Campaign, and Sandbox — sharing the same 26 services. Tower-defense framing turns ideas like queue-depth scaling and circuit breaking into things you watch fail live.

Key Cloud Architecture Concepts Taught

• ✓Observability — placing a Monitoring node turns on a live metrics dashboard (RPS, error rate, latency); campaign level 15 deliberately withholds it first so you feel what flying blind is like.
• ✓Auto-scaling with cold start — Compute fleets scale out once utilization crosses 70% and back in below 30%, but new instances need warm-up time, and Container Clusters warm up even slower.
• ✓Queue-depth scaling — a queue-fed fleet scales on backlog pressure, not just CPU, so a growing queue triggers scale-out on its own.
• ✓Circuit breaking and retries — an overloaded downstream trips a breaker once its rolling error rate hits 50%, and traffic gets retried with backoff.
• ✓Multi-region failover — GeoDNS splits traffic across regional stacks, and region-outage events show why a single region is never enough.
• ✓GPU inference economics — the GPU Cluster batches inference requests into one job, so a full batch is profitable while a half-fed GPU just burns its $60/min upkeep.
• ✓Power as a hard constraint — the grid caps out at 8 kW, each GPU draws 6 kW, and a Substation adds another 6 kW, so capacity runs out in watts before it runs out in dollars.
• ✓For failure reasons, badges are provided; each time a request is dropped, a badge appears on the node responsible for the drop, specifying the actual cause (no route, over capacity, read-only replica, and so on).

Game Modes: Learn, Survive, Experiment

• •Survival mode for open-ended pressure-testing: traffic accelerates on a timer (×1.3 at 1 minute up to ×4.0 at 10 minutes) while DDoS spikes and random events keep forcing you to adapt an existing build.
• •Campaign mode for structured, level-by-level learning: 25 levels across five chapters move from a first server and CDN through observability, auto-scaling, multi-region failover, and GPU inference, each with a pre-built starting architecture and a debrief.
• •Sandbox mode for testing one idea in isolation: set any starting budget, dial traffic mix across all seven traffic types, and turn upkeep off entirely, with no game-over to interrupt experimenting with the full set of 26 services.
• •Sharing a build with someone else: the share panel exports a PNG of your diagram or a URL with the architecture encoded in an `?arch=` parameter, so anyone who opens the link gets it pre-loaded.

Getting Started with Server Survival

Server Survival needs no installation to actually play: clone the repository, open index.html directly in a modern browser, and the game loads — there's no build step, since the same static files are what GitHub Pages serves live. The objective is to survive as long as possible while managing Budget ($), Reputation (%), and Service Health; the game ends if Reputation hits 0% or you go bankrupt at $-1000. If you want to touch the code instead, run `npm install` once and then `npm run check`, which lints with ESLint and runs the Vitest suite — that tooling is only for contributors, not for playing.

Alternatives for Cloud Learning

Frequently Asked Questions

The problem it solves

Most ways to learn cloud architecture are either a wall of vendor documentation or a certification course that tests memorized service names, and neither shows you what happens three minutes after you under-provision a Compute fleet. Server Survival turns that consequence into gameplay: a queue backs up, a circuit breaker trips, reputation drains, and you watch it happen instead of reading a diagram that explains it after the fact.

Who should try it — and who should skip

Try Server Survival if you're a developer or student who wants scaling concepts — auto-scaling thresholds, circuit breaking, multi-region failover, GPU batching economics — to click through as consequences instead of definitions; the debrief at the end of each campaign level makes it usable for self-study, and free browser access means trying one level costs nothing. Skip it if you need something mapped to an actual certification exam or a specific cloud vendor's console, since none of the 26 services here are AWS, GCP, or Azure services by name — they're generic stand-ins for the concepts.

Related repositories