17.5 Cloud Deployment
Cloud deployment means running your model on rented infrastructure (AWS, Google Cloud, Azure, or local providers like AletCloud) instead of a single machine you own. The cloud gives you on-demand compute, automatic scaling when traffic spikes, and managed services so a small team can serve millions of requests reliably.
An Ethiopian e-commerce startup gets normal traffic on weekdays but a huge spike during a holiday sale. With cloud auto-scaling, extra model servers start automatically during the rush and shut down afterward — so the company pays for big capacity only during the few peak hours instead of owning idle servers all year.
Scenario
Your traffic is low most of the time but spikes 20x for two hours each evening. Which cloud approach is most cost-effective?
Check your understanding
1/4 · 40 XPWhat is a key advantage of cloud auto-scaling for model serving?