Containers without infrastructure.
WHAT IT DOES
Runs a container, scaling instances with traffic, including to zero.
WHAT SCALING TO ZERO MEANS
No cost when nothing is happening.
WHY THAT MATTERS
Intermittent workloads cost almost nothing.
WHAT IT COSTS
Cold start latency when scaling from zero.
WHAT TO DO ABOUT THAT
Configure a minimum number of instances, where latency matters.
WHAT YOUR CONTAINER MUST DO
Listen on the port provided Start quickly Be stateless
WHY STATELESS
Any request may reach any instance, and instances are created and destroyed freely.
WHAT TO DO WITH STATE
Keep it in a database or object storage.
WHAT TO CONFIGURE
Concurrency, or how many requests one instance handles Memory and processor allocation Maximum instances, as a cost control Request timeout
WHY MAXIMUM INSTANCES MATTERS
Without it, a traffic spike or a loop scales without bound, and bills accordingly.
WHAT TO USE FOR AUTHENTICATION TO OTHER SERVICES
The attached service account.
WHAT TO MONITOR
Request latency, instance count, and errors.
