We were paying $18/month for an idle t3.micro EC2 instance just to run a lightweight campus dining API that only receives traffic at lunch and dinner. Porting to Cloudflare Workers (V8 isolates at edge) reduced cold starts to 0ms, eliminated server maintenance, and stayed comfortably within the 100,000 free daily request tier. Serverless edge is perfect for university projects.

Setting `resources.requests.memory` tells the kube-scheduler where to place the pod based on guaranteed capacity. Setting `resources.limits.memory` is a hard ceiling: if your Node/Go process exceeds it, the Linux kernel invokes the OOMKiller and terminates the container with exit code 137. Always set memory limits to prevent memory leaks from starving neighbouring pods on the worker node.