Autoscaling

Rightsizing TPS

Three Go microservices modeling a Black Friday checkout flow, designed for RPS-based autoscaling with Prometheus metrics.

Overview

Rightsizing TPS is a Black Friday services lab: three independently runnable Go services — checkout, inventory, and payment — implement a bounded checkout flow with inventory reservation, mock payment authorization, and best-effort compensation on decline. Each service exposes health, readiness, and Prometheus-compatible `/metrics` endpoints with a bounded-label RPS contract, designed to drive Kubernetes autoscaling (KEDA/HPA) by request rate. The Kubernetes deployment, Prometheus scraping, and autoscaler configuration shown here are architectural design targets for a follow-up Minikube exercise, not infrastructure running today.

Flow diagram

Client (user traffic)

Kubernetes cluster

Ingress / load balancer

Node

Microservices

Prometheus

Kubernetes autoscaler (KEDA/HPA)target RPS

Idle — baseline traffic.

Client-side simulation — not live metrics.