Faaaster

Redesigning Backend Infrastructure

We undertook a complete redesign of the backend infrastructure using Google Cloud Platform (GCP), Go, and Kubernetes, transitioning from a monolithic architecture to a robust microservices setup. This change allowed us to dynamically spawn new Kubernetes pods to host user websites efficiently.

Scale-to-Zero System with KEDA

To optimize costs and resource management, I implemented a scale-to-zero system for WordPress using KEDA (Kubernetes Event-Driven Autoscaler). KEDA is composed of three primary components:

• Scaler: Monitors the metrics to decide when scaling should occur.
• Metrics Adapter: Integrates custom metrics into Kubernetes’ Horizontal Pod Autoscaler (HPA).
• Controller: Manages the lifecycle of scaling actions based on defined metrics.

If a user’s website received no requests for 15 minutes, the system automatically shut down the pod. Upon receiving a new request through our load balancer, the pod quickly spun back up. This optimization was especially effective as most of our users were independent professionals with lower traffic sites. As a result, we successfully reduced infrastructure costs by approximately 50%.

REST API and WordPress Plugin

I developed a custom REST API in Go for seamless communication between Kubernetes and WordPress. Additionally, I built a dedicated PHP plugin for WordPress, enabling users to:

• Restart pods
• Manage various backend operations

This empowered users to perform essential administrative tasks directly from their dashboards, eliminating the need for support intervention and enhancing user experience significantly.

Kubernetes Controller for ProxySQL & GCP Resources

I also developed a specialized Kubernetes controller responsible for efficiently managing ProxySQL and associated GCP resources. This component significantly streamlined our database proxy handling and resource management, contributing to overall improved infrastructure stability and performance.