Managed Hosting Platform Infrastructure
Infrastructure behind a managed WordPress hosting platform — Kubernetes orchestration, managed databases, and the DevOps workflows that keep it running.
Photo: Brian Harris, Public DomainProblem
Hosting a large fleet of customer sites at scale means the infrastructure layer has to be invisible — any downtime, slow query, or billing error is a direct hit to customers' businesses.
Approach
Own the Kubernetes cluster running the platform's containerized workloads, the Amazon RDS databases backing both the core API and customer WordPress sites, and the billing and monitoring systems that keep operations visible. Led a proof of concept splitting the WordPress docroot into three tiers by how often it changes: immutable platform code baked into the container image, customer-writable plugins/themes on a per-site volume, and media offloaded to S3 behind a CDN — removing the network filesystem from the request path entirely.
Architecture
Kubernetes-orchestrated container platform behind a global CDN, with a three-tier storage split (image-baked platform code, per-site persistent volumes for customer code, S3 + CDN for media), per-site Redis object caching, per-tenant IAM-scoped S3 access, managed RDS databases per workload, auto-scaling, and a WAF/DDoS layer in front of it all.
Results
Proof of concept cut cached-page first-byte latency from 7.4s to ~70ms (over 100x) and reduced CPU throttling under concurrent load by 13x, by getting the WordPress docroot off a network filesystem, adding per-site object caching, and right-sizing compute. Separately led AWS cost optimization: cut monthly cloud spend from a $21.2K peak to $16K (~35%, ≈$3,950/mo) via a 1-year EC2 Savings Plan (~$900/mo, $10.8K/year, zero upfront cost), decommissioning idle test infrastructure, and consolidating five dormant AWS accounts — on track to an $8K/month target.