A rare datacenter outage has inspired me to build a backend that would persist globally.

Andromeda - The web gateway

  1. Exposing services (data fetching, user signup, etc) via HTTPS API
  2. Monitored and checked by AWS Route 53
  3. Connects to any healthy Messier nodes

Whirlpool - The storage handler & service router

  1. Multi-master Percona XtraDB cluster to persist and replicate data
  2. Data is then exposed via service handler, routed via gnatsd
  3. Able to survive an outage of the entire region and maintain availability

Messier - The service handler

  1. Responds to web gateway request, and manipulate data
  2. Responds to events (notification, updates, etc), and manipulate data
  3. Scalable and pluggable

Monitored by Vigil.