Project:Infrastructure/Mirrors/Rsync-PoC

Objective: Reduce the maintenance burden of rsync mirrors by reducing the number of machines and aiming for higher reliability.

Implementation:


 * Keep the masterrsync host around, use it to feed the rest of the network nodes.
 * Run a network per continent (Asia, EU, US).
 * Automatically scale the network based on CPU usage.

A PoC of this is on GCP.

We have n1-standard-1 instances serving rsync out of a container.

Cost-wise:

Per region:
 * 2.5 cents/h => month is 18$/month for NLB.
 * $25/mo for a single n1-standard-1.
 * 1 TB of bandwidth is about $120.