We don't use AWS or GCE. The other providers don't charge anywhere close to as much for bandwidth... to the point that deploying across multiple providers and u

We don't use AWS or GCE. The other providers don't charge anywhere close to as much for bandwidth... to the point that deploying across multiple providers and using the Internet as our backplane network is no issue. One trick I like is to split infrastructure across 2-3 different providers at at least three different nearby data centers. Example: Chicago (Vultr), Toronto (DO), Montreal (OVH). Latency is low enough for most replication purposes and if one experiences a massive failure the other two will be fine. 3/3 failure probably requires alien invasion or global thermonuclear war.