GCP publishes a 99.99% uptime SLA for GCE.
How would I calculate an estimated SLA if my application used services deployed on 1,000 GCE instances? Assuming that the services were dependent on each other, would the estimated downtime per year of my application be 1,000 * 0.0001 = 0.10 or 10% per year?
99.99% SLA uptime roughly translates to 52.5 minutes a year of downtime (calculator here). Generally if the service is down, it is down for everything in that geographical area. So if you had 1000 servers in the same geographical area and that area goes out of service, your downtime would be tied together.
However, most of the "big" cloud providers (e.g. Google) provide their services across different regions and zones. By spreading your infrastructure out across different areas, you can make your network more resilient to cloud provider outages.
