Increasing numbers of Internet-connected fridges and grandparents, together with cloud-based service delivery hysteria, are pushing availability requirements for web-accessible services through the roof. Subscribers are less interested in the reasons for failure, and are largely disinclined to try and call anybody for help (who would they call?) Service unavailability leads to lost subscribers, lost momentum and fear of lost investment and business failure. Being up is important.
Small, upstart web properties have options for outsourcing pieces of their infrastructure and operations to get a leg up on network and platform availability. With escalating availability requirements and a desire to be able to serve hot markets opportunistically, we consider how deep we can dig this particular rabbit-hole. We describe some of our thinking about how to scale our current service delivery platform from 20 sites globally to something much, much, much (much) bigger. We consider logistics, security, provisioning, manageability, monitoring and measurement, and begin to paint a picture of DNS service at a scale not previously seen on the Internet.