Glance's session load is spread across global centers. Each session is assigned to a server in a data center which is geographically close to the session's host, in order to minimize path latency. Sessions at each location are balanced across a pool of available servers.
Glance's distributed architecture ensures there is no single point of failure. Should one server or even an entire data center drop offline, the remaining data centers assume the load and new sessions are rerouted. Sessions that were running on a failed server will drop, but hosts typically are able to restart them within a minute.
Glance's servers are hosted by Amazon Web Services (AWS) and Microsoft Azure. Both AWS and Azure data centers offer excellent security against physical intrusion, power and network outages, and other adverse events. Glance operates backup servers for business continuity in a different geographical region from primary servers.
All production servers are hardened. Non-essential services are disabled and security patches applied as appropriate. Remote management is by secure SSH only, from pre-authorized IP addresses.