Optimizing ingress with global server load balancing (GSLB)

Consider the situation shown in Figure 2.8:

Figure 2.8 – A geographically distributed application

This application has a server on the west coast of the US, and another one on the east coast. How do we make sure Source 1 in Seattle goes to Server 1 in Los Angeles while Source 2 in NYC goes to Server 2 in Atlanta? The simplest answer would be to tell Customer 1 to go to server1.myapp.io and tell Customer 2 to go to server2.myapp.io. That might work for a handful of customers:

Figure 2.9 – GSLB

Realistically, though, we probably want a single address of myapp.io. How do we make it so both customers automatically get routed to the server that is geographically closest to them? Historically, there have been two main approaches: GSLB and IP Anycast.

GSLB is just DNS with some extra intelligence added. Figure 2.9 shows how it works. The GSLB servers are...