At GCP:NEXT, we learned about Google's global cloud load balancer with a (hypothetical) clever Santa tracker story. If you have amazing internal assets like Google, you can create load balancers in totally different way. The story was about a Santa tracker app that shows the status of gift delivery across Christmas day, but unfortunately, the app servers are misconfigured, down or overloaded. How do you deal with that? The session was presented by Anna Berenberg, Matt Smart, and Prajakta Joshi of Google; and Dmitry Shestak of Wix, a Google customer.
Unlike traditional hardware, with load balancers, software based load balancers (a big step forward), or even DNS-based load balancers, Google manages to use a single IP (VIP) to take requests, deal with different levels of resources and demand around the world and direct traffic to systems most capable of responding to it, and give best performance. WIthout going into details, this illustrates the point that great infrastructure can result in business benefits and better end-user experience. This is not unlike the brief that I wrote about Equinix's power of its systems and network as a platform for better business resuts. But this time, it is built for modern apps built on GCP, and rethinks how load balancers can be designed from the ground up. We saw an example of Wix.com, a customer who managed to scale as they grew at an extremely rapid pace.
My takeaway? If you architect applications in a modern way, you need to revisit the architecture to take advantage of it fully, and to exploit new possibilities open to you. This applies not only to the other levels of the softwrare stack, but underlying network infrastructure as well.