So you've hacked/cached/patched your way to a robust and modular microservice architecture. We did the same at LinkedIn, and it served our needs for years and years... but then we realized that in order to scale that architecture across multiple data centers, we had to revisit all of our past design decisions. User experiences and services could no longer assume millisecond eventual-consistency latencies and some of our databases were simply too large to replicate wholly in multiple DCs. In this talk, I'll discuss some of the design decisions we had to revisit and how you might architect your systems to be multi-datacenter compatible in the first place, so that if your company decides to pursue its own active/active architecture, you'll know what to do.