Chapter 5. Conclusion

The case studies we’ve discussed provide two tales of foundational, large-scale infrastructure change: one of replacing Google’s distributed file system with its successor; the other of decoupling storage and compute. In both instances, the jet (or, at least, an engine) was rebuilt mid-flight and, in both cases, at least one lesson was clear: think about all the ways your infrastructure change can break your users, then don’t let the infrastructure change break them.

This requires a pre-flight checklist and some paranoia: know how things can go wrong, consider the whole spectrum of bugs, from benign to catastrophic, and their range in rarities.

Regardless of the type of infrastructural jet you’re piloting and fixing along the way, we hope these case studies will be as instructive for you as they were for us.