Software design is often seen as the process of introducing abstractions and protocols to simplify the act of composing larger programs from smaller ones.

However, these days, hardly any complex system is a monolith and is often composed of many interacting components (services, modules, other applications etc). In addition, infrastructure-as-code tools such as Chef are as much a part of the codebase as application code. When development teams run such systems, a good chunk of their time involves triaging errors or triaging user facing problems. The process of analysis and reasoning about a running system or designing for deployment concerns is very different from the process of synthesis and composition involved in writing application software. In this talk, I would like to explore some patterns and lessons learned in deploying & maintaining a complex system (https://snap-ci.com) over the past few years.