KubeCon + CloudNativeCon North America 2017

Never before have organizations wrestled with as much choice in how they compute and where they compute. The public cloud offers freedom from lead times and elasticity to manage changing workloads, but once a workload reaches a certain size or can be forecasted over a longer period of time it may be much more expensive than building and operating the compute infrastructure in-house. In the following talk we estimate when this threshold is crossed. We then explore what a modern datacentre should look like, why running an efficient compute infrastructure requires a spirit of radical simplification, and finally how focusing on important abstractions enables workload portability in an era with an abundance of choice.