Conway’s Law is a bear. Alternative approach, also from a financial services world: VMs are created with a DSL on top of qemu/firecracker, containers with Dock

Conway’s Law is a bear. Alternative approach, also from a financial services world: VMs are created with a DSL on top of qemu/firecracker, containers with Dockerfiles. Cyber are part of an image review group alongside other engineers that validates the base images. But: no interactive access to any of these VMs at all. There’s hypervisors running on bare metal, but SRE teams have that scripted pretty well to the point a physical server can be added in a day or so. It does mean you’ve to be serious about logging, monitoring and health. This is one instance where we got it right (I think). We do have some legacy servers we’re trying to get rid of. But we’ve learnt we can run even complex vendor apps this way. Conway’s Law comes to bite us in other ways though! Like I said, it’s a bear.