Gordon Haff wrote an insightful column today about âskeuomorphic virtualizaton,â in other words the notion of thinking of virtual servers as having the same look and feel of the physical machines lying underneath. The upshot is that virtualization ends up just being a form of partioning; users view virtualized instances as no different from a single physical server.
Gordon urges the end of skeuomorphism's reign of terror (my words, not his) now that containers are (back) in the news.
So don't simply stuff âa hodepodge,â as Gordon says, of operating systems versions along with your app and all that goes with it (web server, database, storage, comms) into a container as if it's just another virtualized machine.
Instead, he writes that it's time to think of âcontainerized operating systems, container packaging systems, container orchestration like Kubernetes, DevOps practices, microservices architectures, âcattleâ (shorter term) workloads, software-defined everything, and pervasive open source as part of a new platform for cloud apps.â
Old habits die hard. The relentless drive for efficiency in business focuses itself evermore on IT operations, breeding an instinct to view the new tool as simply an improved hammer and to simply try to slam things through with it more quickly. An increasing focus on DevOps, as seen from the high management level, can certainly reinforce this view.
But that would be wrong. Containers give everyone another shot at imagining their software as a number of services rather than a unified application. Further, as Gordon notes, they let you throw in an OS here, some orchestration there, a VPN there, and truly become efficient (and more future-proof) in the way you design and deploy what you're trying to do.
The original SOA days seem far away to me, but the idea of provisioning services is still a radical one. Too often today, software-as-a-service in the cloud still means software-as-an-application from the sky. The inexorability of Moore's Law allows people to slam things through for short-term gain rather than long-term efficiency, and this will be particularly true as containers become prized for their relative abstemious use of resources.
But an expeditious, rather than more profound, use of containers will just create tomorrow's hodgepodge. The potential of containers to reboot the way we think of services and how they should be delivered can logically lead to re-imagining how users should experience what the services offer â and how skeuomorphic those experiences should be, of course.