After a couple decades building technology, you come to realize that technology development is really a cycle rather than a vector. For instance, the way in which AppLogic packages an application and its infrastructure is analogous to the way you’d launch an app on an SMP system with a shell script that starts processes and connects them through sockets. Jason Brooks has a short but interesting piece about this subject, and the way in which the iteration adds value.

Often the parallels aren’t obvious at first, because you start out trying to solve a problem rather than recreate a technology in a new space. As an example, we weren’t trying to create an operating system when we started AppLogic - we simply wanted to help folks scale web services. We wanted to give them a way to express the application structure so it could be implemented by a system rather than operators. Only after we had the first prototype running did it became clear we’d built a new kind of operating system, a meta operating system.

What’s unique is that a meta operating system has no APIs of it’s own, because the code runs in guest OS’s. Still, the meta OS controls resources, schedules processes and provides the user a console.

Realizing this was an important step and made completing the project easier. Realizing we had an OS meant we now had a template to refer to whenever we weren’t sure how to solve a problem.