A senior fellow at Oracle told me earlier this year, "there is a line drawn between applications and the Operating System, and we don't cross it". The line is actually more like a solid wall, not to be crossed - or so says traditional wisdom.

The application space belongs to Oracle, BEA, SAP, and the like. Systems vendors - IBM, HP & SUN - don't cross that line; nor would it make sense to do so.

Oddly enough, application management doesn't cross the line either. Much of the management of an application is inferred through the underlying infrastructure. Values from the OS are interrogated to infer that the application is healthy and is within resource constraints. Network traffic is used to determine load on an application by examining specific packet content. This information is used to infer application activity.

To infer what an application is doing by examining values in the host infrastructure is not necessarily a bad approach. But it doesn't explicitly say what the application is using. What files are actually required? Which processes are parts of an application stack as opposed to the system services? What is the actual traffic across a specific socket connection? For that matter, what connections are being used?

This kind of application management works where a server image is equated to an application. An application is the server and they are effectively one in the same. But when demand increases for additional applications (and for increased capacity of existing applications), there is a corresponding explosion in the number of servers required. The additional servers and applications become increasingly difficult to manage over time.

Virtual machine technology and virtualization attempt to create an environment in which an individual application can be isolated from a single server. This is a step forward because a single hardware platform can be used to host more than one application. But what does it actually accomplish? If application demand continues increase, there are just as many virtual machines to manage as there had been physical machines. The problem has shifted around but hasn't really changed much.

What if that line - the solid wall between the application and the OS - could be crossed? What if the application could be extracted in a way that it could be used independent of the infrastructure? Direct interaction could take place with the application to determine what it requires, with certainty not inference.

What if the most difficult application stuck on an aging Solaris 6 platform could be picked up with all its dependencies and placed a Solaris 9 platform? Better yet, what if the application were placed on a Solaris 10 platform in a specific application zone, with all the isolation and individual identity that zones offer?

We built Trigence AE to more than cross the line: we built it to blow a big hole in the solid wall between the application and the OS.