As a high school student I had to go through a Philosophy class, even though my curriculum was in sciences. Zeno's paradox, or the negation of movement, was one of the subjects of that class through which I suffered enormously. Years later, my daughter came back one day with some math homework: the subject was to explain why Zeno's paradox was wrong. And I suffered through it again. This familiarity with Zeno, which I really could have done without, lead me to apply it to IT and what I consider to be the ball and chain that slows IT progress. In Zeno's paradox a runner (Achilles) cannot catch a turtle which started a race earlier than him because each time the runner reaches the point where the turtle was, the turtle has of course moved forward. Repeating this reasoning leads to the conclusion that the interval will become very small, but that the runner will never catch the turtle. What's wrong with the reasoning is that it explains a continuous movement variation through a set of discrete events. But this is what we do in IT: we have a continuous progress of IT technology, hardware and software, and IT projects which are discrete events. When we decide to start an IT project, all hardware and software components are frozen for the duration, while technology continue to progress. The obvious conclusion is that IT is always behind the latest technology and will never be able to catch up. The consequence is that we drag legacy applications running on legacy hardware for sometimes decades. What it does to IT organization is that it forces them to support all their past sins: this diversity in the data center is a major impediment to progress. Standardization of infrastructures would lead to a major reduction in costs.

Unless we abstract the application from the technology. Virtualization is what enables this abstraction. In the past, we have applied it to files and disks, to internal memory, etc. to great success.

The only issue today is that virtualization is still proprietary, as is the computing environment. We can virtualize Intel based systems, or AIX based systems, etc. But we cannot relocate an AIX application and its OS on an Intel box. Implementing a proprietary "decor" on an Intel platform has been done before: mainframers have re-implemented their mainframes using this type of technology. But it does voraciously consume cycles and calls for a myriad of CPUs. What we can hope is that, in a couple of technology cycles, enough power will be available at a reasonable price and will let us build hypervisors that include "decors" for proprietary environments.

I strongly believe that it would enable IT to do so much more.

I welcome you opinion, even if you want to talk about Zeno's paradox.


JP Garbani