Latency and progress in computing
by
Charlie Bess
I went to the University of North Texas recently with a number of other folks to talk with a class completing their master’s degree in computer science. One of the topics we discussed was the rate of change of computing capability. A presenter provided a chart similar to the following:
- 2010 - 3,000 MIPS: Lizard-scale intelligence
- 2020 - 100,000 MIPS: Mouse-scale intelligence
- 2030 - 3,000,000 MIPS: Monkey-scale intelligence
- 2040 - 100,000,000 MIPS: Human-scale intelligence
His numbers were more aggressive than these (saying we are already at mouse level computing capability today, but I looked around and pulled these numbers from another source). The details don't really matter, it's the concept of advancing technology that is important.
What struck me about this level of advancement is that the hardware that is capable of performing computation at the level of a mouse is actually being written by people who are most familiar with lizard-scale capabilities. It takes at least one generation to take real advantage of any new technology, once it comes on the scene.
There are many reasons why this happens, but one is: the zealot of today is the Luddite of tomorrow. We've probably all experienced the extinction of technologies we thought were leading edge at one time: structured programming – OO programming – SOA.
That's why EDS creates tools like BuzzTracker to allow us to see the relationship between technologies. It makes you realize that everything has a time when it is delivery-ready and a time when its usefulness draws to a close. Many times, its usefulness can be directly related to another complementary technology.
Organizations don’t currently appear to take into account these migration costs as part of their technology implementation. Anything that goes into production will be removed someday. With concepts like SOA, utility computing – additional flexibility and lower disposal costs will be available to those who want to take advantage of it.
