Saturday, December 17, 2005

Challenge to Traditional Software Engineering

Microsoft and Google have combined with Sun to fund an new Lab in University of California, Berkeley to the development of new approach for software development.

In traditional software engieerning, work is completed in sequential stages starting from system concept design to development, assessment or test, deployment and operation. This approach is often too slow. Instead of infrequent, well-tested upgrades, code for internet services is continually being modified on the fly. This fix-it-as-you-go approach enables speedier deployment, but it also requires a large technical support group to make sure operations are not disrupted as bugs are resolved.

The message from Prof. David Patterson in UBC, the founding director of this lab: "right now, it takes a large company employing hundreds of really smart people to support internet service". The new approach from the lab is to develop technology that eliminates the need for such a large organization, opening up innovation opportunities for small groups or even individual entrepreneurs. It is obvious that the big software companies, such as Microsoft and Google, welcome the birth of this new approach. Thus they could save lots of human resources and financial resources for the company.

According to Prof.David Patternson, the new approach could achieve this goal by applying statistical mahine learning - the same technology used successfully in the recent antonomous vehicle grand challenge - to the development of computer systems.

No comments: