Computer technology has become pervasive in all aspects of our society in part because computers keep getting more powerful, both in how fast they run and in how much data they can store and process. The tremendous progress in semiconductor technology has given us smartphones with processing capability that far exceed that of the world's largest supercomputers from 40 years ago. Meanwhile, supercomputers and other large-scale systems have also benefitted from improved technology, reaching the ability to perform petaflops (10^15 floating-point operations per second) and to manage petabytes (10^15 bytes) of data. Such machines are vital for tasks such as predicting climate change and for powering Internet search engines.
Several important trends will influence the development of high performance computers:
- Whereas current large scale computer systems are designed either to maximize computational speed or to manage and process large amounts of data, future applications call for machines that can provide both capabilities.
- The challenges of reliability and massive parallelism will require major changes in all aspects of computer system design, including hardware, operating systems, programming, and applications.
- Physical limits to device scaling will make it increasingly difficult to continue the trends of increased performance and capacity that have spurred progress in computer technology for the past 50 years.
Randy Bryant was for many years Dean of the School of Computer Science at Carnegie Mellon University and is currently an advisor for Big Data in the Office of Science and Technology Policy in the White House.
This talk is part of the TU Vision 2025+.