Computation for Interdisciplinary Data

 March 19, 2012, Interdisciplinary Engineering Design Education Conference, Santa Clara, CA—Masoud Nikravesh, CITRIS (Center for Information Technology Research in the Interest of Society) director of computation science and engineering at UC Berkeley presented information on the intersection of computer science and other science fields. This computational science can be a driver for changes to improve many areas in society.

As compute power continues its exponential increases, it enables researchers the ability to apply simulations to their theories and expert experiments. The parallel growth of machine-to-machine communications allows smart sensors to provide more data in context. Together, these capabilities plus the social data networks form a triumvirate that can change the world.

As Moore’s Law approaches its limits, we are seeing more multicore systems. The main barrier to more compute power is energy use. Technology in the compute space continues to change rapidly. The top bracket of the 500 high performance computers are all obsolete in 6-8 years, and in 8-10 years all of them are out of date. Now all computers are moving towards the exascale and possibly the singularity.

Now a petaflop computer takes about 1 million cores, but a GPU has the compute capacity of a early ’90’s supercomputer. The computer power exists, but is not easily accessible. The onset of big data is changing the types of data we look at and the addition of sensors will only make the data volume increase faster. In ’10, the world generated 1.2 petaflops of data, and this data volume is projected to increase to over 35 zettabytes. In comparison, the human population is about 6 B people and is expected to grow to 9 B by ’50.

Therefore, the drivers for improvements in data processing are the data and compute capabilities. Computational science takes concepts from mathematics, computer science, and applications to create and manage large scale models. These large scale simulations can address non-physical models and also can handle simulations that other technologies cannot handle due the size of the data. Some of the applications are energy, environmental issues, health information, climate, biological sciences, and neuroscience. The efforts to address these issues requires a change in the paradigm of large platforms to emphasize their efficiency for the application.

The CITRIS program has detached some research from Lawrence Berkeley Lab with 120 faculty, 120 researchers, and 22 departments. They collaborate with industry to develop courses and emphasize large scale simulations and analysis. The programs they are developing are trying to address the growing need for more computer scientists, especially with knowledge of other areas of science.

One problem is that the State of California has been reducing its support for the University. At one time, Cal was considered in the top 3 worldwide, but has dropped to #9. We have to look at a new economy to support the research in institutions if higher learning. Silicon Valley has grown because of the high grade graduates from Stanford and Berkeley. Cal, unfortunately, has dropped in ratings to number 7 in the US, due to the lack of funding.

New funding models will help transfer the acquired knowledge into new designs and will encourage innovations. The state needs to enhance education to change its rankings. The ongoing reductions in funding are creating an unsustainable decline in the University’s ability to bring the next generation of graduates out for the next industrial movements.

The types of changes to build dynamic big data sets and analytics for simulation require the integration of areas like statistics and crowd sourcing. The societal impacts of technology have to be evaluated to increase our ability to address environmental and climate issues, among others. The tools have to predict and simulate changes in the next generation. For example, people rioted in Mexico because of a shortage of corn. The corn shortage was due in part to the US subsidies to convert corn to biofuel.

There are many opportunities for applications to address the content of large data, including energy, health, digital media, and smart cities. An intelligent cyber-physical infrastructure could repel attacks and self-heal. Biology and medicine could move into other industries and make both better. The research has to show how to shift jobs and train people in new skills to address the bigger social and economic impacts of technology.

We need to change the way we train students to help them keep up with technology changes. We need innovation in education to make the educational process as fluid and dynamic as the technologies they are creating. As the world becomes more interconnected, all transformations in technology, jobs, and education need to address the changes needed to fit people into the new environment. The educational cycles are not updating fast enough.

CITRIS is a young program and is growing fairly quickly. Many students are taking CITRIS courses, but are not in the program. The growing body of results from some of the programs is encouraging and leads one to believe that we have a chance to realign technology to be good for more people.