Rich Knepper’s talk as part of the International Conference on Infrastructures and Cooperation in E-science and E-humanities, in Leipzig, Germany, tackles the subject of “Living cyberinfrastructures and their impact on scientific research.”
Knepper, who is Manager of Campus Bridging and Research Infrastructure at Indiana University, says he started thinking about the idea for the presentation when it became clear to him that technology and scientific outcomes are very closely linked. The idea is that science becomes a part of the technology that supports it. The premise is used as a jumping off point to take a fresh look at the current research computing paradigm.
Today’s research cyberinfrastructure consists of systems, storage, networks, software and people in support of scientific research – these are large systems that require a substantial investment in technology and people.
This tier of advanced cyberinfrastructure has traditionally been focused on physics, chemistry, and molecular dynamics, domains which require complex solutions, involving massively parallel code, inter-node communication, and a low-latency network.
Knepper distinguishes how living cyberinfrastructure is different from standard infrastructure, noting that today’s leadership-class CI is not used the same way throughout its lifecycle nor is it designed for a single purpose. Also, not every user is using it the same way. This is very different from, say, a microscope, which has a more standard deployment.
Knepper highlights some of the shortcomings of this paradigm:
- Systems are designed to meet the needs of the domains that use them the most, so models tend to be used as the basis for new models, which can lead to inertia of purpose.
- Despite the “free capacity” offered by living cyberinfrastructure, there are other costs having to do with allocation, adaptation, sharing and attribution.
- Unsustainability of the collaboration/compete cycle.
With the looming exascale horizon and the proliferation of computational science outside traditional HPC domains, there is an opportunity to take a fresh look at our national cyberinfrastructure strategy.
To kick off this conversation, Knepper offers the following questions:
- Are new methods such as big data analytics a way of fitting some problems into highly computational models?
- What would cyberinfrastructure in support of e-humanities and e-social sciences look like?
- What are possible arrangements between traditional science and living cyberinfrastructure?