HPC Clouds

This week’s HPCWire contains my article on the state of science research funding and the future of high-performance computing infrastructure, with some associated comments by the editor, Michael Feldman. In the article, I argue that we need to rethink the approaches, mechanisms and funding processes to deploy infrastructure in support of scientific discovery.  All of this is related to each group identifying their core competencies and focusing on what they do best. 

In this world of computing clouds, researchers and universities largely focus on the science, leaving resource provisioning to service providers who exploit economies of scale to deploy large hardware and software capabilities at lower cost.  As I noted in the article,

In the facility-less research computing model, researchers focus on the higher levels of the software stack – applications and innovation, not low-level infrastructure. Administrators, in turn, procure services from the providers based on capabilities and pricing. Finally, the providers deliver economies of scale and capabilities driven by a large market base.

Cloud computing offers the intriguing potential to democratize access to multi-thousand processor computations and petascale data analysis. Considerable work remains to be done if we are to realize this vision, but the opportunities are great for broad access to computing and analysis at unprecedented scale.  For HPC, we must find the right software interfaces that reach beyond map-reduce in computational generality but eschew unnecessary complexity.

Let’s build a world where scientists can focus on science and not worry about who has the biggest or the best computational toys.  Success should accrue to those who have the best ideas, not those who have access to restricted infrastructure.

Coda: Science Funding

In the same HPCWire article, I also lamented the sad outcome of the push for additional physical science research funding in the U.S. At this week’s PCAST >meeting, we discussed this topic at length.  Watch this space for some outcomes and position statements. Lots of very senior people are very unhappy.