Attracted by challenges in large scale computing, and particularly inspired by the SKA compute problems on which I worked during the last five years, I seek opportunities to leverage novel systems software and hardware design. With the programming language community I am exploring concepts for hybrid memory systems, a logical sequel to my work on storage. Accelerators and other novel concepts in hardware design, such as number formats, are on my radar. I study machine learning for the sciences, as an alternative to traditional simulation, and as a scientific method. I give lectures and keynotes every year, about work I am involved in, sometimes happily explaining work of others.
Lustre File System
In 1995 I started to work on distributed file systems, and in 1999, unexpectedly, an opportunity emerged to design a new cluster file system which I named Lustre. Guided almost entirely by the requirements of the largest US government laboratories I led its architecture until 2008. Hundreds of individuals and hundreds of companies have made important contributions to it. It is now a cornerstone of scientific computing as one of the most used HPC software packages. My work on Lustre originated from and led to many other storage projects.
The square kilometer array (SKA) radio telescope requires very data intensive computing to convert the input from the antennas into usable scientific data. I work with the SKA Science Data Processing (SDP) group in the Cavendish Laboratory at Cambridge University to develop an architecture to approach these problems, and I advise the SDP project on other matters. A more detailed page about this can be found on this site.