HPC DIRECTOR
More than just an investigation into a potential future, my centre is involved in a joint project to purchase the next ‘capability’ machine for use primarily by Los Alamos and Sandia.
Capability systems are built to solve the largest problems at the labs and typically land at around the top of the Top500 list of supercomputers.
Since this machine will be located at Los Alamos, it only makes sense to try to increase the capabilities of their similar central supercomputer store (which also uses a Panasas-based solution). That laboratory has deployed a central store; four petabytes of storage supports six supercomputers, including RoadRunner, which is currently number two on the Top500.
Currently, that system delivers 55GB/s to Los Alamos applications and, as part of this project, we will add an additional 160GB/s of performance. To support that effort, Sandia purchased four new shelves of Panasas gear and attached it to a brand new Cray XT5. The Cray IO system, called DVS,
currently supports Lustre and GPFS. Sandia, Panasas, and Cray have been working together to make the necessary design and software changes required to support a solution at its best possible capabilities.
Our missions are very demanding, computationally. We will continue to pursue the acquisition of state-of-the-art HPC resources in support of these missions. In addition to all of this, Sandia has a long history of research into new technologies that may be expressed and incorporated into our production machine purchases.
‘Currently we are working on some extremely advanced architecture designs’
Our research efforts in this area encompass a very wide range – everything from architecture, to high-speed networking, operating systems, and, of course, IO. Probably the best-known
instances of these research efforts were our partnerships with Intel, which resulted in the ASCI Red machine (the very first machine to achieve one teraflop/sec), and Cray, which resulted in the XT line of machines. Currently, we are working on extremely advanced architecture designs, memory technologies, a new operating system called ‘Kitten’, and a radical departure in high-performance file system design. We will be attempting to have these technologies influence or incorporated into any products that we might buy in the future.
One way we do that is by working with standards committees. For instance, our extensive experience with these very large machines has demonstrated that there are some potentially troublesome side-effects with the current IEEE POSIX file system interface. Together with Panasas, and the other US energy labs, we have been working on a set of extensions in order to enhance portability and future-proof our applications.
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44