ACADEMIA
IU, TeraGrid facilitate massive data generated from powerful new electron microscope
- Written by: Writer
- Category: ACADEMIA
The new IU transmission electron microscope, acquired by IU late in 2008, is used for electron cryomicroscopy, a technique in which samples such as microscopic proteins are structurally analyzed at very low temperatures. Samples in an electron microscope are also subjected to an extremely harsh vacuum and intense radiation, and these biological specimens are easily damaged in the process. To overcome this obstacle and obtain a usable result, scientists record a large number of images of many samples and, using supercomputers, mathematically average the images to obtain a final composite image.
Composite images are often made using multiple images from hundreds of thousands of individual protein particles -- a process which can require tens of thousands of hours of computing time, and creates significant challenges when moving and storing these massive data sets.
The IU Research Technologies High Performance Applications (HPA) Group was approached by David Morgan, director of the Cryo-Transmission Electron Microscopy Facility (cryoEM), the IU facility that is home to the new microscope, to assist in building applications to optimize the microscope's performance and help to address the challenges associated with the data-intensive process.
"We especially needed a way for scientists using the microscope to be able to access and move exceptionally large data sets to and from the instrument quickly and easily," said Morgan. "The Pervasive Technology Institute HPA technologists helped us to achieve this using the Data Capacitor and other IU resources and applications. We were pleased to find the resources we needed to make the microscope work successfully available right here at home at IU and via the TeraGrid."
The resources include IU's Big Red and Quarry Supercomputers, High Performance Storage System (HPSS), and the Data Capacitor, a system to temporarily move and manipulate massive data sets -- all of which are available nationwide via the TeraGrid, the National Science Foundation's collaborative national network of high performance computing resources. Applications tools developed by the Pervasive Technology Institute (PTI) Research Technologies High Performance Applications (HPA) group optimized these resources to support the JEOL JEM 3200FS transmission electron microscope, making the process of using the microscope easier for scientists.
The IU Data Capacitor (DCWAN) uses the Lustre file system over a wide area network, allowing the microscope control system to write data directly to the locally mounted DCWAN. Using DCWAN along with the IU supercomputers and other TeraGrid resources, the HPA group has been able to achieve a high level of scaling and stability for the microscope.
"We are very pleased with the results thus far," said Robert Henschel, manager of HPA group. "The data handling and life cycle challenges have been addressed by seamlessly integrating the Data Capacitor and HPSS into the data acquisition process. Careful optimization of applications has more than doubled execution speed and enabled the use of TeraGrid resources beyond IU. We look forward to continued work on this project."