amed “BuTinah” after the marine-protected archipelago reserve off the coast of Abu Dhabi, NYU Abu Dhabi is now operating the fastest high-performance computer (HPC) in the UAE. It already plays a critical role in supporting the computational requirements of the university’s research community.
BuTinah is also connected to Ankabut, the UAE’s national research and education network, supporting high-speed connectivity between educational, research, and non-profit organizations in the country. The state-of-the-art resource was built by Hewlett-Packard and is currently maintained at the Injazat headquarters in Abu Dhabi.
“As NYU Abu Dhabi continues to develop into a world-class center for cutting-edge research, having access to a computational resource that can manage a high volume of complex numerical calculations is imperative,” says NYU Abu Dhabi Provost Fabio Piano. “Scientific research is increasingly moving in the direction of mathematical model-based experiments as they provide an efficient way to develop and test theories on some of science’s most challenging questions.”
BuTinah runs at approximately 70 teraflops (the general measure of computational power, which equates to one trillion floating point operations per second) and consists of 512 super-dense compute nodes, each of which has a memory capacity of at least 48 gigabytes. Additional memory nodes supply 192 gigabytes of RAM, with an additional terabyte node dedicated for NYU Abu Dhabi’s Center for Genomics and Systems Biology. Graphics processing units and visualization nodes included in the system are used for specialized functions, such as the translation of data into images.
Researchers can access the HPC through their personal computers to run experiments and computations on applications that have been installed on the system, enabling efficient processing of experiments.
“If someone has a serial requirement, needing only one processor, but has 100 jobs with each taking a week to run, it would take 100 weeks to complete the job using a regular PC,” explains HPC manager Muataz Al-Barwani. “The HPC provides a form of task farming, allowing the researcher to submit all 100 serial jobs to be processed at the same time to have the results within a week.”