CCES Unicamp

CCES researchers accelerate scientific cloud computing

A new technique developed by researchers at the Center for Computing in Engineering & Sciences (CCES) at the University of Campinas (UNICAMP) accelerates in over 100 times the execution of complex scientific calculations by computers. The acceleration is achieved by the OmpCloud software, developed to make it easier to adjust any scientific program to cloud computing services.

“This software takes to the cloud a program initially written for a local cluster by changing a single code line”, explains Guido Araújo, a researcher at the Institute of Computing at UNICAMP who developed the software along with PhD scholarship holder Hervé Yviquel and grad-student Lauro Cruz. The research has been published by ACM Transactions on Architecture and Code Optimization.

Yviquel illustrates in this video how OmpCloud works. By changing a single line in its code, a section of the computer graphics algorithm Ray Trace is rapidly executed by cloud service processors, generating a tridimensional image.

Araújo tells us that the development of OmpCloud begun with the need to take to the cloud scientific calculations developed in association with a team led by Munir Skaf, a researcher at the Institute of Chemistry at UNICAMP and head of CCES. The team developed a new parallel computing code that simulates the result of the collision of inert ionized gas molecules with a given protein molecule.

The research, published in July in the Journal of Computational Chemistry, simulates the collisions inside a mass spectrometer, a lab tool that identifies the presence of a protein in a sample. Each protein spreads the inert gas molecules in a distinct way, producing a different pattern of cross section. Computer simulations are essential to identify patterns corresponding to a certain protein presence in the experiments data.

These calculations usually take hours to be done in a conventional computer. Using a single processors core, for example, the calculations of the cross section of nitrogen gas with beta-lactoglobulin, a protein present in cow milk, takes around 5 hours. However, with the parallel computing code developed by Araújo’s team, the same calculation was completed in 6 minutes by 256 processor cores form Microsoft Azure cloud service.

Research institutes from around the globe increasingly turn to large scale cloud computing services. According to Araújo, these services offer the advantage of taking care of the maintenance and update of large computing centers, giving their clients on demand remote access to tens of thousands processor cores.

OmpCloud’s performance has drawn the attention of big companies of the segment. Microsoft Azure service has offered 100 thousand dollars in cloud credits to fund Araújo’s team’s research, which has also received 60 thousand dollars in cloud credit from Amazon Web Services.

Scientific Articles

YVIQUEL, H.; CRUZ, L.; ARAÚJO, G. Cluster Programming using the OpenMP Accelerator Model. ACM Transactions on Architecture and Code Optimization, New York, v. 15, 3. is., art. 35, oct. 2018. Available at: Access on: jan. 27, 2019.
ZANOTTO, L.; et al. High performance collision cross section calculation – HPCCS, Journal of Computational Chemistry, v. 39, 21. is., 5 ago. 2018, p. 1675-1681. Available at: Access on: jan. 27, 2019.

Related posts

Modeling and Building IoT Data Platforms with Actor-Oriented Databases

cces cces

Bio‐inspired Aluminum Composite Reinforced with Soft Polymer with Enhanced Strength and Plasticity

cces cces

Líquidos iônicos e suas interações com proteínas: uma visão molecular

cces cces
WP Twitter Auto Publish Powered By :