Internet2 calls on two research teams for the final phase of the E-CAS project

0

WASHINGTON, DC, September 3, 2020 – Internet2 has confirmed the selection of two research teams using an external academic review board for the second and final phase of the Exploring Clouds for Acceleration of Science (E-CAS) project which was first announced in November 2018. The two research projects are:

  1. Iinvestigate heterogeneous computing at the Large Hadron Collider, Philip Harris, MIT. Only a small fraction of the 40 million collisions per second at the Large Hadron Collider (LHC) is stored and analyzed due to the huge volumes of data and the computing power needed to process them. This project proposes a redesign of algorithms using modern machine learning techniques that can be integrated into heterogeneous computer systems, allowing more data to be processed and therefore greater physical output and potentially fundamental discoveries in the field.
  2. Decipher the neural code of the brain through detailed large-scale simulation of cortical circuitry, Salvador Dura-Bernal and William Lytton, SUNY Downstate. This project aims to help decipher the neural coding mechanisms of the brain with far-reaching applications including the development of treatments for brain disorders, the improvement of brain-machine interfaces for paralyzed people, and the development of new algorithms for ‘artificial intelligence. Using a brain modeling software tool, researchers will run thousands of parallel simulations exploring different conditions and inputs for simulating brain cortical circuits.

The second phase of the E-CAS project builds on the lessons learned and best practices identified by the six research proposals selected in March 2019 with the aim of better understanding the use of cloud computing to accelerate scientific discovery.

“The first phase of the E-CAS project helped six teams develop their IT workflows and test them at scale, and the results from all teams were very impressive,” said Howard Pfeffer, president and CEO of E-CAS. ‘Internet2. “Now in phase two, teams at MIT and SUNY Downstate have the opportunity to build on their technological achievements using commercial cloud platforms with a focus on the scientific outcomes of their work.”

Investigating Heterogeneous Computing at the Large Hadron Collider

The MIT research team has developed a range of new tools and codes to take advantage of the latest graphics processing units (GPUs) and field-programmable gate arrays (FPGAs) to perform accelerated machine learning tasks in Amazon Web Services (AWS) and Google Cloud. using remote procedure calls from their core workflows running on high-performance clusters at MIT and FermiLab.

Fig. A: Overview of the computing grid used to process data from the LHC. Collisions like the one shown in the black box, which is a Higgs boson identified with an advanced deep neural network, start in Geneva, Switzerland, and are then sent around the world to high-performance computing (HPC) clusters to be processed. As part of the E-CAS project, the MIT-led team integrated GPUs in Google Cloud and FPGAs in AWS into existing infrastructure to enable HPCs to operate seamlessly by applying deep neural networks to speeds several orders of magnitude faster than local HPCs.

This allows the MIT-led team to accelerate the processing of data generated by the Large Hadron Collider by offloading deep neural network algorithms to cloud platforms, while running the CPU-intensive work of processing (CPU) on campus or in national research infrastructures. . The use of hardware accelerators and machine learning has increased the throughput of some algorithms by up to 20 times, up to 1,000 times that of CPU-based clusters, and allowed the integration of new, more advanced algorithms.

Deciphering the Brain’s Neural Code Through Detailed Large-Scale Simulation of Cortical Circuitry

The SUNY Downstate research team used a cluster of more than 100,000 cores on Google Cloud to run large-scale, highly detailed models of the brain’s motor cortex and auditory cortex. This allowed them to quickly explore large parameter spaces and use computationally demanding scalable algorithms to further refine models and reproduce experimental results.

Fig. B: GUI of the Multiscale Brain Circuit Modeling Tool (NetPyNE) showing a simplified version of the motor cortex model. The tool is hosted on Google Cloud Kubernetes and is freely available to all research students who can create, simulate and analyze brain circuit models online.

The simulations then provided insights into the molecular, cellular, and network mechanisms driving movement and perception, which can help develop new treatments for brain disorders. The team also deployed a self-scaling, multi-user Google Cloud Kubernetes cluster that provided the wider scientific community with an easy-to-use graphical user interface-based tool to develop their own brain circuit models in line: www.netpyne.org.

For more information on the E-CAS project, visit www.internet2.edu/ecas.

This material is based on work supported by the National Science Foundation under grant number 1904444.

About Internet2

Internet2 is a non-profit, member-driven, cutting-edge technology community founded by the nation’s leading institutions of higher education in 1996. Internet2 serves 323 U.S. universities, 60 government agencies, 43 regional and state education networks and, through them, supports more than 100,000 community anchor institutions. , more than 1,000 InCommon participants and 56 leading companies working with our community, and 70 national research and education network partners who represent more than 100 countries.


Source: Sara Aly, Interent2

Share.

Comments are closed.