NEC Embraces Open Source Frameworks for SX-Aurora Vector Computing

Print Friendly, PDF & Email

In this video from ISC 2019, Dr. Erich Focht from NEC Deutschland GmbH describes how the company is embracing open source frameworks for the SX-Aurora TSUBASA Vector Supercomputer.

NEC recently opened the Vector Engine Data Acceleration Center (VEDAC) at its Silicon Valley facility. VEDAC is focused on fostering big data innovations using NEC’s emerging technologies while tapping into Silicon Valley’s rich ecosystem.

NEC X is excited about the opening of our vector engine data acceleration center,” said Robbert Emery, director of technology marketing and business development at NEC X. “The SX-Aurora AI platform will allow people familiar with the Apache Spark framework to experience firsthand how straightforward it is to use and develop applications with vector processors.”

The VEDAC Lab provides servers configured with multiple NEC SX-Aurora vector engine cards running NEC’s Frovedis middleware that is compatible with the Apache Spark Machine Learning Framework. It offers remote access to the SX-Aurora TSUBASA AI platform for qualified companies, universities and government labs; or physically by entrepreneurs. After login, the VEDAC provides a short series of tutorials to quickly familiarize users with vector processing data analysis acceleration, allowing them to experience big data processing that is an order of magnitude faster than traditional servers.

NEC SX-Aurora Vector Engine

SX-Aurora brings supercomputing performance to a broad range of applications, with its compact and cost-effective PCIe card form factor that fits in everything from workstation towers for a single card, to rack-mounted servers for multiple cards. This platform is also compatible with standard programming languages, such as C, C++ and Fortran.

Back in April, NEC and NEC X also announced today the expansion of their open-source Frovedis framework for SX-Aurora, with several new machine learning (ML) algorithms. Frovedis is an extensive and growing collection of Apache Spark- and Python-compatible algorithms that utilize the framework APIs. Frovedis now has more than 18 algorithms for ML, including linear and logarithmic regression, ALS, SVM, k-means and word2vec. The newly added graph algorithms leverage the benefits of vector processing for rapid analysis of large, complex graphs. The Frovedis framework is equipped to handle large data frames and transformations, as well as dense and sparse matrix operations that are designed to offload the CPU and perform all operations on the vector processor card.

Until now, with the existing server processing capabilities, developing complex models on graphical information for AI has consumed significant time and host processor cycles. NEC Laboratories has developed the open-source Frovedis framework over the last 10 years, initially for parallel processing in Supercomputers. Now, its efficiencies have been brought to the scalable SX-Aurora vector processor.

As part of its mission to foster big data innovation, NEC X provides support to new and existing SX-Aurora customers in the Americas region. The Frovedis libraries enhance the Spark MLlib libraries with optimal vector processing, providing a 100x speed improvement when running on the SX-Aurora vector engine. This scalable AI platform, powered by SX-Aurora’s advanced, in-memory processing, accelerates AI, ML and big data analytics with its industry leading memory bandwidth (1.2TB/s) and capacity (48GB), which processes many data elements simultaneously. The unique architecture significantly increases speeds and reduces power consumption for memory-intensive statistical ML and data frame applications, such as recommendation engines, demand/price prediction, data traffic analysis and risk/risk-mitigation analysis.

Check out our insideHPC Events Calendar