NVIDIA releases Magnum IO software line to help reduce data bottlenecks for data scientists and AI, HPC researchers

NVIDIA introduced NVIDIA Magnum IO, a suite of software to help data scientists and AI and high performance computing researchers process massive amounts of data in minutes, rather than hours. This offering ahs been developed in close collaboration with vendors in networking and storage, including DataDirect Networks, Excelero, IBM, Mellanox and WekaIO.

Optimized to eliminate storage and input/output bottlenecks, Magnum IO delivers up to 20 times faster data processing for multi-server, multi-GPU computing nodes when working with massive datasets to carry out complex financial analysis, climate modeling and other HPC workloads.

At the heart of Magnum IO is GPUDirect, which provides a path for data to bypass CPUs and travel on “open highways” offered by GPUs, storage and networking devices. Compatible with a wide range of communications interconnects and APIs — including NVIDIA NVLink and NCCL, as well as OpenMPI and UCX — GPUDirect is composed of peer-to-peer and RDMA elements.

Its newest element is GPUDirect Storage, which enables researchers to bypass CPUs when accessing storage and quickly access data files for simulation, analysis or visualization.

Magnum IO integrates infrastructure elements to maximize storage and network I/O performance and functionality. It optimizes IO performance by bypassing the CPU to enable direct IO between GPU memory and network storage. It also relieves CPU contention to create a more balanced GPU-accelerated system and delivers peak IO bandwidth with up to ten times fewer CPU cores, and provides optimized implementation for current and future platforms, whether the data transfers are latency sensitive, bandwidth sensitive, or collectives.

NVIDIA Magnum IO software is available, with the exception of GPUDirect Storage, which is currently available to select early-access customers. Broader release of GPUDirect Storage is planned for the first half of next year.

HPE, Cray release comprehensive HPC and AI offerings optimized for exascale computing on premises, in the cloud, or as-a-service

Hewlett Packard Enterprise (HPE) announced Thursday that it will deliver comprehensive high-performance computing (HPC) and Artificial Intelligence (AI) portfolio for the exascale era, which is characterized by explosive data growth and new converged workloads such as HPC, AI, and analytics.

The addition of Cray Inc., which HPE recently acquired, bolsters HPE’s HPC and AI solutions to now encompass an end-to-end supercomputing architecture across compute, interconnect, software, storage and services, delivered on premises, hybrid or as-a-Service. Now every enterprise can leverage the same foundational HPC technologies that power the world’s fastest systems, and integrate them into their data centers to unlock insights and fuel new discovery.

Digital transformation is driving new data-intensive workloads and real-time analytics operating at an unprecedented scale. New software, compute, interconnect, and storage capabilities are required for customers to unlock the potential of their data and accelerate innovation. 

HPE delivers solutions for any experience from single, small systems all the way to exascale-class supercomputers with tailored software, interconnect and storage capabilities. This includes solutions for modeling and simulation in weather forecasting, manufacturing and energy sectors, and AI and big data analytics in precision medicine, autonomous vehicles, geospatial imaging and financial services.

Aldec delivers comprehensive portfolio of FMC daughter cards

Aldec Inc., vendor of mixed HDL language simulation and hardware-assisted verification for ASIC and FPGA designs, expanded its  FPGA Mezzanine Card (FMC) product line to support various applications in the areas of IIoT, networking, embedded vision and automotive.

The FMC standard from ANSI/VITA has been a key enabler for extending I/O support for a range of applications.

Established in 1984, Aldec provides electronic design certification and offers a patented technology suite including RTL Design, RTL simulators, hardware-assisted verification, SoC and ASIC prototyping, design rule checking, CDC verification, IP cores, high-performance computing platforms, embedded development systems, requirements lifecycle management, DO-254 functional verification and military/aerospace solutions. 

“We now provide the most comprehensive portfolio of FMC daughter cards that can be used for FPGA development, ASIC emulation and prototyping,” said Zibi Zalewski, hardware division general manager. “Our FMC connectors are based on High Pin Count (HPC) type that is compliant with ANSI/VITA 57.1, and provides individual signaling speed of up to 25 Gb/s per link and overall bandwidth of 200 Gb/s between the daughter card and carrier card for selected configurations.”

The FMC daughter cards are compatible with TySOM FPGA development boards or HES ASIC emulation/prototyping boards.

DARPA picks teams for virtual air combat competition; to develop AI-enabled “pilots” for series of simulated aerial dogfights

DARPA has selected eight teams to compete in the AlphaDogfight Trials, a virtual competition designed to demonstrate advanced artificial intelligence (AI) algorithms that can perform simulated within-visual-range air combat maneuvering, colloquially known as a dogfight. 

The Autonomy Research Collaboration Network (ARCNet) Consortium issued a solicitation for the AlphaDogfight Trials in the summer of 2019. The eight organizations -- Aurora Flight Sciences, EpiSCI, Georgia Tech Research Institute, Heron Systems, Lockheed Martin, Perspecta Labs, physicsAI and SoarTech -- were chosen from a larger pool and are leading teams that will develop and demonstrate their dogfighting AIs in a series of three trials between November 2019 and March 2020.

The Trials aim to energize and expand a base of AI developers and potential proposers prior to an anticipated algorithm-development solicitation to be released under DARPA’s Air Combat Evolution (ACE) program. 

The ACE program will train AI in the rules of aerial dogfighting similar to how new fighter pilots are taught, starting with basic fighter maneuvers in simple, one-on-one scenarios. While highly nonlinear in behavior, dogfights have a clearly defined objective, measurable outcome, and the inherent physical limitations of aircraft dynamics, making them a good test case for advanced tactical automation. 

Like human pilot combat training, the AI performance expansion will be closely monitored by fighter instructor pilots in the autonomous aircraft, which will help co-evolve tactics with the technology. These subject matter experts will play a key role throughout the program.

Announced earlier this year, ACE seeks to automate air-to-air combat and build human trust in AI as a step toward improved human-machine teaming. DARPA’s vision is that with trusted AI able to manage lower-order operations, pilots could focus on higher-order strategic challenges such as orchestrating teams of unmanned aircraft across the battlespace under the Mosaic Warfare concept. 

The first two trials will take place November 2019 and January 2020 near Baltimore at the Johns Hopkins University Applied Physics Laboratory, which is developing and managing the simulation air environment for the AlphaDogfight Trials. During the trials, teams will pit their AIs against DARPA-provided adversarial “Red” AIs in a series of one-versus-one dogfights. The level of difficulty will increase in the second trial. 

The third and final trial will take place in two locations near Las Vegas March 2020, first at the Air Force’s innovation hub, AFWERX, and then at Nellis Air Force Base, Nevada. Teams will fly their AI algorithms against “Red” AIs as in the first two trials, but they’ll also compete against each other in a round-robin-style tournament. 

The final matchup between the top two AIs is scheduled to take place at Nellis AFB in front of a live audience of Air Force fighter pilots and the other contractor teams. The winning AI will then be matched in a simulated dogfight against a Weapons School fighter pilot while the other pilots observe and critique the live Weapons Officer in the final showdown.

Although there are no monetary prizes associated with any of the trials, all of the teams may benefit from demonstrating their technologies to potential Defense Department end users.

The AlphaDogfight Trials are related to the ACE program but are not formally part of it. Proposers to the upcoming ACE solicitation on algorithm development are not required to be part of the Trials.