Application Accelerator Software

NVIDIA® is all about data acceleration. With the right tool an application can effectively use server resources and reach extremely low latency and unparalleled throughput performance. Selecting the right tool for your application is the most important step to utilizing your server to the fullest.

Applications

DPDK

DPDK is a set of libraries and optimized NIC drivers for fast packet processing in user space. DPDK provides a framework and common API for high speed networking applications. NVIDIA is part of the DPDK open source community, contributing not only to the development of high performance Mellanox drivers but also by improving and expanding DPDK functionalities and use cases.

HPC-X ScalableHPC

To meet the needs of scientific research and engineering simulations, supercomputers are growing at an unrelenting rate. The NVIDIA Mellanox HPC-X® Toolkit is a comprehensive MPI and SHMEM/PGAS software suite for high performance computing environments. HPC-X provides enhancements to significantly increase the scalability and performance of message communications in the network.

NVIDIA Mellanox Rivermax

As the Media and Entertainment (M&E) live video broadcasting industry shifts from a Serial Digital Interface (SDI) to IP-based infrastructure, broadcasters are reaping such benefits as production scalability and flexibility. IP-based solutions also addresses the high bandwidth requirement of 4K UHD and 8K UHD video formats.

Messaging Accelerator

The NVIDIA Mellanox Messaging Accelerator (VMA) boosts the performance of message-based and streaming applications across a wide range of industries. These include the Financial Service market's High-Frequency Trading (HFT) platforms, Web 2.0 clusters, and more. For Media and Entertainment broadcasting and streaming (supporting SMPTE 2110-21) please refer to NVIDIA Mellanox Rivermax®.

Unified Communication X

The UCX framework is a collaboration between industry, laboratories, and academia to create an open-source production grade communication framework for data centric and high-performance applications. UCX is performance oriented for low-overheads in communication path allowing near native-level performance while enabling cross platform unified API supporting various network Host Card Adapters (HCAs) and processor technologies (x86, ARM and PowerPC). UCX is part of the HPC-X® ScalableHPC Software Toolkit.

SparkRDMA

Apache Spark is an open-source, fast and general engine for large-scale data processing. Spark provides an interface for programming entire clusters with implicit data parallelism and fault-tolerance.

Key Features

  • Data acceleration to increase performance characteristics
  • Optimize networking for faster packet processing
  • A common API framework for high speed networking
  • Removes interrupt-driven processing in the kernel

Benefits

  • Decreases the amount of data traversing the network
  • Achieve higher server utilization
  • Reduce time for scientific research and engineering simulations
  • Enhance the scalability and performance of message passing functionalities

Specifications

  Application Profile Ethernet / InfiniBand Required Hardware Server Technology
DPDK - Data Plane Development Kit DPDK exposes API for fast packet processing

High Throughput (even for small packet sizes e.g. 148 Mpps @ 64 Bytes on a single ConnectX-5 100GbE port)

Utilizes hardware offloads on the NIC
Ethernet ConnectX-3 Pro
ConnectX-4 / Lx
ConnectX-5 / Ex
x86,
Arm,
PPC
HPC-X - High Performance Computing X HPC-X is a software package for HPC applications enabling MPI, SHMEM and UPC

Significantly increase the scalability and performance of message communications in the network utilizing NVIDIA adapters, switches and NVIDIA Mellanox SHARP technology
Ethernet /InfiniBand ConnectX-3 Pro
ConnectX-4 / Lx
ConnectX-5 / Ex
x86,
Arm,
PPC,
GPU-Direct*
Rivermax - Video Streaming Rivermax is a software library API for media streaming applications

Use common off-the-shelf (COTS) servers for HD to Ultra HD flows

Enables compliance with SMPTE 2110-21 packet pacing

Reach 82 Gbps with a single CPU core
Ethernet ConnectX-5 / Ex x86
VMA - Messaging Accelerator Exposes standard UDP and TCP sockets, no application code changes necessary

For extremely low latency and high throughput applications over standard socket API
Ethernet /InfiniBand ConnectX-3 Pro
ConnectX-4 / Lx
ConnectX-5 / Ex
x86
UCX - Unified Communication X Simplified API for point-to-point messaging over RDMA

Leverages a multitude of transports and offloads supported by NVIDIA NICs
Ethernet /InfiniBand ConnectX-3 Pro
ConnectX-4 / Lx
ConnectX-5 / Ex
x86,
Arm,
PPC,
GPU-Direct
Apache Spark SparkRDMA plugin Any application using Apache Spark currently can be accelerated using the SparkRDMA plugin Ethernet /InfiniBand ConnectX-3 Pro
ConnectX-4 / Lx
ConnectX-5 / Ex
x86,
Arm,
PPC
* Note: Roadmap

Resources

Configure a Cluster

Academy Online Courses

Ready to Purchase