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.

Learn More

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.

Learn More

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.

Learn More

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^®.

Learn More

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.

Learn More

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.

Learn More

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