Skip to content

Long-Term Support for NVIDIA DOCA 1.5

Today, NVIDIA announced the long-term support (LTS) release of NVIDIA DOCA 1.5. NVIDIA DOCA is the open cloud SDK and acceleration framework for NVIDIA…

Today, NVIDIA announced the long-term support (LTS) release of NVIDIA DOCA 1.5.

NVIDIA DOCA is the open cloud SDK and acceleration framework for NVIDIA BlueField DPUs. It unlocks data center innovation by enabling you to rapidly create applications and services for BlueField DPUs by using industry-standard APIs.

The new NVIDIA DOCA 1.5 release includes several important platform updates, making this an LTS release due to the stability and robustness of the code base. In addition, NVIDIA DOCA now supports NVIDIA ConnectX SmartNICs to simplify the transition from NIC/SmartNIC to the NVIDIA BlueField DPU.

New NVIDIA DOCA 1.5 features focus on adding advanced programmability, security, and functionality to support new storage use cases.

NVIDIA DOCA 1.5 software

Figure 1. NVIDIA DOCA Software 1.5 architecture

Highlights of the NVIDIA DOCA 1.5 release include the following:

  • Platform updates
    • Long-term support (LTS) version for BlueFIeld-2
    • Support for ConnectX SmartNICs (ConnectX-6/7 family)
  • Advanced programmability
    • NVIDIA DOCA FLOW, which is a superset of functionality compared to DPDK
  • New storage use cases:
    • SHA2 library for hash operations and crypto acceleration
    • Compression and decompression library

The NVIDIA commitment to forward and backward compatibility ensures that applications developed with NVIDIA DOCA will run seamlessly on future versions of the BlueField DPU. They can take advantage of future hardware upgrades for higher performance and increased scale.

For more information, see the full release notes.

The adoption of NVIDIA DOCA has been driven by the delivery of substantial performance gains and through the option of a dual development approach through either NVIDIA DOCA drivers or libraries.

  • NVIDIA DOCA drivers provide customization for experienced developers.
  • NVIDIA DOCA libraries give the best per-use case performance and scale, for those looking for lower coding complexity.

NVIDIA DOCA services​ and performance enhancements

This release adds live migration​ support for VirtIO-blk and support for transitional virtio-blk emulation devices, with the ability to support a mix of both Virtio0.95 and Virtio1.0 devices simultaneously.

Platform security and cryptography acceleration​

Other additions include AppShield for ransomware inspection and a regular expression (regex) library with reference applications to enable the security matching of repeated code and text patterns.

A TPM firmware trusted application is designed to support the deployment of sensitive applications on the Arm TrustZone. This adds an additional level of security that enables the use of hardware keys to authenticate and encrypt data on the DPU Arm cores.

Telemetry aggregator and logging​

NVIDIA DOCA now exposes collected telemetry data for logging and metrics. BlueField can be used to sample data on demand and log metrics for later querying by implementing Prometheus, a free software application for event monitoring.

NVIDIA DOCA FireFly: A synchronized data center

Precision timing is in the heart of the data center. NVIDIA DOCA FireFly is a timing service for the data center that supports all timing needs in one place. With Nanosecond-level clock synchronization, we can enable a new spectrum of timing and delay-critical applications.  

Improving the accuracy of data center timing represents an order of magnitude improvement as accuracy changes from milliseconds to nanoseconds with FireFly.

With a synchronized data center, you can accelerate globally synchronized data centers, AI, high-performance computing, professional media production, telco virtual network functions, and precise monitoring.  All the servers in the data center can be harmonized to provide something that is bigger than a compute node.

Storage acceleration​

Storage data compression/decompression is a CPU-intensive operation. The NVIDIA DOCA Compression library implements storage data compression and decompression onto the BlueField DPU. This offloads storage operations from the CPU to free up cycles for other compute functions and lowering server TCO. ​

ConnectX SmartNICs

With added support for NVIDIA ConnectX SmartNICs, the open NVIDIA DOCA software unlocks the benefits of the most comprehensive software APIs, libraries, and services. Developers and IT leaders can foster data center innovation on the most widely deployed, high-performance SmartNICs.

This introduces a broad range of networking, storage, and security capabilities and enhancements to deliver breakthrough performance for software partners, server and storage vendors, end users, and global system integrators. NVIDIA DOCA support for ConnectX SmartNICs helps to speed and simplify the transition from ConnectX SmartNICs to the BlueField DPU.

Open data center innovation

NVIDIA DOCA is built on open APIs such as DPDK for networking, OFED for RDMA, and SPDK for storage. It’s fully compatible with all major OS and hypervisors.  Applications written with NVIDIA DOCA run on BlueField-2 and future versions of the BlueField DPU.

DOCA Hackathon in China

The recent hackathon in China focused on BlueField DPU innovations that use the NVIDIA DOCA software framework to streamline the development process. Participants continue to find new ways to use the DPU for offloading, accelerating, and isolating a broad range of services. There were 13 teams competing over 24 hours, with four winners announced:

  • First place:  SDIC, Research on RDMA Virtualization based on the BlueField DPU
  • Second place:  Zhindex-Numa, Distributed intelligent key-value storage engine
  • Third place:  Network Needs, Key-value storage acceleration based on DPU cache
  • Fourth place:  ScaleFlash, DPU cloud-native high-speed storage

Congratulations to the winners and thanks to all the teams that participated in making this NVIDIA DOCA Hackathon such a success!

For more information about how leading companies are using NVIDIA software-defined, hardware-accelerated data center solutions to change the world, see the Modernize Your Data Center with Accelerated Networking free ebook.

More resources:

  • Demystifying NVIDIA DOCA
  • Understanding When To Use DOCA Drivers and DOCA Libraries
  • Introduction to DOCA for DPUs course (free)
  • Getting Started with DOCA Flow course (self-paced)
  • Delivering an AI-Ready Infrastructure Today for Powering the AI Factories of Tomorrow GTC session

Source:: NVIDIA