Features and applications of NVIDIA DOCA architecture

DOCA is to DPU what CUDA is to GPU

I believe that many people are familiar with CPU and GPU. In the past ten years, various new ultra-large-scale data centers are connected together through powerful CPU and GPU processors, and DPU (Data Processing Unit) is also becoming a data processing unit. The third computing unit of the centered accelerated computing model: CPU is used for general computing, GPU is used for accelerated computing, and DPU is used for data processing. CPU, GPU, and DPU are becoming the three pillars of future computing.

NVIDIA BlueField DPU is a new type of programmable processor that focuses on data processing and can meet the increasing demands of enterprises for performance, security, and manageability. It has a high-performance and software-programmable multi-core CPU, high-performance Network interface, flexible and programmable acceleration engine. In order to accelerate the deployment of the data center and support the majority of developers in the BlueField DPU for software development, NVIDIA has also tailored a software development kit for the DPU-DOCA.

What is the charm of DOCA?

NVIDIA DOCA is an architecture that integrates data center functions on chips, which can help developers easily program future data center infrastructure. More importantly, while DOCA provides programmers with a simple development interface, it can be smoothly downward compatible and supports each generation of BlueField DPU products.

Through BlueField DPU and DOCA, its data center can be transformed into an advanced virtual private cloud, which not only achieves acceleration, is fully programmable, safe and reliable, but also provides the scalability of public cloud.

DOCA supports BlueField DPU to generate time-stamped telemetry data for the data traffic of all servers, and to configure, update and monitor hundreds or even thousands of DPUs throughout the data center.

DOCA can offload data center infrastructure services to the BlueField DPU, thereby freeing up the core so that the CPU can be fully used for business applications.

In short, DOCA simplifies the programming journey. Developers can use APIs, libraries and sample codes to create a program framework to program accelerated applications running on the DPU, resulting in excellent efficiency and performance.

The applications and services created with DOCA will have high performance, software-defined, cloud-native and other features, and use DPU acceleration to meet the increasing performance and security requirements of modern data centers.

What problems can DOCA help developers solve?

With the powerful functions of DOCA, developers can simplify the development and deployment process of infrastructure applications using DPU, and accelerate the speed of applications entering the market.

The four magic weapons provided by DOCA for developers include:

Unified access to all DPU functions: save developers the cost of learning and using a variety of different tools.

Provide an abstraction layer to the upper-layer library on the bottom API of the DPU: developers can develop faster and easier, realize the integration with the upper-layer business, and provide excellent performance after optimization, or cooperate with the lower-layer interface to achieve better performance. Fine control.

Forward/backward compatibility: Applications developed using DOCA can run seamlessly on future versions of BlueField DPU, and get higher performance and scalability.

DPU provisioning and deployment based on containerized services: DOCA includes tools to simplify DPU setup, configuration, and service orchestration.

Why do users choose DOCA?

Since DPU and DOCA have so many advantages, are there any users who have begun to use DPU? In fact, as early as 2018, UCloud, a well-known domestic cloud service provider, began to actively explore high-performance bare metal physical cloud solutions based on NVIDIA BlueField DPU, and successfully launched bare metal physical cloud 1.0 at the end of 2020, and added it in 2021. Bare metal physical cloud 2.0 product with cloud storage function.

Since its establishment, UCloud has faced the problem of how to provide a large number of cloud computing tenants with high-throughput, low-latency physical and virtualized networks. They believe that the combination of software and hardware is the future trend. Through the cooperation of software and hardware, software definition and hardware acceleration are the best ways to truly improve system performance and security.

DOCA is a software framework tailored for DPU, with the purpose of supporting developers to develop software on BlueField DPU. UCloud simplifies the UCloud bare metal network architecture through DPU and DOCA, reconstructs the storage architecture, realizes the unification of virtualization and bare metal architecture, and improves the security of the UCloud data center.

A strong lineup of DOCA ecology

Currently, NVIDIA is cooperating with leading platform vendors and partners such as Red Hat, Palo Alto, SUSE, VMWare, Check Point, Guardicore, F5, etc., to integrate and expand the support of NVIDIA BlueField DPU and DOCA on various commercial distribution software. The DOCA-centric ecology will continue to expand the available resources for developers.