Data Center / Cloud

GTC 21: 5 Data Center Networking and Ecosystem Sessions You Shouldn’t Miss!

As NVIDIA CEO Jensen Huang stated in last year’s GTC, “the data center is the new unit of computing.” Traditional way of using the server as the unit of computing is fading away quickly. More and more applications are moving to data centers that are located at the edge, in different availability zones or in private enterprise clouds. Modern workloads such as AI/ML, edge computing, cloud-native microservices and 5G services are becoming increasingly disaggregated, distributed and data hungry. These applications demand efficient, secure and accelerated computing and data processing across all the layers of the application stack. 

Computing accelerated by NVIDIA GPUs and data processing units (DPUs) is at the heart of modern data centers. DPUs are a game-changing new technology that accelerates GPU and CPU access to data while enabling software-defined, hardware-accelerated infrastructure. With DPUs, organizations can efficiently and effectively deploy networking, cyber security, and storage in virtualized as well as containerized environments.

We at NVIDIA are on a mission to bring the next generation data center vision to reality. Join us at NVIDIA GTC’21 (Apr 12-16, 2021) to witness the data center innovation we are pioneering.  Register for the top DPU sessions at GTC to learn how NVIDIA Networking solutions are powering the next generation data centers.

Palo Alto Networks and NVIDIA present  Accelerated 5G Security: DPU-Based Acceleration of Next-Generation Firewalls [S31671]

5G offers many new capabilities such as lower latency, higher reliability, throughput, agile service deployment through cloud-native architectures, greater device density, etc. A new approach is needed to achieve L7 security at these rates with software-based firewalls. Integrating Palo Alto Network next generation firewall with the NVIDIA DPU enables industry-leading high-performance security. NVIDIA’s Bluefield-2 DPU provides a rich set of network offload engines designed to address the acceleration needs of security-focused network functions in today’s most demanding markets such as 5G and the cloud.


  • Sree Koratala, VP Product Management Mobility Security, Palo Alto Networks
  • John McDowall, Senior Distinguished Engineer, Palo Alto Networks
  • Ash Bhalgat, Senior Director, Cloud, Telco & Security Market Development, NVIDIA

China Mobile, Nokia and NVIDIA present Turbocharge Cloud-Native Applications with Virtual Data Plane Accelerated Networking [S31563]

Great progress has been made leveraging hardware to accelerate cloud networking for a wide range of cloud-based applications. This talk will examine how cloud networking for public cloud infrastructure as a service can be accelerated using the NVIDIA’s new open standards technology called virtual data plane acceleration (vDPA). In addition, this presentation will examine the early validation results and acceleration benefits of deploying NVIDIA ASAP2 vDPA technology in China Mobile’s BigCloud cloud service.


  • Sharko Cheng, Senior Network Architect, Cloud Networking Products Department, CMCC
  • Mark Iskra, Director, Nokia/Nuage Networks
  • Ash Bhalgat, Senior Director, Cloud, Telco & Security Market Development, NVIDIA

Red Hat and NVIDIA present Implementing Virtual Network Offloading Using Open Source tools on BlueField-2 [S31380]

NVIDIA and Red Hat have been working together to provide an elegant and 100% open-source solution using the BlueField DPU for hardware offloading of the software-defined networking tasks in cloud-native environments. With BlueField DPUs, we can encrypt, encapsulate, switch, and route packets right on the DPU, effectively dedicating all the server’s processing capacity to running business applications. This talk will discuss typical use cases and demonstrate the performance advantages of using BlueField’s hardware offload capabilities with Red Hat Enterprise Linux and the Red Hat OpenShift container platform.


  • Rashid Khan, Director of Networking, Red Hat
  • Rony Efraim, Networking Software and Systems Architect, NVIDIA

NVIDIA DPU team presents  Program Data Center Infrastructure Acceleration with the Release of DOCA and the Latest DPU Software [S32205]

NVIDIA is releasing the first version of DOCA, a set of libraries, SDKs, and tools for programming the NVIDIA BlueField DPU, as well as the new version 3.6 of the DPU software. Together these enable new infrastructure acceleration and management features in BlueField, simplify programming and application integration. DPU developers can offload and accelerate networking, virtualization, security, and storage features including VirtIO for NFV/VNFs, BlueField SNAP for elastic storage virtualization, regular expression matching for malware detection, and deep packet inspection to enable sophisticated routing, firewall, and load-balancing applications.


  • Ami Badani, VP Marketing, NVIDIA
  • Ariel Kit, Director of Product Marketing for Networking, NVIDIA

F5/NGINX and NVIDIA presents kTLS Hardware Offload Performance Benchmarking for NGINX Web Server [S31551]

Encrypted communication usage is steadily growing across most internet services. TLS is the leading security protocol implemented on top of TCP. kTLS (kernel TLS) serves as a unit that offers TLS operations support in the Linux kernel layer. It was introduced in kernel v4.13 as a software offload for user-space TLS libraries and was extended in kernel v4.18 with an infrastructure for performing hardware-accelerated encryption/decryption in the SmartNIC and DPUs. This session will review the life cycle of a hardware-offloaded kTLS connection and the driver-hardware interaction to support it while demonstrating and analyzing the significant performance gain by offloading kTLS operations to the hardware, using NGINX as the target workload and NVIDIA’s mlx5e driver on top of a ConnectX-6 Dx SmartNIC.


  • Damian Curry, Business Development Technical Director, F5
  • Bar Tuaf, Software engineer, NVIDIA

Register today for free and start building your schedule. 

Once you are signed in, you can explore all GTC conference topics here. Topics include areas of interest such as data center networking and virtualization, HPC, deep learning, data science, and autonomous machines, or industries including healthcare, public sector, retail, and telecommunications.

See you at the GTC’21!

Discuss (0)