Monday, October 5, 2020

NVIDIA unveils DPUs and Data-Center-Infrastructure-on-a-Chip SDK

 NVIDIA has begun sampling its second-generation family of BlueField data processing unit (DPUs) for accelerating data center infrastructure.

The new BlueField-2 DPU is designed to offload critical networking, storage and security tasks from server CPUs. The company says a single chip can deliver the same data center services that could consume up to 125 CPU cores. This frees up valuable CPU cores to run a wide range of other enterprise applications.

NVIDIA’s current DPU lineup includes two PCIe products:

  • The NVIDIA BlueField-2 DPU, which features all of the capabilities of the NVIDIA Mellanox ConnectX-6 Dx SmartNIC combined with Arm cores. Fully programmable, it delivers data transfer rates of 200 Gbps and accelerates key data center security, networking and storage tasks, including isolation, root trust, key management, RDMA/RoCE, GPUDirect, elastic block storage, data compression and more.
  • The NVIDIA BlueField-2X DPU, which includes all the key features of a BlueField-2 DPU enhanced with an NVIDIA Ampere GPU’s AI capabilities that can be applied to data center security, networking and storage tasks. Drawing from NVIDIA’s third-generation Tensor Cores, it is able to use AI for real-time security analytics, including identifying abnormal traffic, which could indicate theft of confidential data, encrypted traffic analytics at line rate, host introspection to identify malicious activity, and dynamic security orchestration and automated response.

BlueField-2 DPUs are sampling now and expected to be featured in new systems from leading server manufacturers in 2021. BlueField-2X DPUs are under development and are also expected to become available in 2021.

NVIDIA says its BlueField DPUs are being adopted by leading server manufacturers worldwide, including ASUS, Atos, Dell Technologies, Fujitsu, GIGABYTE, H3C, Inspur, Lenovo, Quanta/QCT and Supermicro.

NVIDIA is developing a novel data-center-infrastructure-on-a-chip (DOCA) architecture and programming that is intended to be analogous to its CUDA environment for GPUs. A new NVIDIA DOCA software development kit enables developers to rapidly create applications and services on top of NVIDIA BlueField DPUs. he DOCA SDK provides industry-standard open APIs and frameworks, including Data Plane Development Kit (DPDK) and P4 for networking and security and the Storage Performance Development Kit (SPDK) for storage. 

  • NVIDIA is also participating in VMware's recently announced Project Monterey to integrate SmartNICs with VMware Cloud Foundation.
  • Red Hat plans to offer support for BlueField-2 DPUs with Red Hat Enterprise Linux and Red Hat OpenShift, components of Red Hat’s open hybrid cloud portfolio, which is used by 95 percent of the Fortune 500.
  • Canonical announced support of BlueField-2 DPUs and DOCA in its Ubuntu Linux platform, the most popular operating system among public clouds.
  • Check Point Software Technologies, a leading cybersecurity provider, is integrating BlueField-2 DPUs into its technologies, which more than 100,000 organizations worldwide use to protect themselves from cyberattacks.

NVIDIA envisions EDGE AI with servers under Fleet Command service

NVIDIA announced growing momentum for its EGX edge AI platform, bringing a new wave of secure, GPU-accelerated software, services and servers to enterprise and edge data centers. The company cites traction from  AI, 5G, CloudRAN, security and networking companies who are teaming with major server manufacturers, including Dell Technologies, Inspur, Lenovo and Supermicro, as well as leading software infrastructure providers, including Canonical, Cloudera, Red Hat, SUSE and VMware, to leverage the NVIDIA EGX platform in edge applications.

NVIDIA also announced that the EGX platform is expanding to combine the NVIDIA Ampere architecture GPU and BlueField-2 DPU capabilities on a single PCIe card, giving enterprises a common platform to build secure, accelerated data centers. BlueField-2 DPUs include powerful Arm CPU cores that can be used to create a trusted enclave for AI inference models running on the GPU.

“Companies around the world will come to use AI to improve virtually every aspect of their business,” said Jensen Huang, founder and CEO of NVIDIA, who unveiled the news in his keynote at the GPU Technology Conference. “To support this massive shift, NVIDIA has created an accelerated computing platform that helps companies modernize their data centers and deploy AI anywhere.”

Servers in many locations

Rather than having 10,000 servers in one location, future enterprise data centers will have one or more servers across 10,000 different locations, including inside office buildings, factories, warehouses, cell towers, schools, stores and banks.

To simplify and secure the deployment and management of AI applications and models on these servers at scale, NVIDIA will offer a new service called NVIDIA Fleet Command with a unified control plane to centralize the management of servers spread across vast areas, including factories, hospitals, retail stores and even city streets. Fleet Command promises one-touch provisioning, over-the-air software updates, remote management and detailed monitoring dashboards for AI applications.

OpenFive and AnalogX target chip-to-chip interface

OpenFive, which offers silicon-focused solutions with differentiated IP, along with AnalogX, a provider of SerDes interface IP,  announced a sub-system solution and implementation for Chip-to-Chip (C2C) interface with ultra-low latency and power.

OpenFive’s 8th generation Interlaken Controller IP is silicon-proven on multiple process nodes with tier-1 customers. OpenFive’s 8th gen. Interlaken IP supports from 1 up to 48 SerDes lanes with up to 112G SerDes rates, providing a scalable interface that offers end-to-end reliability using optional re-transmission and flow control mechanisms. For high-frequency native chip interfaces, OpenFive 8th gen Interlaken has been proven to support up to 1.6Tbps data throughput, and optional forward error correction engines (FEC) further provide bit error rate improvements.

AnalogX specializes in ultra-low power SerDes IP with a rich portfolio of interconnectivity solutions ranging from USR, XSR, VSR, to MR applications. AnalogX USR die-to-die (D2D) SerDes offers sub-pico-joule per bit (pJ/bit) and multi-terabit per mm area efficiency. AnalogX XSR/VSR chip-to-chip SerDes offers a1.5pJ/bit solution for low-energy scaling of multiple SoCs in a single PCB while providing approximately one terabit-per-mm area efficiency. AnalogX XSR/VSR configurations support up to 20 parallel lanes with very low latency, and AnalogX multi-protocol SerDes offers 2.5 pJ/bit energy efficiency for PCIe Gen 5 connectivity while offering compatibility with CCIX, CXL, and other standard protocols. AnalogX IPs are currently available across multiple foundries and 22nm, 16nm, 12nm, 7nm, and 6 nm technology nodes.

AirTrunk to build largest data center in APAC (outside China)

AirTrunk announced plans to build the biggest data centre market in Asia (excluding China) -- a new 300+ megawatt (MW) hyperscale data centre campus in Inzai, Tokyo. The initial ~60 MW phase of the campus is targeted to open in late 2021 to support anchor customer demand.

The data center is the company’s sixth in the Asia-Pacific (APAC) region, bringing its platform to a total capacity of more than 750 MW across five tier one markets.

“Japan is a highly developed market with strong international connectivity, underpinning its position as a technology and data centre hub in Asia. The rapid increase of cloud adoption in Japan will be enabled by critical infrastructure, including hyperscale data centres like TOK1,” states Robin Khuda, founder and CEO of AirTrunk.

“TOK1 is part of our ongoing commitment to deliver secure, reliable, scalable, and cost-effective infrastructure for our cloud customers in key Asia-Pacific markets. We’re ensuring operational excellence and a consistent experience for our customers across our data centre platform,” said Mr Khuda.

Japanese construction conglomerate, Daiwa House Industry Co. Ltd, has been appointed as the general contractor and will also take a stake in the project. 

Earlier this year, a consortium led by Macquarie Asia Infrastructure Fund 2 (MAIF2), a Macquarie Infrastructure and Real Assets-managed infrastructure fund, and including Public Sector Pension Investment Board (PSP Investments), acquired an 88 per cent stake in AirTrunk, valuing the company at more than $A3 billion and providing necessary capital and expertise to further realise AirTrunk’s expansion plans across APAC.

NeoPhotonics trims costs cutting and announces preliminary results

 NeoPhotonics announced preliminary results for the third quarter 2020 that incorporate cost cutting measures. As previously announced, NeoPhotonics has adopted a conservative approach to exclude future contributions from Huawei in its financial outlook, following the August 17th tightening of Department of Commerce BIS restrictions.

Revenue for Q3 is now expected to be in the range of $101 to $103 million compared to prior guidance of $95 to $105 million.

NeoPhotonics said it has taken steps to tighten production operations, account for Huawei-specific assets and inventory, consolidate Indium Phosphide production and implement an approximately 4% reduction in force. The costs to implement these changes are expected to be approximately $12.1 million, with $1.1 million in severance costs and $11.0 million in inventory and idle asset charges. The company expects to incur approximately $10.7 million of these costs in the third quarter, $0.7 million in the fourth quarter and the remainder as accelerated depreciation charges through 2021.

“Our actions better align our capacity and production infrastructure with expected demand levels, and accelerate our goal of returning to profitability,” said Tim Jenks, Chairman and CEO of NeoPhotonics. “We are maintaining our focus on developing products for next generation coherent systems and modules, wherein our silicon photonics, lasers and advanced hybrid photonic integration technologies provide the highest value, fully supporting our expansion into the data center market with coherent products. We are increasingly optimistic about our ability to drive growth both in the near-term with our 64 Gbaud solutions and in the mid-term with 96 Gbaud solutions and as our 400ZR products ramp in mid-2021. With these changes, we continue to pursue growth opportunities and deploy our best-in-class products and solutions for the highest speed over distance applications, and with a more diverse customer set,” concluded Mr. Jenks.

Cisco loses patent case with Centripetal Networks

 Cisco Systems was found to have wilfully infringed four patents held by Centripetal Networks, a threat intelligence gateway developer based in Herndon, Virginia.

The United States District Court for the Eastern District of Virginia ordered Cisco to pay $1.89 billion in damages, according to reporting by Reuters. Cisco is expected to appeal.