Wednesday, December 6, 2023

Marvell debuts two PAM4 Optical DSPs

 Marvell introduced two optical PAM4 digital signal processors (optical DSPs) for connecting data infrastructure:

Perseus - a 400/800 Gbps 5nm device that monolithically integrates the primary electrical components of a short-reach pluggable optical module — an optical DSP, a transimpedance amplifier (TIA) and a vertical-cavity surface-emitting laser (VCSEL) driver. Perseus, which is also available with an integrated silicon photonics driver, is optimized for both active optical cables (AOCs), which replace passive copper cables for connecting equipment within racks, and short-reach single mode and multi-mode optical interconnects for distances of five to 500 meters. Perseus is currently sampling.

Select features of Perseus include:

  • Quad / Octal 100 Gbps/channel optical PAM4 DSP
  • Support for both 400G and 800G optical module applications
  • Low power
  • Integrated TIA
  • Integrated linear driver (VCSEL and SiPho PIC)
  • CMIS compliant with advanced diagnostic features
  • Supports independent lane operation
  • Compatible with 800 Gbps QSFP-DD and OSFP optical modules for 800 Gbps applications and QSFP112 modules for 400 Gbps AOC/single-mode/multi-mode deployments  
Spica Gen2 - an 800 Gbps 5nm optical DSP optimized for longer-reach connections such as the high-bandwidth optical connections linking servers within an AI cluster or the optical connections between racks in a hyperscale data center. Spica Gen2 enables pluggable optical module manufacturers to reduce the power consumption of products to below 12 watts, a 25% savings over the previous generation of devices. Spica Gen2 is now in volume production.

Select features of Spica Gen2 include:

  • Qualified and proven 800G Spica DSP architecture for data center connectivity
  • Support for 1x800G, 2x400G, 8x100G connectivity for 51.2T, 25.6T switches, and advanced AI/ML clusters
  • Supports 100m SR8 and 500m DR8, 2Km 2xFR4 and 10Km LR8 optical modules
  • IEEE, MSA and CMIS compliant DSP that supports all 200G/400G/800G based data center optical connectivity applications
  • Low power DSP with integrated drivers supporting EML and Sipho
  • Advanced diagnostic features that provide insights into the health of the network
  • Support for bare die design for cost sensitive applications and small BGA package designs, for faster time to market
  • Compatible with 800 Gbps QSFP-DD and OSFP modules

“Accelerated computing requires accelerated infrastructure,” said Xi Wang, vice president of product marketing for Optical Connectivity at Marvell. “Perseus and Spica Gen2 are the latest examples of our strategy to enable our partners and end users to continuously improve the efficiency, performance and resiliency of their critical assets while creating a platform for new services.”

Video: Scaling Data Centers Optics to 800G, 1.6T, and Beyond

What does it take to scale up a data center to accommodate large AI clusters powered by 32,000 GPUs? Loi Nguyen, EVP and GM, Optical from Marvell explains:

  • The importance of connectivity in scaling up data centers: Not only do you need the GPUs themselves, but you also need high-speed optics and low latency switches to connect these GPUs together. A 32k GPU cluster would require around 70,000 optics connected via approximately 2,000 switches.
  • The power consumption of a large-scale data center: A 32k GPU cluster would consume about 45 megawatts of power, which is significantly more than the typical data center's 32 megawatts. This presents a major scaling challenge.
  • The role of Marvell in scaling bandwidth: The optic industry has done a good job in scaling optics, with optic bandwidth increasing by about 400% in the past 10 years. 
  • 800G optics are here. Next up, 1.6 Tbps.  And beyond?

Have a tech update that you want to brief us on? Contact!

Video: 3 Types of Data Center Pluggable Optics

Linear Pluggable Optics (LPO) has generated a lot of interest. Achyut Shah, SVP and GM, Connectivity from Marvell addresses the topic:

  • The significant growth of optical interconnects in data centers: As networks scale to accommodate accelerated compute-optimized clouds, the number of interconnects is set to increase dramatically. A majority of these will be optical interconnects, which are becoming more prevalent in modern data centers compared to a few years ago.
  • The segmentation of data center stages based on AI architecture: With the scale required in AI and considerations on high volume and power requirements, data centers are now being segmented into different stages. These stages are defined by the reach of optical interconnects, which are optimized based on the AI architecture adopted by different customers.
  • The optimization and fragmentation of optical modules: As data centers evolve, there is a trend towards optimization and fragmentation of optical modules within different parts of the data center network. This is driven by the need for faster market readiness, scalability, reliability, and lower latency. Different types of pluggable optical modules are used to achieve these goals, including full DSPs, link-optimized DSPs, and linear pluggable optics (LPO).

Join Achyut Shah as he provides insights into the future of data centers and AI architecture.

Have a tech update that you want to brief us on? Contact!

Video: The Data Movement Challenge

Ultimately, the challenge of how to move data as quickly and efficient as possible can be understood as an equation. Sandeep Bharathi, Chief Development Officer from Marvell, explains:

  • The importance of transferring large amounts of data across a distance with the highest energy efficiency. This involves different signaling standards and techniques to apply the right error correction algorithms, shaping waveforms to get the lowest latency and the lowest bit error rate.
  • The need to focus on short reach, long reach, and medium reach to surmount the data movement challenge. This involves understanding the difference between offboard communications and what can be done within the die.
  • How Marvell is focusing on when to move to advanced semiconductor process technology. This includes using a combination of chiplets, advanced packaging, and their advanced analog mix signal capabilities to meet the power targets of data centers.

Have a tech update that you want to brief us on? Contact!

Marvell intros two new OCTEON 10 DPUs

Marvell boosted its silicon portfolio with two new OCTEON 10 data processing units (DPUs) designed for data and control plane applications in routers, firewalls, 5G small cells, SD-WAN appliances, and control plane applications in top-of-rack switches and line card controllers. 

The new OCTEON 10 CN102 and CN103 DPUs contain up to eight Arm Neoverse N2 cores, delivering 3x the performance of Marvell current DPU solutions for devices while reducing power consumption by 50% to 25 watts. Achieving SPEC CPU® (2017) integer rate (SPECint®) scores of 36.5, OCTEON 10 CN102 and CN103 are able to deliver nearly 1.5 SPECint points per watt2. The chips can serve as an offload DPU for host processors or as the primary processor in devices; advanced performance per watt also enables OEMs to design fanless systems to simplify systems and further reduce cost, maintenance and power consumption.

Marvell says several of the world’s largest networking equipment manufacturers have already incorporated the OCTEON 10 CN102 into a number of product designs.

“The idea behind the Marvell OCTEON 10 DPU platform is that every market and every customer segment deserve optimized processors. With Moore’s Law becoming more challenging, optimized design methodologies will become the primary engine propelling innovation in semiconductors,” said Will Chu, senior vice president, Compute and Custom Business Unit at Marvell. 

Key Features of OCTEON 10 CN102 and CN103

  • 5nm process
  • 64-bit Arm Neoverse N2 core
  • High integration with 20 Ethernet media access control (MAC) with MACsec support
  • Inline IPsec engine to enable 50 Gbps IPsec with 50% of single N2 core
  • Hardware packet acceleration with VPP optimization
  • DDR5 up to 5600MT/s
  • PCIe 5.0 (CN103xx only)
  • 56G SerDes (CN103xx only)

T-Mobile US hits 4.3 Gbps with 5G standalone millimeter wave

T-Mobile US achieved a peak downstream rate of 4.3 Gbps using 5G standalone millimeter wave (mmWave) on its production network. 

The test, which was carried out in partnership with Ericsson and Qualcomm Technologies, aggregated eight channels of mmWave spectrum to reach download speeds topping 4.3 Gbps without relying on low-band or mid-band spectrum to anchor the connection. T-Mobile also aggregated four channels of mmWave spectrum on the uplink, reaching speeds above 420 Mbps.

“We’ve been industry leaders – rolling out the first, largest and fastest 5G standalone network across the country – and now we’re continuing to push the boundaries of wireless technology,” said Ulf Ewaldsson, President of Technology at T-Mobile. “We’ve always said we’ll use millimeter wave where it makes sense, and this test allows us to see how the spectrum can be put to use in different situations like crowded venues or to power things like fixed-wireless access when combined with 5G standalone.”

Versa Networks selected for DISA's Thunderdome

Versa Networks has been selected to provide SD-WAN, Zero Trust access and Customer Edge Security Stack (CESS) functions for Thunderdome, the Defense Information Systems Agency’s (DISA) zero trust network access (ZTNA) and application security architecture.

Thunderdome is DISA’s ZTNA architecture. DISA’s recently completed Thunderdome prototype successfully proved that commercial technologies, including Secure Access Service Edge (SASE), Customer Edge Security Stack (CESS) and Application Security Stacks, can improve both security and network performance in an existing enterprise environment. Following the successful completion of the prototype, DISA awarded a follow-on production Other Transaction Authority (OTA) agreement for Thunderdome to Booz Allen Hamilton.

SD-WAN functions delivered by Versa will provide software-defined routing across the Defense Information System Network (DISN), and Versa’s CESS functions will deliver Next-Generation Firewall (NGFW), IPS/IDS, and on-premises ZTNA (conditional access) services.

“We’re deeply thrilled to be selected as part of Thunderdome, which is pioneering the application of Zero Trust to secure network access for both remote and on-network users,” said Kelly Ahuja, CEO of Versa Networks. “We are proud that our team was able to successfully demonstrate the use of our secure SD-WAN and Customer Edge Security Stack services to help meet Thunderdome’s rigorous Zero Trust requirements. This selection further establishes Versa as a premier provider of secure networking and Zero Trust networking technology.”

Intel integrates 224Gbps SerDes with NewPhotonics' Engine

Intel and NewPhotonics announced a milestone achievement: the integration of Intel new 224Gbps electrical SerDes design with NewPhotonics' Photonics Engine, resulting in an end-to-end direct modulation electrical-to-optical link utilizing PAM4 modulation.

The electrical-to-optical link incorporates a wide bandwidth photodetector (PD) and Modulator, supporting the end-to-end 224Gbps data transfer.

"We are thrilled to have achieved the successful integration of our photonics technology with Intel most advanced 224Gbps SerDes." Yaniv Ben Haim, CEO and Founder of NewPhotonics, said, - "The end-to-end electrical-to-optical link at exceptional performance is a key innovation for products with the lowest power consumption, virtually no processing delay, and cost-effective advanced photonics-based products."

Noam Avni, Intel Vice President, added, "Our advanced SerDes designs, combined with NewPhotonics optical engine, demonstrates the power of pushing the boundaries of technology. The ability to establish an end-to-end electrical-to-optical link at 224Gbps direct modulation over 10 kilometers is a testament to our commitment to innovation and the advancement of high-speed data transmission solutions."

Oracle opens Bogotá Hyperscale Region

In partnership with Claro, Oracle opened a hyperscale cloud region in Bogotá, Colombia.

Claro is one of the host partners for the new Oracle Cloud Bogotá Region, and through its partnership with Oracle it becomes the first telecommunications operator to offer Oracle Cloud Infrastructure (OCI) services to organizations across Colombia.

The new region is Oracle's 47th public cloud region and offers Oracle Cloud's full capabilities across more than 100 OCI services and cloud applications to migrate, modernize, and innovate, including Oracle Autonomous Database, MySQL HeatWave Database Service, Oracle Container Engine for Kubernetes, Oracle Cloud VMware Solution, and AI infrastructure.

"Establishing a cloud region in Colombia is a reflection of Oracle's commitment to the country's technological development," said Germán Borromei, managing director, Oracle Colombia and Ecuador. "With the new region we are helping to accelerate the adoption of cloud solutions, democratize access to innovative technology, support the modernization of organizations, and foster their growth to help them achieve success in a competitive business environment."

"Partnering with Oracle to establish the first public cloud region in Bogotá is a major milestone in Colombia's technological evolution, as the new region will help accelerate the digital transformation of businesses nationwide – including small and medium-sized enterprises, which represent 90 percent of businesses in Colombia," said Carlos Zenteno, president, Claro Colombia. "Oracle selected our data center, certified with ICREA 5, so that organizations in Colombia and other countries like Perú, Ecuador, and Panamá gain access to innovative cloud services with low latency and high availability."

"Public cloud infrastructure, such as the new region Oracle is opening in the Colombian territory, generates impact not only on the country's economy but also brings social benefits and the generation of new opportunities," said Saúl Kattan, senior presidential advisor for digital transformation, Colombia.

CTI acquires 525 mobile towers across U.S.

CTI Towers, one of the largest private tower companies in the U.S., acquired rights for 525 towers located in Louisiana, Texas, North Carolina, South Carolina and California from Conterra Ultra Broadband Holdings. Financial terms were not disclosed.

With this acquisition, CTI Towers now has over 1800 towers that it owns, manages and/or markets throughout the United States. 

"We are proud to announce this partnership with Conterra Networks, as expanding access to wireless and broadband connectivity across the country is at the core of both of our businesses," said David Bacino, CEO of CTI Towers. "As we continue to execute upon our expansion strategy, adding towers located in these key areas of the U.S. will allow us to even better serve our carrier clients and their end customers."

Vantage Data Centers to open 16MW facility in Taipei

Vantage Data Centers plans to open a 16MW, 215,000 square foot (20,000 square meters) data center in Taipei, Taiwan. The company will open TPE1 in mid 2024. It is the company's first data center campus in Taiwan and its eighth in development or operational in APEC.

The new data center is sited on two acres (one hectare) just 35 minutes from the Taiwan Taoyuan International Airport. In alignment with the local government’s focus on sustainable energy sources, the TPE1 campus will be constructed in accordance with Vantage’s “sustainable by design” blueprint for energy-efficient operations to deliver industry-leading power usage effectiveness (PUE) and water usage effectiveness (WUE). 

The campus will be powered by Taiwan Power Company (Taipower), the local utility provider who is actively developing renewable energy and is committed to introducing low carbon, clean energy. TPE1 will deliver redundancies across all systems to ensure Vantage’s world-class uptime.

“Taipei is a prime location to serve customers across Asia as demand across the region continues to skyrocket, especially given the use of modern technologies such as AI,” said Raymond Tong, president of Vantage Data Centers, APAC. “Leveraging a skilled workforce, we look forward to opening the doors of this facility to continue building Taiwan’s digital hub and contributing to the market’s ongoing technological growth in sustainable ways. Vantage fully supports the government’s focus in adopting sustainable resources, and we are committed to reaching net zero carbon emissions across our global portfolio by 2030.”