Broadcom introduces Tomahawk 4, its high bandwidth Ethernet switch chip at 25.6 terabits per second

Broadcom announced that it has delivered the StrataXGS Tomahawk 4 switch series, demonstrating 25.6 terabits/sec of Ethernet switching performance in a single device— double the bandwidth of any other switch silicon currently available.

Delivered in less than two years after the previous 12.8Tbps product generation, Tomahawk 4 is implemented in 7nm technology with 512 50G PAM4 SerDes. Tomahawk 4’s high level of integration and highly efficient architecture delivers up to 75 percent lower power and cost compared to alternative solutions.

The growth of hyperscale cloud networks continues its unrelenting upward trend, with bandwidth increasing by 25 percent and storage by 36 percent year-over-year. Workloads for deep learning, using massively distributed neural networks and alternative computing nodes, require a dramatic increase in bandwidth with dense 100/200/400GbE links spanning high-radix, ultra-low-latency network topologies.

The continued trend towards disaggregated, high-density flash storage is concurrently demanding step functions in network throughput and reduced end-to-end access times. With its switching performance, high 100/200/400GbE port density, and rich feature set optimized for cloud use cases, Broadcom’s Tomahawk 4 is ideally suited to serve as the backbone for the next generation of hyperscale data center networks.

StrataXGS Tomahawk 4 Series enables next generation of high-throughput, low latency hyperscale networks with 64 ports of 400GbE switching and routing; comes with 256 ports supported on a single chip, enabling low-latency, single-hop networks for massive alternative compute cluster; and delivers connectivity using 512 instances of the industry’s highest performance and longest-reach 50G PAM4 SerDes core, enabling long-reach East-West optical links and Direct-Attached-Copper in-rack cabling in the data center.

The offering includes 25.6Tbps shared-buffer architecture, offering up to five times higher incast absorption and providing the highest performance and lowest end-to-end latency for RoCEv2 workloads; advanced load balancing mechanisms, virtually eliminating hash polarization and providing extremely efficient, controllable link utilization; and advanced congestion management, enabling new traffic management paradigms.

Broadcom also announced the introduction of Broadcom Open Network Switch APIs (OpenNSA), opening its SDK APIs for StrataXGS and StrataDNX products. Multiple open source network operating system initiatives are underway in the disaggregation ecosystem, focused on hyperscale and service provider markets. OpenNSA enables these initiatives on merchant silicon and allows the larger community to build on top of these efforts. 

OpenNSA also expands Open Compute Project efforts, like Switch Abstraction Interface (SAI), by simplifying the process of translating SAI APIs to Broadcom SDK APIs. Moreover, OpenNSA accelerates the SDN ecosystem by enhancing the toolset available for the developer community.

“The Tomahawk franchise is the flagship for cutting-edge, single-chip performance and integration among Broadcom’s multi-vectored Ethernet switch silicon portfolio, tailored to the unique and rigorous demands of hyperscale data center operators,” said Ram Velaga, senior vice president and general manager, Core Switching Group, Broadcom. “We are proud of our world-class engineering team for innovating and delivering the 25.6Tbps Tomahawk 4 chip in less than two years after we released Tomahawk 3.  Broadcom is proving yet again that customers can rely on us to lead the industry on switch silicon performance and execution at every generation.”

“Tomahawk 4 is an engineering tour de force, packing an unprecedented number of serdes into a monolithic 7nm chip,” said Bob Wheeler, principal analyst, The Linley Group. “As a result, Broadcom is once again first in delivering the next switch density, which hyperscalers need as they deploy 400G Ethernet.”

Tomahawk 4 accelerates the adoption of 100/200/400GbE Ethernet solutions at a point where optics utilizing 50G PAM4 electrical connectivity are shipping in high volumes. Products from Broadcom’s Optical Systems Division, which offers a suite of optical interconnect solutions supporting 50G PAM4 and 100G PAM4 optical data rates, complement Broadcom’s switch chips. The combination of switches and pluggable optics from Broadcom provides a complete system-level solution with plug-and-play functionality.

Rambus completes acquisition of Verimatrix silicon IP, secure protocols and provisioning business

Rambus Inc. announced on Monday that the company has completed acquisition of the silicon IP, secure protocols and provisioning business from Verimatrix, formerly Inside Secure, for US$45 million at closing, and up to an additional $20 million, subject to certain revenue targets for the transferred business for the calendar year 2020. 

The combined portfolio of products, including the secure silicon IP and provisioning solutions from both companies, creates the most comprehensive silicon-proven security IP portfolio in the industry.

“This acquisition is a natural fit for the company in line with our areas of focus in semiconductor,” said Luc Seraphin, president and CEO of Rambus. “The addition of this business from Verimatrix augments our portfolio of mission-critical embedded security products and expands our offerings for data center, AI, networking and automotive.”

Although this transaction will not materially impact 2019 financial results due to the timing of close and acquisition accounting, Rambus expects this acquisition to be accretive in 2020.

Microsoft validates Lenovo ThinkSystem SE350 edge server for Azure Stack HCI

Microsoft and Lenovo have teamed up to validate the Lenovo ThinkSystem SE350 for Microsoft's Azure Stack HCI program. The ThinkSystem SE350 was designed and built with the unique requirements of edge servers in mind. It is versatile enough to stretch the limitations of server locations, providing a variety of connectivity and security options and can be easily managed with Lenovo XClarity Controller. 

The ThinkSystem SE350 solution has a focus on smart connectivity, business security, and manageability for the harsh environment.

The ThinkSystem SE350 is the latest workhorse for the edge. Designed and built with the unique requirements for edge servers in mind, it is versatile enough to stretch the limitations of server locations, providing a variety of connectivity and security options and is easily managed with Lenovo XClarity Controller. 

The ThinkSystem SE350 is a rugged compact-sized edge solution with a focus on smart connectivity, business security, and manageability for the harsh environment.

The ThinkSystem SE350 is an Intel Xeon D processor-based server, with a 1U height, half-width and short depth case that can go anywhere. Mount it on a wall, stack it on a shelf, or install it in a rack. This rugged edge server can handle anything from 0-55°C as well as full performance in high dust and vibration environments.

Information availability is another challenging issue for users at the edge, who require insight into their operations at all times to ensure they are making the right decisions. The ThinkSystem SE350 is designed to provide several connectivity options with wired and secure wireless Wi-Fi and LTE connection ability. This purpose-built compact server is reliable for a wide variety of edge and IoT workloads.

Azure Stack HCI solutions bring together highly virtualized compute, storage, and networking on industry-standard x86 servers and components. Combining resources in the same cluster makes it easier to deploy, manage, and scale, while managing command-line automation or Windows Admin Center.

Consumers can achieve virtual machine (VM) performance for server applications with Hyper-V, the foundational hypervisor technology of the Microsoft cloud, and Storage Spaces Direct technology with built-in support for non-volatile memory express (NVMe), persistent memory, and remote direct memory access (RDMA) networking.

IDC reveals that the global Ethernet switch and router sector made marginal gains in the third quarter this year

Research firm IDC released Friday data on the global Ethernet switch market (Layer 2/3) that recorded US$7.32 billion in revenue in the third quarter of 2019, an increase of 0.1 percent year over year. Meanwhile, worldwide total enterprise and service provider (SP) router market revenues grew 0.8 percent year over year in the third quarter this year to $3.74 billion. These results were published Friday in the International Data Corporation (IDC) Worldwide Quarterly Ethernet Switch Tracker and Worldwide Quarterly Router Tracker. 

The Worldwide Quarterly Ethernet Switch Tracker and the Worldwide Quarterly Router Tracker provide total market size and vendor shares for the Ethernet switch and router technologies in an Excel pivot table format. The geographic coverage for both the Ethernet switch market and the router market includes eight major regions (USA, Canada, Latin America, Asia/Pacific (excluding Japan), Japan, Western Europe, Central and Eastern Europe, and Middle East and Africa) and 60 countries. 

The Ethernet switch market is further segmented by speed (100Mb, 1000Mb, 10Gb, 25Gb, 40Gb, 50Gb, 100Gb), product (fixed managed, fixed unmanaged, modular), and layer (L2, L3, ADC). Measurement for the Ethernet switch market is provided in vendor revenue, value, and port shipments. The router market is further split by product (high-end, mid-range, low-end, SOHO), deployment (service provider, enterprise), connectivity (core, edge), and the measurements are in vendor revenue, value, and unit shipments.

From a geographic perspective, the third quarter of this year Ethernet switch market delivered mixed results across the globe. The Middle East and Africa region grew 9.3 percent with the region's largest market, the United Arab Emirates, growing 6.9 percent year over year while Israel's market grew 18.8 percent. Across Europe, growth was stagnant. 

The Central and Eastern Europe (CEE) region lost 9.0 percent compared to a year earlier with Russia dropping 13.9 percent year over year. The Western Europe market fell 6.1 percent with Germany losing 5.6 percent year over year. Denmark was a bright spot in the region with 18.6 percent year-over-year growth. 

The Asia/Pacific (excluding Japan) (APeJ) region grew 1.3 percent year over year. In China, the market grew 4.6 percent year over year while the Philippines rose 25.9 percent. Japan was off 3.8 percent compared to its growth in the third quarter last year. In Latin America, the market dropped 4.4 percent year over year, while the U.S. market grew 2.6 percent annually and Canada declined 4.9 percent year over year. 

"The over-arching trend driving both the Ethernet switch and router markets continues to be enterprise, communications service provider, and cloud SP investments in agile cloud connectivity to meet the needs of their business and customers," said Rohit Mehra, vice president, Network Infrastructure at IDC. "The demand for higher speeds across a range of network connectivity options will spur growth in these markets in the coming years as bandwidth needs increase and mission-critical applications demand even lower latencies." 

Growth in the Ethernet switch market continues to be driven by the highest-speed switching platforms. For example, port shipments for 100Gb switches rose 57.2 percent year over year to 5.6 million. 100Gb revenues grew 32.8 percent year over year in the third quarter this year to $1.44 billion, making up 19.6 percent of total market revenue compared to 14.8 percent a year earlier. 

25Gb switches also saw growth with revenues increasing 69.3 percent to $463.4 million and port shipments growing 68.0 percent year over year. Lower-speed campus switches, a more mature part of the market, saw moderate growth in port shipments but declining revenue, pointing to a decrease in average selling prices (ASPs). 

10Gb port shipments rose 7.9 percent year over year, but revenue declined 8.7 percent to deliver 26.3 percent of total market revenue. 1Gb switches grew 3.9 percent year over year in port shipments but declined 7.4 percent in revenue. 1Gb now accounts for 39.2 percent of the total Ethernet switch market revenue. 

The worldwide enterprise and service provider router market increased 0.8 percent on a year-over-year basis in the third quarter this year with the major service provider segment, which accounts for 75.5 percent of revenues, decreasing 0.1 percent and the enterprise segment of the market growing 3.6 percent. 

From a regional perspective, the combined service provider and enterprise router market fell 1.5 percent in APeJ with the enterprise segment growing 1.5 percent and the service provider segment declining 2.2 percent year over year. Japan's total market grew 3.9 percent year over year. 

Revenues in Western Europe were off 7.3 percent year over year while CEE revenues for the combined enterprise and service provider market grew 3.5 percent year over year. The Middle East & Africa region was up 8.9 percent. In the U.S., the enterprise segment was up 7.3 percent while service provider revenues grew 1.8 percent, giving the combined markets 3.2 percent year-over-year growth. In Latin America, the market grew 11.7 percent. 

Cisco finished the thrid quarter this year with a 5.6 percent year-over-year decline in overall Ethernet switch revenues and market share of 51.3 percent. In the hotly contested 25Gb/100Gb segment, Cisco was the market leader with 38.2 percent revenue share. Cisco's combined service provider and enterprise router revenue declined 10.6 percent year over year with enterprise router revenue increasing 3.4 percent and SP revenues falling 18.1 percent. Cisco's combined SP and enterprise router market share increased to 37.9 percent, up from 36.8 percent in the second quarter of this year. 

Huawei's Ethernet switch revenue rose 4.4 percent on an annualized basis, giving the company market share of 8.9 percent. The company's combined SP and enterprise router revenue rose 20.6 percent year over year, giving the company a market share of 28.1 percent. 

Arista Networks saw Ethernet switch revenues increase 14.3 percent in the third quarter, bringing its share to 7.6 percent of the total market, up from 6.6 percent a year earlier. 100Gb revenues accounting for 68.8 percent of the company's total revenue, indicating the company's focus on hyperscale and cloud providers. 

HPE's Ethernet switch revenue declined 7.0 percent year over year, resulting in overall market share of 5.3 percent. 

Juniper's Ethernet switch revenue rose 4.5 percent year over year in the third quarter this year, bringing its market share to 3.2 percent. Juniper saw a 17.9 percent decline in combined enterprise and SP router sales, bringing its market share in the router market to 10.9 percent. 

"Dynamics in the Ethernet switch and routing markets continue to evolve," said Petr Jirovsky, research director, IDC Networking Trackers. "In Ethernet switching, the seemingly insatiable demand for higher-speed networking platforms continue to drive investments. Meanwhile, lower-speed campus switching continues to moderate as enterprises build out connectivity platforms to support mobile workers and the Internet of Things. In the routing segment, the enterprise continues to buoy the broader market as enterprises and service providers augment their networks to support improved cloud connectivity."

Richweb boosts rural broadband with hyperscale carrier-grade network address translation from A10 Networks

A10 Networks announced that Richweb, a Virginia-based managed service provider, has adopted the company’s bare metal carrier-grade network address translation (CGNAT) solution. Richweb is helping to bring digital equity to rural communities across Virginia. 

As part of its infrastructure services portfolio expansion, Richweb aspired to provide CGNAT services to help its electric co-op customers bring high-speed broadband to more Virginia residents. It chose A10 Thunder Carrier-grade Networking (CGN) Bare Metal to extend its customers’ IPv4 address pools and ensure that critical network services are always available and reliable.

A10 Networks’ Thunder CGN solution provides high-performance, highly transparent network address and protocol translation, allowing service providers and enterprises to extend IPv4 network connectivity while simultaneously transitioning to the IPv6 protocol. In addition to hyperscale and rich application-level address translation features, the Thunder CGN solution also provides increased security to protect the services and infrastructure.

A10 Thunder CGN Bare Metal leverages the same high-performance architecture and features of the A10 Thunder CGN appliances. Bare-metal software deployments benefit from greater performance by not using a hypervisor, since they have direct access to the underlying hardware. A single instance of A10 Thunder CGN Bare Metal offers up to 40 Gbps of throughput.

“The performance of A10 Thunder CGN Bare Metal doesn’t waver,” says Jon Larsen, CIO and co-founder of Richweb, noting that open-source NAT products can experience unpredictable performance, impacting the subscriber experience.

Richweb deployed A10 Thunder CGN Bare Metal on its preferred server platform, which lowered capital costs and simplified operations. Servers running the bare-metal software can be upgraded independently from the hardware, giving the engineering team greater flexibility as business expands. 

“I like the bare metal because we can throw more cores at it and we can scale on-demand,” says Larsen.

Richweb’s engineering and operations team was new to A10 Thunder CGN, but deployment was smooth. “I expected it to be more of a challenge to configure Thunder CGN Bare Metal,” says Larsen. “It was simple. Thunder CGN just worked.”

“A10 Thunder CGN allows Richweb to be at the center of rural broadband services. Richweb provides the infrastructure services, such as routing, network address translation and peering, that electric co-ops need to deliver broadband to rural communities,” said Mark Lea, CEO and co-founder, Richweb.

High-performance, scalable CGNAT is part of a sustainable growth strategy for any service provider or enterprise. With CGNAT services, Richweb can make the most of their IPv4 address allocations, enabling them to serve more subscribers and more connected devices, while being ready for IPv6.

AutoStore delivers flexible, high-density storage solution enabling retailers to deliver products to consumers in record time

AutoStore, vendor of high-density storage with an existing robotic cube-based, automated storage and retrieval system that has transformed the logistics and delivery process for businesses across the globe. The system has evolved to accommodate the increasing demands on retailers to change and stock inventory at speed and deliver goods to consumers within same-day or several hour timeframes.

Consumers are demanding more choices and are more willing to change brand allegiance to get better value and service. To remain competitive, retailers, whether big box or independent, require a presence across multiple channels – stores, online, social networking and mobile. Same-day delivery and immediate local-store pick-up are no longer perks for online shoppers, they are expectations.

Micro-fulfilment presents one of the largest challenges for business owners and retailers to not only store ever-changing inventory within a warehouse, but to ensure the warehouses are close enough to consumers. Managing fulfilment of consumer goods and groceries from regional distribution centres does not accommodate customised needs of consumers, nor can it sustain cost efficiency.

AutoStore provides a solution for the future of logistics and warehousing with micro-fulfilment at the retail store level. The pioneering Cube Storage Automation system offers the densest product/inventory storage solution in existence, where up to 15,000 SKUs can be stored within 604 sq meters. 

Modular in design, the system can be configured to fit different store ceiling heights and obstacles and suits a range of environments – from back-of-house of a grocery retailer to retail spaces for consumer products.

The AutoStore system contains a three-dimensional aluminum grid structure with self-supporting crates that are moved to pick stations by independently operating robots, providing swift and accurate movement of orders. Each robot has two sets of wheels that enable it to move along perpendicular axes, and this makes it possible for all robots to reach any position, and any crate on the grid independently – providing near 100 percent system availability.

For customers who would prefer to collect their items, AutoStore's pick-up points play a critical role in maintaining high-volume throughout for output workstations. Online orders can be consolidated in less than 10 minutes and confirmed in the system for customers to collect at any time. Orders can be retrieved from the system in just a few minutes, and customers can choose their preferred option to collect pre-picked orders, whether in-store or for home delivery.

“AutoStore delivers the flexibility and scalability needed for streamlining automated fulfillment of online orders, which is absolutely critical to accommodate increasingly higher numbers of SKUs, and seasonal spikes in ordering," said Karl Johan Lier, CEO and president of AutoStore. "It gives retailers the flexibility to make more targeted decisions about where they believe automated micro-fulfilment will be most beneficial."

Supermicro aligns with Intel to offer large scale distributed training AI systems to meet needs of deep learning training models

Super Micro Computer is collaborating with Intel on AI solutions that are validated on the Intel Nervana Neural Network Processor for Training (NNP-T). Intel's NNP-T is a purpose-built AI training ASIC supporting the growing compute needs of deep learning training models.

The Intel Nervana NNP-T solves memory constraints and is designed to scale out through systems with racks easier than today's solutions. As part of the validation process, Supermicro integrated 8 NNP-T processors, dual 2nd Generation Intel Xeon scalable processors, up to 6TB DDR4 memory per node supporting both PCIe card and OAM form factors. Supermicro NNP-T systems are expected to be available mid-year 2020.

With high compute utilization and high-efficiency memory architecture for complex deep learning models, the Supermicro NNP-T AI System is built to validate two key real-world considerations: accelerating the time to train ever-complex AI models and doing it within a given power budget. The system enables faster AI model training with images and speech, more efficient gas & oil exploration, more accurate medical image analytics, and faster autonomous driving model generation.

"Supermicro is excited to cooperate with Intel for the Nervana NNP-T platform," said Charles Liang, president, and CEO of Supermicro. "Striking a balance among computing, communication, and memory, the validated NNP-T ASICs on Supermicro systems can train large AI models with near-linear scaling efficiency via intra- and inter-chassis links."

"Supermicro has validated our Deep Learning (DL) solution and is helping us prove the Nervana NNP-T system architecture, including card and server design, interconnect, and rack," said Naveen Rao, corporate vice president and general manager, Artificial Intelligence Products Group, Intel.