Monday, December 16, 2019

Intel acquires Habana Labs for $2 billion - AI chipset

Intel has acquired Habana Labs, an Israel-based developer of programmable deep learning accelerators for the data center, for approximately $2 billion.

Habana’s Gaudi AI Training Processor is currently sampling with select hyperscale customers. Large-node training systems based on Gaudi are expected to deliver up to a 4x increase in throughput versus systems built with the equivalent number of GPUs. Gaudi is designed for efficient and flexible system scale-up and scale-out.

Additionally, Habana’s Goya AI Inference Processor, which is commercially available, has demonstrated excellent inference performance including throughput and real-time latency in a highly competitive power envelope. Gaudi for training and Goya for inference offer a rich, easy-to-program development environment to help customers deploy and differentiate their solutions as AI workloads continue to evolve with growing demands on compute, memory and connectivity.

Habana will remain an independent business unit and will continue to be led by its current management team. Habana will report to Intel’s Data Platforms Group, home to Intel’s broad portfolio of data center class AI technologies.

“This acquisition advances our AI strategy, which is to provide customers with solutions to fit every performance need – from the intelligent edge to the data center,” said Navin Shenoy, executive vice president and general manager of the Data Platforms Group at Intel. “More specifically, Habana turbo-charges our AI offerings for the data center with a high-performance training processor family and a standards-based programming environment to address evolving AI workloads.”

Habana Labs chairman Avigdor Willenz will serve as a senior adviser to the business unit as well as to Intel Corporation after Intel’s purchase of Habana.

“We have been fortunate to get to know and collaborate with Intel given its investment in Habana, and we’re thrilled to be officially joining the team,” said David Dahan, CEO of Habana. “Intel has created a world-class AI team and capability. We are excited to partner with Intel to accelerate and scale our business. Together, we will deliver our customers more AI innovation, faster.”


Interview: Habana Labs targets AI processors



Habana Labs, a start-up based in Israel with offices in Silicon Valley, emerged from stealth to unveil its first AI processor. Habana's deep learning inference processor, named Goya, is >2 orders of magnitude better in throughput & power than commonly deployed CPUs, according to the company. The company will offer a PCIe 4.0 card that incorporates a single Goya HL-1000 processor and designed to accelerate various AI inferencing workloads,...

Habana Labs, a start-up based in Tel-Aviv, Israel, raised $75 million in an oversubscribed series B funding for its development of AI processors.

Habana Labs is currently in production with its first product, a deep learning inference processor, named Goya, that is >2 orders of magnitude better in throughput & power than commonly deployed CPUs, according to the company. Habana is now offering a PCIe 4.0 card that incorporates a single Goya HL-1000 processor and designed to accelerate various AI inferencing workloads, such as image recognition, neural machine translation, sentiment analysis, recommender systems, etc.  A PCIe card based on its Goya HL-1000 processor delivers 15,000 images/second throughput on the ResNet-50 inference benchmark, with 1.3 milliseconds latency, while consuming only 100 watts of power. The Goya solution consists of a complete hardware and software stack, including a high-performance graph compiler, hundreds of kernel libraries, and tools.

Habana Labs expects to launch an training processor - codenamed Gaudi - in the second quarter of 2019.

The funding round was led by Intel Capital and joined by WRV Capital, Bessemer Venture Partners, Battery Ventures and others, including existing investors. This brings total funding to $120 million. The company was founded in 2016.

“We are fortunate to have attracted some of the world’s most professional investors, including the world’s leading semiconductor company, Intel,” said David Dahan, Chief Executive Officer of Habana Labs. “The funding will be used to execute on our product roadmap for inference and training solutions, including our next generation 7nm AI processors, to scale our sales and customer support teams, and it only increases our resolve to become the undisputed leader of the nascent AI processor market.”

“Among all AI semiconductor startups, Habana Labs is the first, and still the only one, which introduced a production-ready AI processor,” said Lip-Bu Tan, Founding Partner of WRV Capital, a leading international venture firm focusing on semiconductors and related hardware, systems, and software. “We are delighted to partner with Intel in backing Habana Labs’ products and its extraordinary team.”

https://habana.ai/

Intel ships its Nervana Neural Network Processors

Intel announced the commercial production of its Nervana Neural Network Processors (NNP) for training (NNP-T1000) and inference (NNP-I1000).

The new devices are Intel’s first purpose-built ASICs for complex deep learning for cloud and data center customers. Intel said its Nervana NNP-T strikes the right balance between computing, communication and memory, allowing near-linear, energy-efficient scaling from small clusters up to the largest pod supercomputers. Both products were developed for the AI processing needs of leading-edge AI customers like Baidu and Facebook.

Intel also revealed its next-generation Movidius Myriad Vision Processing Unit (VPU) for edge media, computer vision and inference applications. Additionally, Intel’s next-generation Intel Movidius VPU, scheduled to be available in the first half of 2020, incorporates unique, highly efficient architectural advances that are expected to deliver leading performance — more than 10 times the inference performance as the previous generation — with up to six times the power efficiency of competitor processors.

“With this next phase of AI, we’re reaching a breaking point in terms of computational hardware and memory. Purpose-built hardware like Intel Nervana NNPs and Movidius Myriad VPUs are necessary to continue the incredible progress in AI. Using more advanced forms of system-level AI will help us move from the conversion of data into information toward the transformation of information into knowledge,” stated Naveen Rao, Intel corporate vice president and general manager of the Intel Artificial Intelligence Products Group.

“We are excited to be working with Intel to deploy faster and more efficient inference compute with the Intel Nervana Neural Network Processor for inference and to extend support for our state-of-the-art deep learning compiler, Glow, to the NNP-I,” said Misha Smelyanskiy, director, AI System Co-Design at Facebook.



Cisco to acquire Exablaze for FPGA-based solutions

Cisco agreed to acquire Exablaze, a privately-held designer and manufacturer of FPGA-based devices aimed at reducing latency and improving network performance. Financial terms were not disclosed.


Exablaze, which is based in Sydney, Australia, is focused on low-latency networking, layer 1 switching, timing and time synchronization technologies, and low-latency FPGA solutions. The company was founded in 2013. Exablaze targets applications such as high-frequency trading (HFT), financial services, high-performance computing, and emerging AI/ML clusters.

https://blogs.cisco.com/news/cisco-corporate-news-december-2019
https://exablaze.com/

Brookfield acquires 130,000 towers from Reliance Jio for $3.7 billion

by James E. Carroll

Brookfield Infrastructure Partners will acquire a 100% stake in a telecom tower company in India from Reliance Industrial Investments and Holdings Limited for US$3.7 billion, of which Brookfield Infrastructure will invest approximately $375 million, with the balance being funded by its institutional partners.

The deal includes approximately 130,000 communication towers that forms the infrastructure backbone of Reliance Jio’s mobile network.

Brookfield says the majority of the towers are connected by fiber backhaul, which provides a unique platform to capitalize on the rollout of 5G and future technologies. Jio is an anchor tenant of the tower portfolio under a 30-year Master Services Agreement, providing a secure source of revenues for the tower company.

“This is a unique opportunity to invest in a large-scale, high-quality telecom business and participate in India’s high-growth data industry,” said Sam Pollock, Chief Executive Officer of Brookfield Infrastructure. “This is an attractive business that offers downside protection with meaningful upside by co-locating equipment from other Mobile Network Operators on the towers, which to-date, have only carried Jio equipment. Further growth is anticipated through a tower build-out program, which is expected to bring the portfolio to approximately 175,000 towers.”


AT&T sells 31 data centers to Brookfield Infrastructure for $1.1 billion



AT&T completed the sale of its data center colocation operations and assets to Brookfield Infrastructure and its institutional partners for $1.1 billion. This includes 18 Internet Data Centers (IDC) in the United States and 13 outside the United States. The colocation data center operations serve a diversified customer base of more than 1,000 companies. Brookfield has established a wholly owned company, Evoque Data Center Solutions ("Evoque"),.

Google Cloud announces security partnerships

Google Cloud Platform announced more than a dozen new solutions with security partners, including:

  • GCP is working with McAfee, Palo Alto Networks, and Qualys to help customers manage the deployment of agent-based endpoint security and vulnerability management solutions automatically and at scale.
  • Palo Alto Networks is expanding its usage of services on Google Cloud to jointly develop new solutions for Anthos and threat detection.
  • McAfee will integrate its MVISION Cloud solution for data security, threat prevention, governance, and compliance capabilities for container workloads with Google Cloud, as well as its endpoint security solution for Linux and Windows-based workloads.
  • Citrix Workspace is now available on Google Cloud. This includes integration with G Suite that provides a single sign-on experience, multi-factor authentication, enhanced security policies for G Suite, web filtering policies for G Suite, and end-to-end visibility and analytics. Also, users will be able to seamlessly authenticate using G Suite credentials early next year to provide simple, secure access to the apps and information they need to do their jobs anywhere, on any device.
  • Exabeam, a leading SIEM vendor, will expand its SaaS Cloud security management platform on Google Cloud, helping customers bring the scale and speed of the cloud to their existing, trusted SIEM platform.
  • ForgeRock will deliver its Digital Identity Platform on Google Cloud. ForgeRock’s platform helps customers build and maintain cloud-ready architecture to automate multi-cloud deployments. ForgeRock joins Google Cloud as a Premier Partner in the identity space and has named Google Cloud as its primary cloud provider for its cloud-native suite of identity products.
  • Fortinet will provide a new reference architecture for customers to connect facilities to Google Cloud with secure SD-WAN solutions, to make its FortiWeb Cloud WAF-as-a-Service available on Google Cloud, and to integrate its FortiCWP service with Google Cloud Security Command Center.
  • Semperis and STEALTHbits are working with GCP to enable customers to manage, audit, and protect their Microsoft Active Directory-dependent apps and workloads running on Google Cloud from service outages, data breaches, and cyberattacks.
  • Tanium will integrate its endpoint security telemetry and Chronicle’s Backstory platform with GCP to provide full visibility into endpoint events across an enterprise.

In addition, Google Cloud Platform is expanding its work with leading systems integrators and managed services providers, including: Deloitte, IBM Security, Wipro, Arctic Wolf, Comm-IT, Cyderes, and Optiv.

https://cloud.google.com/blog/products/identity-security/making-your-cloud-more-secure-through-google-cloud-partners

DE-CIX tops 8 Tbps at Frankfurt Internet Exchange

DE-CIX Frankfurt hit a new world record: more than 8.1 Terabits per second across its Internet Exchange. In September, the Frankfurt Internet Exchange handled peak rate of 7.1 Tbps for the first time.

In a timeframe of just a few weeks, data traffic at DE-CIX Frankfurt increased by almost 14 percent at peak times.

“This record figure not only underscores our position in the global market, but also that of the FrankfurtRhineMain metropolitan region as one of the most important data hubs in the world.  We are very happy with the development that we have been able to observe for many years now with regard to data throughput and the customer response to it at DE-CIX in Frankfurt. This data record can be explained purely on the basis of the increase in the number of connected networks since the beginning of the year. Here we were able to record an increase of more than 10 percent, which brings us to just under 950 connected networks. An additional factor coming into play is being in the season when Internet users particularly like to stream films or play online. All of these components contribute to such a record,” says Thomas King, Chief Technology Officer (CTO) at DE-CIX.

Generally, the data traffic at Internet Exchanges moves in waves and reflects the daily rhythm of Internet usage, beginning at 6 am and reaching its peak at around 9 in the evening. Seasonal changes – in summer people tend to be online less than in the fall and winter months – can also be registered.

https://www.de-cix.net/en/locations/germany/frankfurt/statistics

DE-CIX brings its Internet Exchange to Equinix Istanbul

DE-CIX will bring its Internet Exchange into Equinix’s IL2 International Business Exchange (IBX) data center in Istanbul, Turkey. By moving into Equinix’s IL2 data center, DE-CIX will be able to leverage Equinix’s Metro Connect services to provide direct, dedicated, carrier-grade network links between Turkey’s MedNautilus subsea cable landing station and IL2. These cable systems enable accelerated traffic globalization and data consumption, as well as rapid growth of cloud and online services.

DE-CIX currently operates more than 20 neutral Internet Exchanges in Europe, India, Southeast Asia, the Middle East and North America. The company’s Frankfurt Internet Exchange is the largest IX worldwide in terms of peak traffic – with a maximum data throughput of more than 7.8 terabits per second.

"Extending our presence in Istanbul with our long-term partner Equinix, shows we are building robust, secure and reliable interconnection bridges across continents. Istanbul itself owns a unique strategic position within Eurasia and DE-CIX is here to strengthen this position with its customers and partners. We believe Istanbul is the interconnection gateway to Europe, Asia and Middle East, with a huge future potential," states Ivo Ivanov, CEO of DE-CIX International.

IDC: 1.1 billion 5G connections in 2023

IDC is predicting that the number of 5G connections will grow from roughly 10.0 million in 2019 to 1.01 billion in 2023. By 2023, IDC expects 5G will represent 8.9% of all mobile device connections.

"While there is a lot to be excited about with 5G, and there are impressive early success stories to fuel that enthusiasm, the road to realizing the full potential of 5G beyond enhanced mobile broadband is a longer-term endeavor, with a great deal of work yet to be done on standards, regulations, and spectrum allocations," said Jason Leigh, research manager for Mobility at IDC. "Despite the fact that many of the more futuristic use cases involving 5G remain three to five years from commercial scale, mobile subscribers will be drawn to 5G for video streaming, mobile gaming, and AR/VR applications in the near term."

SpaceX launches JCSAT-18 and Kacific1 satellites

SpaceX successfully completed its JCSAT-18/Kacific1 satellite launch mission from Cape Canaveral.

The JCSAT-18/Kacific1 payload was equipped with two unique payloads.

The JCSAT-18 satellite was built by Boeing for SKY Perfect JSAT, one of the largest providers of multichannel pay TV broadcast services in Japan, which operates the largest satellite communications business in Asia. The JCSAT-18 satellite will provide Ku-band coverage and improve mobile and broadband services for SKY Perfect JSAT Corporation customers in the Asia-Pacific region,  including the far eastern part of Russia. Boeing has built 13 satellites, including two high-throughput satellites, for SKY Perfect JSAT Corporation and its predecessors since the 1980s.

Kacific1 is a next-generation geostationary satellite operating in the Ka-band frequency spectrum. Its 56 high-throughput spot beams will place capacity over selected regions in South East Asia and the Pacific Islands. Deployed to a geostationary orbital position above Asia Pacific, Kacific1 will transmit to state-of-the-art gateways, designed and built by Kratos.

Falcon 9’s first stage was successfully recovered on the “Of Course I Still Love You” droneship stationed in the Atlantic Ocean.

Telenor sticks with Ericsson, Nokia and Huawei for 5G

Telenor completed its vendor selection process for the future network in Norway. Ericsson will modernize and build the Telenor’s 5G radio access network (RAN). The core network is being upgraded by Nokia and Ericsson, as previously announced. Huawei, which is an existing RAN vendor for Telenor, was selected to maintain the 4G network and also upgrade to 5G coverage in selected areas of Norway.

Telenor is currently deploying 5G in the city of Trondheim.

Currently, the Telenor Group is testing 5G in ten locations in Norway as well as selected areas in Denmark, Sweden, Finland, Thailand and Malaysia.

https://www.telenor.com/media/press-release/telenor-completes-5g-vendor-selection-for-norway

China Mobile tests 5G with Spirent

China Mobile is using Spirent’s radio frequency (RF) channel emulation technology to help compare the performance of new 5G chipsets and consumer smartphones from leading global manufacturers.

The new YuHeng system allows operators and device manufacturers to test their RF solutions in the lab with the latest 5G base stations (gNodeB) that employ sophisticated new radio technologies such as massive MIMO beamforming. Devices under test are positioned within an RF anechoic chamber and are driven by a commercial base station, allowing the system to evaluate the performance of the device’s antenna configuration, baseband algorithms, hardware and software design.

Spirent’s Vertex RF Channel Emulator is used to replicate real-world indoor and outdoor effects that can influence radio efficiency, such as multi-path loss, reflections, absorption, delay, doppler frequency shift and antenna polarization. Both downlink and uplink propagation channels are modeled by the system, and multiple types of device-in-motion simulations can be created to replicate real-world scenarios such as walking in an urban environment or traveling in a vehicle on a highway.

“We are delighted to have had the opportunity to collaborate with the China Mobile Group Device Company team on the YuHeng project,” said Peter Tan, Spirent’s VP for Greater China & East Asia. “This innovative solution is the first of its kind and enables China Mobile to thoroughly assess and benchmark the growing number of new 5G smartphones, devices and chipsets for performance in test environments that realistically replicate a wide variety of authentic radio and motion conditions.”

See also