Tuesday, February 13, 2018

Google Cloud offers TensorFlow processors, delivering teraflops for ML

Cloud Tensor Processing Units (TPUs) are now available in beta on the Google Cloud Platform (GCP).

The Cloud TPUs are a family of Google-designed hardware accelerators that are optimized to speed up and scale up specific machine learning (ML) workloads programmed with TensorFlow.

Google's TPUs are built with four custom ASICs. Each Cloud TPU packs up to 180 teraflops of floating-point performance and 64 GB of high-bandwidth memory onto a single board. The boards can be connected via an ultra-fast network to form multi-petaflop ML supercomputers. These "TPU pods" will be available on GCP later this year.

The Cloud TPUs can be programmed with high-level, open source TensorFlow APIs. GCP is making a number of reference Cloud TPU model implementations available, including:

  • ResNet-50 and other popular models for image classification
  • Transformer for machine translation and language modeling
  • RetinaNet for object detection

Google adds that its cloud TPUs can also simplify the planning and management of ML computing resources. By purchasing the service, users benefit from the large-scale, tightly-integrated ML infrastructure that has been heavily optimized at Google over many years. The cloud TPUs are protected by the security mechanisms and practices that safeguard all Google Cloud services.

Google Builds a Custom ASIC for Machine Learning

Google has developed a custom ASIC for machine learning and artificial intelligence.

The Tensor Processing Unit (TPU) is tailored for TensorFlow, which is an open source software library for machine learning that was developed at Google.

In a blog posting,  Norm Jouppi, Distinguished Hardware Engineer at Google, discloses that the TPUs have already been in deployment in Google data centers for over a year, where they "deliver an order of magnitude better-optimized performance per watt for machine learning." The stealthy project to develop in-house silicon has been underway for several years.

A number of Google applications are already running on the Tensor Processing Units, including RankBrain, Street View, and the AlphaGo application that recently defeated the Go world champion, Lee Sedol.

Google plans to deliver machine learning as a service on its Google Cloud Platform by providing APIs for computer vision, speech, human language translation, etc.

https://cloudplatform.googleblog.com/2016/05/Google-supercharges-machine-learning-tasks-with-custom-chip.html


  • TensorFlow was originally developed by the Google Brain team and released under the Apache 2.0 open source license in November 2015. At the time, Google said TensorFlow can run on multiple CPUs and GPUs.

Verizon completes its first 5G call

Verizon completed its first over-the-air call on a 3GPP-compliant 5G New Radio (NR) system using licensed millimeter wave spectrum.

The call, which was completed at Nokia’s facility in Murray Hill, New Jersey earlier this month, used Nokia's 5G CloudRAN solution, which is comprised of the Nokia AirScale baseband and radio, AirFrame server, and AirScale Cloud RAN running 5G NR 3GPP-compliant software. Qualcomm provided 5G NR prototype devices in smartphone form factors.

The companies described the test as an important milestone on the road to preparing Verizon’s network for widespread implementation of commercial 5G mobile services for consumers and enterprises.

“With this first 3GPP NR standards-based connection, Verizon continues to lead the development of 5G technology,” said Ed Chan, senior vice president and chief technology architect, Corporate Network & Technology, Verizon. “By partnering with Nokia and Qualcomm to combine 5G technology with our deep millimeter wave spectrum, we’re well on the way to being the first to usher in the next era of wireless communications for customers.”

“Nokia’s 3GPP-compliant high-capacity 5G solution supports pioneering operators like Verizon in leveraging their assets to make a true difference with 5G for their customers,” said Marc Rouanne, president of Mobile Networks, Nokia. “Using the successful interoperability testing we conducted with Qualcomm as a basis, we’re now applying our standard-compliant 5G technology in this trial with Verizon to push the commercialization of 5G.”

Verizon, Qualcomm and Novatel to expedite 5G NR mmWave rollout

Verizon, Qualcomm Technologies, and Novatel Wireless, are pushing ahead with plans to expedite the rollout of 5G New Radio (NR) millimeter wave (mmWave) technology.  The companies have agreed to collaborate on over-the-air field trials based on the 5G NR Release-15 specifications being developed by 3GPP, with hopes of moving the mobile ecosystem towards faster validation and commercialization of 5G NR mmWave technologies at scale before the end of the decade.

The expedited plan call for an initial focus on 5G NR operation in 28 GHz and 39 GHz mmWave spectrum bands. The goal is to achieve robust multi-gigabit per second data rates with mobility at significantly lower latencies than today’s networks. Over-the-air trials are expected starting in 2018, that will be compliant with the first 3GPP 5G NR specification that will be part of Release 15. The trials will utilize 5G NR mmWave mobile test platforms from Qualcomm and will employ advanced 5G NR Multiple-Input Multiple-Output (MIMO) antenna technology with adaptive beamforming and beam tracking techniques.


Qualcomm shows 5G NR mmWave prototype

Qualcomm unveiled a 5G NR mmWave prototype system based on the 5G New Radio (NR) Release-15 specifications.  Peak download speeds of up to 5 Gbps are possible.

The prototype system, which operates in millimeter wave (mmWave) spectrum bands above 24 GHz, demonstrates how advanced 5G NR mmWave technologies can be utilized to deliver robust mobile broadband communications at multi-gigabit-per-second data rates in real-world mobile environments.

Qualcomm's prototype leverages the company's optimized mmWave RF Front-end design in a smartphone form-factor to test and trial real-world mmWave challenges, such as device and hand-blocking. It employs Multiple-Input Multiple-Output (MIMO) antenna technology with adaptive analog beamforming and beam tracking/steering techniques, which is required for robust and sustained mobile broadband communications in non-line-of-sight (NLOS) environments and device mobility. It also supports 800 MHz bandwidth and advanced 5G NR technologies including LDPC channel coding for data channels.

AT&T expands its push into microservices

AT&T has added Accenture to its Microservices Supplier Program, which offers a new approach to delivering business functionality for software developers.

Teams from Accenture and AT&T will develop microservices using scaled, agile and DevOps processes. The collaboration is expected to improve customer experiences, reduce support costs, and get services to market faster.

AT&T describes microservices as one of the key pillars of its next-gen network platform. In 2017, AT&T deployed more than 300 microservices, with hundreds more planned for 2018.

“Our collaboration with Accenture allows us to quickly deliver the products and services our customers need to stay competitive in today’s technology-based economy,” said Pam Parisian, president, Technology Development, AT&T. “Soon, they will be able to access a whole new level of scalability and functionality as we shift legacy business functions into flexible, software-centric capabilities.”

Steve Tibbetts, a managing director at Accenture said, “By decomposing legacy systems into suites of small, modular and independent services, we can co-develop new products more rapidly than ever before. Together with AT&T, we are able to deliver an enhanced experience for their customers.”

AT&T launches Microservices Supplier Program

AT&T is launching its microservices supplier program in partnership with IBM.  The idea is to offer a new approach to delivering business functionality for software developers. AT&T said that it is a heavy user of microservices for its internal systems.

Under this partnership, IBM will collaborate with AT&T to design, develop and deploy microservices that will transform AT&T’s business backend processes. IBM’s creation of microservices across sales, ordering, and enterprise data will enhance and modernize AT&T’s mission-critical systems. This enables higher-performing backend processing with minimal downtime while allowing the whole business to adapt quickly to changing market dynamics through the deployment of new fast and easy to access services and capabilities.

“Our software journey has focused on virtualizing our network functions to give more speed and flexibility than the traditional hardware model,” said Melissa Arnoldi, president, AT&T Technology and Operations. “We’re now looking at the 2,200+ apps in our IT system. And we’re starting to break them up into microservices to create agility, speed, and scalability that wasn’t possible before. Microservices are at the very core of our vision for the future of our network.”

Earlier this week, AT&T introduced Acumos, an artificial intelligence (AI) platform that makes it easy to build, share and deploy AI applications, built with microservices.

AT&T moves Acumos into open source as an AI-enabling platform

AT&T, in collaboration with Tech Mahindra, is backing an open source artificial intelligence (AI) platform called Acumos that can be used to build smart applications.

Acumos, which will be hosted by The Linux Foundation, makes it easy to build, share and deploy AI applications. More specifically, it provides the capability to edit, integrate, compose, package, train and deploy AI microservices. Acumos will provide a marketplace for accessing, using and enhancing those applications.

“Our goal with open sourcing the Acumos platform is to make building and deploying AI applications as easy as creating a website,” said Mazin Gilbert, vice president of Advanced Technology at AT&T Labs. “We’re collaborating with Tech Mahindra to establish an industry standard for AI in the networking space. We invite others to join us to create a global harmonization in AI and set the stage for all future AI network applications and services.”

“We’re opening up AI. We’re focusing on the telecommunication, media and technology spaces, starting with the network. The platform will be available to anyone and the more users who adopt it, the better it will get,” added Gilbert.

AT&T said the Acumos platform is built on open source technologies and can be powered by AT&T Indigo, its next-generation platform for delivering a seamless network experience.

Earlier this year, AT&T moved its ECOMP into the open source curation of The Linux Foundation under the newly launched ONAP project. With Acumos, AT&T said it was acting to place the initial framework into open source as quickly as possible.

Blue Danube intros 96-element, Multiband, Massive MIMO Antenna

Blue Danube Systems, a start-up based in Santa Clara, California, introduced its 96-element, second-generation 5G-ready Massive MIMO system.

Blue Danube's the BeamCraft 600 series supports simultaneous beamforming over multiple frequency bands and across multiple cellular standards, allowing operators to deliver increased capacity through a single compact system.

The company said its Massive MIMO can dynamically direct radio frequency (RF) beams where mobile traffic is high and without changes to existing cell sites or smartphones. The 96-element BeamCraft 600 supports up to two active mid bands using 16 software configurable beams, double the capacity of the company’s previous generation. An integrated passive low-band option is also available to further conserve antenna mount positions.  Blue Danube has deployed active solutions in 3 mid bands to date (AWS, PCS, DCS) and will trial the multi-band BeamCraft 600 in 2Q’18. Additional products supporting higher frequency bands including TDD will be available in 2H’18.

“Blue Danube’s portfolio of Massive MIMO products enables operators to flexibly customize RF energy to optimize capacity, coverage and cell edge performance.” said Mark Pinto, CEO & President of Blue Danube Systems. “Our 5G-ready architecture allows beam-based network management leading to virtualization of the network RF functions. By avoiding the rigidity imposed by other radio architectures, we are enabling a completely flexible RAN ecosystem.”


  • Blue Danube Systems is headed by Dr. Mark Pinto, who previously was was an executive vice president at Applied Materials, where he launched the company’s solar business. Earlier in his career, Dr. Pinto was CTO of the Lucent Microelectronics Group and general manager of a wireless/wireline product division.
  • The company was founded by Dr. Mihai Banu, who developed the Blue Danube concepts independently since 2006. Previously,  he was R&D director at Agere Systems, working on analog circuits, RF systems for wireless LANs and wireless circuits research, and before that was Head of the Communications Circuits Research department at Lucent Technologies.
  • Investors in Blue Danube Systems include Sequoia, Northgate, Silver Lake, and AT&T.

Wix.com is onboard with Google Cloud Platform

Wix.com, which offers website services to small businesses, has chosen G Suite productivity and collaboration apps from Google as its exclusive cloud productivity offering for its users.

Wix.com already uses the Google Cloud Platform, Google Maps API, YouTube and AdWords.

Yuval Dvir, Head of Online Partnerships from Google Cloud added, "We are delighted to be part of Wix's growth and we will continue to help SMBs get the best of Google Cloud as part of our strategic partnership with Wix."

Coriant forms IoT partnership with Prodea

Coriant announced a partnership with Prodea to deliver IoT services. The companies will leverage Coriant’s new Multi-Sided Platform Partnership Program and Prodea’s new IoT Service X-Change platform and ecosystem tp accelerate global adoption of IoT services.

Prodea’s IoT Service X-Change is a platform offering turnkey solutions and a framework for rapid IoT service creation and delivery with a comprehensive ecosystem of best-of-breed offerings including guaranteed quality of service. It was developed to solve a number of real problems in the IoT industry, such as:

1) Empowering service providers to participate in the IoT value chain beyond connectivity;
2) Providing a vehicle for efficient launch of partner-powered IoT solutions; and
3) Enabling monetizable and scalable services, rather than just technology components.

“While service providers, enterprises, and other organizations have been working to bring profitable IoT services to market, the full potential of these services has not been realized. In most cases, the challenge has been less about technology and more about optimal approaches to business models, specific verticals, and ecosystem partners,” said Homayoun Razavi, Chief Customer Officer and Executive Vice President of Global Sales & Digital Marketing, Coriant.

“We chose to partner with Prodea not only because of their technology innovation, but also because their IoT Service X-Change can form the basis for our Multi-Sided Platform Partnership Program’s distribution of IoT services from ecosystem partners to customers worldwide. We believe this brings together all the right go-to-market ingredients, enabling service providers and enterprises to accelerate monetization of IoT services.”

Twilio posts sales of $115 million, up 41% yoy

Twilio, which enables developers to embed messaging, voice, and video capabilities directly into their software applications, reported Q4 2007 revenue of $115.2 million, up 41% from the fourth quarter of 2016 and 15% sequentially from the third quarter of 2017.

There was a GAAP loss from operations of $20.2 million for the fourth quarter of 2017, compared with GAAP loss from operations of $12.8 million for the fourth quarter of 2016. Non-GAAP loss from operations was $3.9 million for the fourth quarter of 2017, compared with non-GAAP profit from operations of $0.1 million for the fourth quarter of 2016.

Twilio was serving 48,979 active customer accounts as of December 31, 2017, compared to 36,606 as of December 31, 2016.

“We are kicking off our tenth year as a company with fabulous momentum. I’m very proud of the team for our fourth quarter performance, but my excitement lies in the foundations we’ve laid for the next ten years of Twilio,” said Jeff Lawson, Twilio’s Co-Founder and Chief Executive Officer. “We are poised for a stellar year ahead, built on our relentless focus on customer success, quality, and software-fueled innovation.”

One Africa Network envisions a regional traffic exchange

The Smart Africa Alliance (SAA) and Global Voice Group (GVG) have agreed to build an African Regional Traffic Exchange and Financial Settlement (ARTEF) Platform as part of the One Africa Network (OAN) initiative.

The idea is to establish a harmonized African telecommunications framework that would give African countries the potential to make significant leaps in development through lower intra-region communication costs. For consumers, the goal would be to lower the costs of communications by eliminating roaming charges. This would boost intra-African traffic, leading to more trade and economic growth. Calls from one OAN country to another would not have to leave the OAN zone to incur international changes.

As Patrice Baker, CEO of GVG, pointed out: “The OAN is a major initiative for the socio-economic development of Africa. The idea of a free roaming zone and the necessity of establishing a harmonized African telecommunications framework has been discussed for many years across the continent. It is time now for action. As a technical partner and provider of the traffic exchange platform, GVG is strongly committed to assist SAA members in the successful implementation of the OAN.”

http://smartafrica.org
https://www.globalvoicegroup.com

See also