HPC Wire

Subscribe to HPC Wire feed
Since 1987 - Covering the Fastest Computers in the World and the People Who Run Them
Updated: 5 hours 7 min ago

Raytheon Developing Superconducting Computing Technology for Intelligence Community

Tue, 12/05/2017 - 12:21

CAMBRIDGE, Mass., Dec. 5, 2017 — A Raytheon BBN Technologies-led team is developing prototype cryogenic memory arrays and a scalable control architecture under an award from the Intelligence Advanced Research Projects Activity Cryogenic Computing Complexity program.

The team recently demonstrated an energy-efficient superconducting/ferromagnetic memory cell—the first integration of a superconducting switch controlling a cryogenic memory element.

“This research could generate a new approach to supercomputing that is more efficient, faster, less expensive, and requires a smaller footprint,” said Zachary Dutton, Ph.D. and manager of the quantum technologies division at Raytheon BBN Technologies.

Raytheon BBN is the prime contractor leading a team that includes:

  • Massachusetts Institute of Technology
  • New York University
  • Cornell University
  • University of Rochester
  • University of Stellenbosch
  • HYPRES, Inc.
  • Canon U.S.A, Inc.,
  • Spin Transfer Technologies, Inc.

Raytheon BBN Technologies is a wholly owned subsidiary of Raytheon Company (NYSE: RTN).

About Raytheon 

Raytheon Company, with 2016 sales of $24 billion and 63,000 employees, is a technology and innovation leader specializing in defense, civil government and cybersecurity solutions. With a history of innovation spanning 95 years, Raytheon provides state-of-the-art electronics, mission systems integration, C5ITM products and services, sensing, effects, and mission support for customers in more than 80 countries. Raytheon is headquartered in Waltham, Massachusetts.

Source: Raytheon

The post Raytheon Developing Superconducting Computing Technology for Intelligence Community appeared first on HPCwire.

Cavium Partners with IBM for Next Generation Platforms by Joining OpenCAPI

Tue, 12/05/2017 - 12:12

SAN JOSE, Calif., Dec. 5, 2017 — Cavium, Inc. (NASDAQ: CAVM), a leading provider of semiconductor products that enable secure and intelligent processing for enterprise, data center, wired and wireless networking is partnering with IBM for next generation platforms by joining OpenCAPI, an initiative founded by IBM, Google, AMD and others. OpenCAPI provides high-bandwidth, low latency interface optimized to connect accelerators, IO devices and memory to CPUs. With this announcement Cavium plans to bring its leadership in server IO and security offloads to next generation platforms that support the OpenCAPI interface.

Traditional system architectures are becoming a bottleneck for new classes of data-centric applications that require faster access to peripheral resources like memory, I/O and accelerators. For the efficient deployment and success of such applications, it is imperative to put the compute power closer to the data. OpenCAPI, a mature and complete specification enables such a server design, that can increase datacenter server performance by several times, enabling corporate and cloud data centers to speed up big data, machine learning, analytics, and other emerging workloads. Capable of 25Gbits per second data rate, OpenCAPI delivers the best in class performance, enabling the maximum utilization of high speed I/O devices like Cavium Fibre Channel adapters, low latency Ethernet NICs, programmable SmartNIC and security solutions.

Cavium delivers the industry’s most comprehensive family of I/O adapters and network acclerators which have the potential to be seamlessly inegrated into OpenCAPI based systems. Cavium’s portfolio includes FastLinQ® Ethernet Adapters, Converged Networking Adapters, LiquidIO SmartNICs, Fibre Channel Adapters and NITROX® Security Accelerators that cover the entire spectrum for data-centric application connectivity, offload and accleration requirements.

“We welcome Cavium to the OpenCAPI consortium to fuel innovation for today’s data-intensive cognitive workloads,” said Bob Picciano, Senior Vice President, IBM Cognitive Systems. “Together, we will tap into Cavium’s next-generation technology, including networking and accelerators, and work in tandem with other partners’ systems technology to unleash high-performance capabilities for our clients’ data center workloads.”

“We are excited to be a part of the OpenCAPI consortium. As our partnership with IBM continues to grow, we see more synergies in high speed communication and Artificial Intelligence applications,” said Syed Ali, founder and CEO of Cavium.  “We look forward to working with IBM to enable exponential performance gains for these applications.”

About Cavium

Cavium, Inc. (NASDAQ: CAVM), offers a broad portfolio of infrastructure solutions for compute, security, storage, switching, connectivity and baseband processing. Cavium’s highly integrated multi-core SoC products deliver software compatible solutions across low to high performance points enabling secure and intelligent functionality in Enterprise, Data Center and Service Provider Equipment. Cavium processors and solutions are supported by an extensive ecosystem of operating systems, tools, application stacks, hardware reference designs and other products. Cavium is headquartered in San Jose, CA with design centers in California, Massachusetts, India, Israel, China and Taiwan.

Source: Cavium

The post Cavium Partners with IBM for Next Generation Platforms by Joining OpenCAPI appeared first on HPCwire.

IBM Unveils Power9 Server Designed for HPC, AI

Tue, 12/05/2017 - 11:38

ARMONK, NY, Dec. 5 2017 — IBM today unveiled its next-generation Power Systems Servers incorporating its newly designed POWER9 processor. Built specifically for compute-intensive AI workloads, the new POWER9 systems are capable of improving the training times of deep learning frameworks by nearly 4x[2] allowing enterprises to build more accurate AI applications, faster.

The system was designed to drive demonstrable performance improvements across popular AI frameworks such as Chainer, TensorFlow and Caffe, as well as accelerated databases such as Kinetica.

As a result, data scientists can build applications faster, ranging from deep learning insights in scientific research, real-time fraud detection and credit risk analysis.

POWER9 is at the heart of the soon-to-be most powerful data-intensive supercomputers in the world, the U.S. Department of Energy’s “Summit”and “Sierra” supercomputers, and has been tapped by Google.

“Google is excited about IBM’s progress in the development of the latest POWER technology,” said Bart Sano, VP of Google Platforms “The POWER9 OpenCAPI Bus and large memory capabilities allow for further opportunities for innovation in Google data centers.”

“We’ve built a game-changing powerhouse for AI and cognitive workloads,” said Bob Picciano, SVP of IBM Cognitive Systems. “In addition to arming the world’s most powerful supercomputers, IBM POWER9 Systems is designed to enable enterprises around the world to scale unprecedented insights, driving scientific discovery enabling transformational business outcomes across every industry.”

Accelerating the Future with POWER9

Deep learning is a fast growing machine learning method that extracts information by crunching through millions of processes and data to detect and rank the most important aspects of the data.

To meet these growing industry demands, four years ago IBM set out to design the POWER9 chip on a blank sheet to build a new architecture to manage free-flowing data, streaming sensors and algorithms for data-intensive AI and deep learning workloads on Linux.

IBM is the only vendor that can provide enterprises with an infrastructure that incorporates cutting-edge hardware and software with the latest open-source innovations.

With PowerAI, IBM has optimized and simplified the deployment of deep learning frameworks and libraries on the Power architecture with acceleration, allowing data scientists to be up and running in minutes.

IBM Research is developing a wide array of technologies for the Power architecture. IBM researchers have already cut deep learning times from days to hours with the PowerAI Distributed Deep Learning toolkit.

Building an Open Ecosystem to Fuel Innovation

The era of AI demands more than tremendous processing power and unprecedented speed; it also demands an open ecosystem of innovative companies delivering technologies and tools. IBM serves as a catalyst for innovation to thrive, fueling an open, fast-growing community of more than 300 OpenPOWER Foundation and OpenCAPI Consortium members.

Learn more about POWER9 and the AC922: http://ibm.biz/BdjCQQ

Read more from Bob Picciano, Senior Vice President, IBM Cognitive Systems:  https://www.ibm.com/blogs/think/2017/12/accelerating-ai/

[1]  Results of 3.7X are based IBM Internal Measurements running 1000 iterations of Enlarged GoogleNet model (mini-batch size=5)  on Enlarged Imagenet Dataset (2560×2560). Hardware: Power AC922; 40 cores (2 x 20c chips), POWER9 with NVLink 2.0; 2.25 GHz, 1024 GB memory, 4xTesla V100 GPU; Red Hat Enterprise Linux 7.4 for Power Little Endian (POWER9) with CUDA 9.1/ CUDNN 7;. Competitive stack: 2x Xeon E5-2640 v4; 20 cores (2 x 10c chips) /  40 threads; Intel Xeon E5-2640 v4;  2.4 GHz; 1024 GB memory, 4xTesla V100 GPU, Ubuntu 16.04. with CUDA .9.0/ CUDNN 7 Software: Chainverv3 /LMS/Out of Core with patches found at https://github.com/cupy/cupy/pull/694 and https://github.com/chainer/chainer/pull/3762

[2] Results of 3.8X are based IBM Internal Measurements running 1000 iterations of Enlarged GoogleNet model (mini-batch size=5) on Enlarged Imagenet Dataset (2240×2240). Power AC922; 40 cores (2 x 20c chips), POWER9 with NVLink 2.0; 2.25 GHz, 1024 GB memory, 4xTesla V100 GPU ; Red Hat Enterprise Linux 7.4 for Power Little Endian (POWER9) with CUDA 9.1/ CUDNN 7;. Competitive stack: 2x Xeon E5-2640 v4; 20 cores (2 x 10c chips) /  40 threads; Intel Xeon E5-2640 v4;  2.4 GHz; 1024 GB memory, 4xTesla V100 GPU, Ubuntu 16.04. with CUDA .9.0/ CUDNN 7.  Software: IBM Caffe with LMS Source code https://github.com/ibmsoe/caffe/tree/master-lms

[3] x86 PCI Express 3.0 (x16) peak transfer rate is 15.75 GB/sec = 16 lanes X 1GB/sec/lane x 128 bit/130 bit encoding.

[4] POWER9 and next-generation NVIDIA NVLink peak transfer rate is 150 GB/sec = 48 lanes x 3.2265625 GB/sec x 64 bit/66 bit encoding.

Source: IBM

The post IBM Unveils Power9 Server Designed for HPC, AI appeared first on HPCwire.

Azure Debuts AMD EPYC Instances for Storage Optimized Workloads

Tue, 12/05/2017 - 11:09

AMD’s return to the data center received a boost today when Microsoft Azure announced introduction of instances based on AMD’s EPYC microprocessors. The new instances – Lv2-Series of Virtual Machine – use the EPYC 7551 processor. Adoption of EPYC by a major cloud provider adds weight to AMD’s argument that it has returned to the data center with a long-term commitment and product roadmap. AMD had been absent from that segment for a number of years.

Writing in a blog, Corey Sanders director of compute, Azure, said, “We’ve worked closely with AMD to develop the next generation of storage optimized VMs called Lv2-Series, powered by AMD’s EPYC processors. The Lv2-Series is designed to support customers with demanding workloads like MongoDB, Cassandra, and Cloudera that are storage intensive and demand high levels of I/O.” The EPYC line was launched last June (see HPCwire article, AMD Charges Back into the Datacenter and HPC Workflows with EPYC Processor.)

The instances make use of Microsoft’s Project Olympus intended to deliver a next generation open source cloud hardware design developed with the Open Compute Community (OCP). “We think Project Olympus will be the basis for future innovation between Microsoft and AMD, and we look forward to adding more instance types in the future benefiting from the core density, memory bandwidth and I/O capabilities of AMD EPYC processors,” said Sanders, quoted in the AMD’s announcement of the new instances.

It is an important win for AMD. Gaining a foothold in the X86 landscape today probably requires adoption by hyperscalers. No doubt some “tire kicking” is going on here but use of an Olympus design adds incentive for Microsoft Azure to court customers for the instances. HPE has also announced servers using the EPYC line.

AMD EPYC chip lineup at the June launch

The Lv2-Series instances run on the AMD EPYC 7551 processor featuring a base core frequency of 2.2 GHz and a maximum single-core turbo frequency of 3.0 GHz. “With support for 128 lanes of PCIe connections per processor, AMD provides over 33 percent more connectivity than available two-socket solutions to address an unprecedented number of NVMe drives directly,” says AMD.

The Lv2 VMs will be available starting at eight and ranging to 64 vCPU sizes, with the largest size featuring direct access to 4TB of memory. These sizes will support Azure premium storage disks by default and will also support accelerated networking capabilities for the highest throughput of any cloud.

Scott Aylor, AMD corporate vice president and general manager of Enterprise Solutions said, “There is tremendous opportunity for users to tap into the capabilities we can deliver across storage and other workloads through the combination of AMD EPYC processors on Azure. We look forward to the continued close collaboration with Microsoft Azure on future instances throughout 2018.”

Link to AMD release: http://www.amd.com/en-us/press-releases/Pages/microsoft-azure-becomes-2017dec05.aspx

Link to Azure blog: https://azure.microsoft.com/en-us/blog/announcing-the-lv2-series-vms-powered-by-the-amd-epyc-processor/

The post Azure Debuts AMD EPYC Instances for Storage Optimized Workloads appeared first on HPCwire.

Bryant Departs Intel For Google Cloud

Tue, 12/05/2017 - 10:24

Google has upped its cloud game with its recruitment of Diane Bryant, the former Intel Corp.’s datacenter boss who becomes chief operating officer of Google Cloud.

Bryant, an engineer who worked her way up through the ranks of Intel during the heyday of the U.S. semiconductor industry, took a leave of absence from Intel in May that was expected to last at least six months. At the time, Intel CEO Brian Krzanich said he looked forward to Bryant’s return.

Instead, Diane Greene, CEO of Google Cloud announced late last week that Bryant would join the company as COO. Bryant “is an engineer with tremendous business focus and an outstanding thirty-year career in technology,” all of it at Intel, Greene noted in announcing the hiring.

Bryant served over the last five years as president of Intel’s Datacenter Group, expanding the chip maker’s focus on cloud computing, big data and network virtualization technologies. The group generated about $17 billion in revenue during 2016 as Intel’s x86-based servers continue to dominate datacenters.

Bryant was instrumental in guiding Intel’s transition away from the fading PC market and its unsuccessful foray into mobile devices. Prior to heading its datacenter operations, Bryant was Intel’s corporate vice president and chief information officer, where she oversaw the chip maker’s IT technology development.

Bryant is the second senior Intel executive to depart this year. In April, Brent Gorda, generation manager of Intel’s High Performance Data Division, left the company. Gorda is the former CEO of Whamcloud, the Lustre specialist acquired by Intel in 2012.

Meanwhile, the addition of Bryant gives Google Cloud another respected technology leader as it challenges cloud giant Amazon Web Services in the booming hybrid cloud market. Bryant’s “strategic acumen, technical knowledge and client focus will prove invaluable as we accelerate the scale and reach of Google Cloud.” Greene noted in a blog post.

Greene, a co-founder and CEO of VMware, took the reins at Google Cloud two years ago with the goal of extending the search giant’s mostly consumer cloud business to the enterprise market still dominated by AWS.

The post Bryant Departs Intel For Google Cloud appeared first on HPCwire.

Microsoft Spins Cycle Computing into Core Azure Product

Tue, 12/05/2017 - 09:31

Last August, cloud giant Microsoft acquired HPC cloud orchestration pioneer Cycle Computing. Since then the focus has been on integrating Cycle’s organization, mapping out its new role as a core Microsoft Azure product, and deciding what to do with those Cycle customers who currently use non-Azure cloud providers. At SC17, HPCwire caught up with Brett Tanzer, head of Microsoft Azure Specialized Compute Group (ASCG, which used to be Big Compute) in which Cycle now lives, and Tim Carroll, formerly Cycle VP of sales and ecosystem development and now a ‘principal’ in ASCG, for a snapshot of emerging plans for Cycle.

Much has already been accomplished they emphasize – for starters “the Cycle organization has settled in” and most are relocating to Seattle. Much also remains to be done – it will probably be a year or so before Cycle is deeply integrated across Azure’s extensive capabilities. In some ways, it’s best not to think of the Cycle acquisition in isolation but as part of Microsoft’s aggressively evolving strategy to make Azure all things for all users and that includes the HPC community writ large. Cycle is just one of the latest, and a significant, piece of the puzzle.

Founded in 2005 by Jason Stowe and Rob Futrick, Cycle Computing was one of the first companies to target HPC orchestration in the cloud; its software, CycleCloud, enables users to burst and manage HPC workloads (and data) into the cloud. Till now, cloud provider agnosticism has been a key Cycle value proposition. That will change but how quickly is uncertain. Tanzer assures there will be no disruption of existing Cycle customers, but also emphasizes Microsoft intends Cycle to become an Azure-only product over time. Cycle founder Stowe has taken on a new role as a solutions architect in the Specialized Compute Group. The financial details of the Cycle acquisition weren’t made public.

Far more than in the past HPC is seen as an important opportunity for the big cloud providers. The eruption of demand for running AI and deep learning workflows has also been a major driving force for cloud providers.

Nvidia V100 GPU

Microsoft, like Google and Amazon (and others), has been investing heavily in advanced scale technology. The immediate goal is to attract HPC and AI/deep learning customers. One indicator is the way they have all been loading up on GPUs. Azure is no exception and offers a growing list of GPU instances (M60, K80, P100, P40, and V100 (announced)); it also offers InfiniBand high speed interconnect. In October, Microsoft extended its high performance gambit further via a partnership with Cray to offer supercomputing in the cloud (see HPCwire article, Cray+Azure: Can Cloud Propel Supercomputing?).

How the latter bet will play out is unclear – Tanzer says, “We are hearing from customers there are some workloads they need to get into the cloud that require a Cray. And Cray itself is a pretty innovative company. We think the partnership has longer legs. Look for more to come.” One wonders what interesting offerings may sprout from that alliance.

For now the plan for Cycle is ever deeper integration with Azure’s many offerings, perhaps eventually including Cray. It’s still early days, of course. Tanzer says, “If Tim looks like he hasn’t slept much for past three months, it’s because he hasn’t.  Strategically, all of these products – Cycle, Azure Batch, HPC pack (cluster tool) – will work together and drive orchestration across all the key workloads.”

“The company is rallying behind the [HPC] category and customers are responding very well,” says Tanzer. “We are investing in all phases of the maturity curve, so if you are somebody who wants a Cray, we now have an answer for you. If you are rewriting your petrochemical workload and want to make it cloud friendly, then Batch is a great solution. We are really just taking care, wherever we can, to take friction out of using the cloud. We looked at Cycle and its fantastic people and knowledge. The relationship with Cycle is very symbiotic. We look at where our customers are and see [that for many], Cycle helps them bootstrap the process.”

It’s not hard to see why Cycle was an attractive target. Cycle brings extensive understanding of HPC workloads, key customer and ISV relationships, and a robust product. Recently it’s been working to build closer relationships with systems builders (e.g. Dell EMC) and HPC ISVs (e.g. ANSYS). From an operations and support perspective, not much has changed for Cycle customers, says Carroll, although he emphasizes having now gained access to Microsoft’s deep bench of resources. No decision has been made on name changes and Tanzer says, “Cycle is actually a pretty good name.”

Cycle’s new home, the Azure’s Specialized Compute Group seems to be a new organization encompassing what was previously Big Compute. As of this writing, there was still no Specialized Compute Group web page, but from the tone of Tanzer and Carroll it seemed that things could still be in flux. SCG seems to have a fairly broad mission to smooth the path to cloud computing across all segments with so-called “specialized needs” – that, of course, includes HPC but also crosses over into enterprise computing as well. To a significant extent, says Tanzer, it is part of Microsoft’s company-wide mantra to meet-the-customer-where-she/he-is to minimize disruption.

“Quite frankly we are finding customers, even in the HPC space, need a lot of help and it’s also an area where Microsoft has a many differentiated offerings,” Tanzer says. “You should expect us to integrate Cycle’s capabilities more natively into Azure. There is much more that can be done in the category to help customers take advantage of the cloud, from providing best practices about how your workloads move, through governance, and more. Cloud consumption is getting more sophisticated and it’s going require tools to help users maximize their efforts even though the usage models will be very different.”

One can imagine many expanded uses for Cycle functionality, not least close integration with an HPC applications and closer collaboration with ISVs to drive adoption. Microsoft has the clout and understanding of both software and infrastructure businesses to help drive that, says Carroll. “Those two things are important because this is a space that’s always struggled to figure out how to build partnerships between the infrastructure providers and software providers; Microsoft’s ability to talk to some of the significant and important ISVs and figure out ways to work with them from a Microsoft perspective is a huge benefit.”

It probably bears repeating that Tanzer’s expectations seem much broader than HPC or Cycle’s role as an enabler. He says rather matter of factly, “Customers are recognizing the cloud is the destination and thinking in more detail about that. It will be interesting to see how that plays out.” When he says customers, one gets the sense he is talking about more than just a narrow slice of the pie.

The conversation over how best to migrate and perform HPC has a long history. Today, there seems less debate about whether it can be done effectively but more around how to do it right, how much it costs, and what types of HPC jobs are best suited for being run in the cloud. Carroll has for some time argued that technology is not the issue for the most potential HPC cloud users.

Tim Carroll

“It’s less about whether somebody is technically ready than whether they have a business model that requires them to be able to move faster and leverage more compute than they had thought they were going to need,” says Carroll. “Where we see the most growth is [among users] who have deadlines and at the end of the day what they really care about is how long will it take me to get my answer and tell me the cost and complexity to get there. That’s a different conversation than we have had in this particular segment over time.”

Some customer retraining and attitude change will be necessary, says Tanzer.

“They are going to have hybrid environments for a while so to the degree we can help them reduce some of the chaos that comes from that and help retrain the workforce easily on what it needs to take advantage of the cloud. We think that’s important. Workforces who run the workloads really understand all they want to do is to take advantage of the technology but some relearning is necessary and that’s another area where Cycle really helps because of its tools and set of APIs and they speak the language of a developer,” he says.

Cycle connections in the academic space will also be beneficial according to Tanzer. There are both structural and financial obstacles for academicians who wish to run HPC workloads in commercial cloud and Cycle insight will help Azure navigate that landscape to the benefit of Azure and academic users, he suggests. The Cray deal will help in government markets, he says.

Stay tuned.

The post Microsoft Spins Cycle Computing into Core Azure Product appeared first on HPCwire.

DDN’s 5th Annual High Performance Computing Trends Survey Cites Complex I/O Workloads as #1 Challenge

Tue, 12/05/2017 - 08:50

SANTA CLARA, Calif., Dec. 5, 2017 – DataDirect Networks (DDN) today announced the results of its annual High Performance Computing (HPC) Trends survey, which reflects the continued adoption of flash-based storage as essential to respondent’s overall data center strategy. While flash is deemed essential, respondents anticipate needing additional technology innovations to unlock the full performance of their HPC applications.  Managing complex I/O workload performance remains far and away the largest challenge to survey respondents, with 60 percent of end-users citing this as their number one challenge.

Run for the fifth year in a row by DDN, the survey results included input from more than 100 global end-users across a wide number of data-intensive industries. Respondents included individuals responsible for high performance computing as well as networking and storage systems from financial services, government, higher education, life sciences, manufacturing, national labs, and oil and gas organizations. As expected, the amount of data under management in these organizations continues to grow. Of organizations surveyed: 85 percent manage or use more than one petabyte of data storage (up 12-percentage points from last year).

Survey respondents continue to take a nuanced approach to cloud adoption. Respondents planning to leverage cloud-based storage (encompassing both private and public clouds) for at least part of their data in 2017 jumped to 48 percent, an 11-percentage point increase from 2016 survey results.  Despite a more positive disposition toward cloud storage, only 5 percent of respondents anticipated more than 30 percent of their data residing in the cloud. Maybe because of the limited use, as well as the ever-improving economics of public cloud services, a full 40 percent of respondents anticipated using public cloud in some way as a solution in the coming year even if for a limited amount of data. This response compares with only 20 percent of respondents last year who said they anticipated using public cloud storage options.

While the basic application of flash storage in HPC data centers remains relatively flat, at approximately 90 percent of respondents using flash storage at some level within their data centers today, the main shift is in how much data is being retained in flash. While the vast majority of respondents (76 percent) store less than 20 percent of their data on flash media, many respondents anticipate an increase in 2018, with a quarter of respondents expecting 20-to-30 percent of their data to be flash based, and another 10 percent expecting 20-to-40 percent of their storage to be on a flash tier.

How customers are applying flash to their workflows is also particularly interesting.  A majority of survey respondents (54 percent) are primarily using flash to accelerate file system metadata.  There is a growing interest in using flash for application-specific data as well, with 45 percent of respondents indicating that they are using at least some of their flash storage this way. On the other end of the flash usage spectrum few customers are using flash user data, which is logical given current cost deltas between flash and spinning disk storage.

“Once again, DDN’s annual HPC Trends Survey reflects the developments we see in the wider HPC community. I/O performance is a huge bottleneck to unlocking the power of HPC applications in use today, and customers are beginning to realize that simply adding flash to the storage infrastructure isn’t delivering the anticipated application level improvements,” said Kurt Kuckein, director of marketing for DDN. “Suppliers are starting to offer architectures that include flash tiers optimized specifically for NVM and customers are actively pursuing implementations utilizing these technologies. Technologies like DDN’s IME are specifically targeted to have the most impact on accelerating I/O all the way to the application.”

I/O bottlenecks continue to be the main concern for HPC storage administrators.  Especially in intensive I/O workflows like analytics, where 76 percent of customers running analytics workloads consider I/O their top challenge.  Given this, it is not surprising that only 19 percent of survey participates consider existing storage technologies sufficient to scale to exascale requirements.

A majority of respondents (68 percent) view flash-native caches as the most likely technology to resolve the I/O challenge and to push HPC storage to the next level, reflecting an eight-percentage point increase versus last year’s survey.  HPC storage administrators have already, or are beginning to evaluate flash-native cache technologies at a greater rate than before, with more than 60 percent of the responses indicating that they have implemented, are evaluating now, or plan to evaluate flash-native cache solutions such as NVM.  Evidence of the impact of these technologies can be seen in the recent io500.org results where JCAHPC utilized a flash-based cache to achieve stellar performance and the top spot in the first annual storage I/O benchmark ranking.

As an increasing number of HPC sites move to real-world implementation of multi-site HPC collaboration, concerns about security remain at the forefront. Perhaps somewhat surprisingly, the second highest barrier to multi-site collaboration has nothing to do with technology or security.   Organizational bureaucracy was identified by 43 percent of respondents as a major impediment to data sharing. This means that even though data sharing has become technically possible as well as cost effective, there are still limiting perceptions that stand in the way of wider collaboration.

About DDN

DataDirect Networks (DDN) is a leading big data storage supplier to data-intensive, global organizations. For almost 20 years, DDN has designed, developed, deployed and optimized systems, software and storage solutions that enable enterprises, service providers, universities and government agencies to generate more value and to accelerate time to insight from their data and information, on premise and in the cloud. Organizations leverage the power of DDN storage technology and the deep technical expertise of its team to capture, store, process, analyze, collaborate and distribute data, information and content at the largest scale in the most efficient, reliable and cost-effective manner. DDN customers include many of the world’s leading financial services firms and banks, healthcare and life science organizations, manufacturing and energy companies, government and research facilities, and web and cloud service providers. For more information, go to www.ddn.com or call 1-800-837-2298.

Source: DDN

The post DDN’s 5th Annual High Performance Computing Trends Survey Cites Complex I/O Workloads as #1 Challenge appeared first on HPCwire.

Verne Global Announces Launch of hpcDIRECT

Tue, 12/05/2017 - 08:29

LONDON and KEFLAVIK, Iceland, Dec. 5, 2017 — Verne Global, a provider of highly optimised, secure, and 100% renewably powered data center solutions, today announced the launch of its new hpcDIRECT service. hpcDIRECT is a powerful, agile and efficient HPC-as-a-service (HPCaaS) platform, purpose-built to address the intense compute requirements of today’s data-driven industries. hpcDIRECT provides a fully scalable, bare metal service with the ability to rapidly provision the full performance of HPC servers uncontended and in a secure manner.

“Building hpcDIRECT was a direct response to overwhelming demand from our customers and tightly correlated with the market’s desire to move from a CapEx to an OpEx model for high performance computing,”said Dominic Ward, Managing Director at Verne Global. “With hpcDIRECT, we take the complexity and capital costs out of scaling HPC and bring greater accessibility and more agility in terms of how IT architects plan and schedule their workloads.”

“hpcDIRECT has been designed and built from the outset by HPC specialists for HPC applications. Whether deployed as stand-alone compute or used as a bare metal extension to existing in-house HPC infrastructure, hpcDIRECT provides an industry-leading solution that combines best-in-class design with our HPC optimised, low cost environment and location.”

hpcDIRECT is accessible via a range of options, from incremental additions to augment existing high performance computing, to supporting massive processing requirements with petaflops of compute. This flexibility makes it an ideal solution for applications such as computer-aided engineering, genomic sequencing, molecular modelling, grid computing, artificial intelligence and machine learning.

hpcDIRECT is available with no upfront charges and can be provisioned rapidly to the size and configuration needed. hpcDIRECT clusters are built using the latest architectures available including Intel’s Xeon (Skylake) processors, and fast intercore connectivity using Mellanox Infiniband and Ethernet networks, with storage and memory options to suit each customer’s needs.

Since Verne Global began operations in early 2012, the company has been at the forefront of data center design and technology, bringing new, innovative thinking and infrastructure to the industry. hpcDIRECT is the latest product in this cycle, and is perfectly optimised for companies operating high performance and intensive computing across the world’s most advanced industries.

Source: Verne Global

The post Verne Global Announces Launch of hpcDIRECT appeared first on HPCwire.

GlobalFoundries, Ayar Labs Team Up to Commercialize Optical I/O

Mon, 12/04/2017 - 18:30

GlobalFoundries (GF) and Ayar Labs, a startup focused on using light, instead of electricity, to transfer data between chips, today announced they’ve entered into a strategic partnership. The companies will develop and manufacture Ayar’s optical I/O technology using GF’s CMOS fabrication process to deliver an alternative to copper interconnects that offers 10 times higher bandwidth and up to five times lower power. As part of the agreement, GF has also invested an undisclosed amount in Ayar.

The partners say the collaboration will create unique and differentiated solutions for cloud servers, datacenters and HPC customers and will benefit from GF’s investment in 45nm CMOS technology.

Ayar Labs’ photonics devices will be manufactured on GF’s 45nm RF SOI (Radio Frequency Silicon on Insulator) process at its East Fishkill fab. GF says it expects to deliver prototype parts for Ayar Lab customers in 2018 and will be ready to support production ramp-up post successful qualification.

GlobalFoundries East Fishkill, NY, facility

Although today’s announcement marks the start of a formal direct relationship between the companies, the researchers who formed Ayar Labs have using the fab’s technology to design silicon photonics components since 2009 but they did so via multi-project wafer runs, relying on aggregators who collect designs from university groups and startups that don’t have the resources to do full wafer runs.

Ayar Labs was launched in 2016 by a group of researchers from MIT, UC Berkeley, and CU Boulder who were part of a 10-year research collaboration funded by DARPA. Their breakthrough was to put advanced electronics and optics on the same chip leading to the development of the first microprocessor chip to communicate using light, implemented via standard CMOS.

“They had a really interesting approach,” explained Alex Wright-Gladstein, CEO of Ayar Labs and one of its cofounders. “Instead of taking manufacturing methods from the optics industry which usually use materials like indium phosphide and different III-V materials, instead of using that set of techniques to make optics, they just said let’s try to use standard CMOS manufacturing, pure silicon, with no change to the standard CMOS process, and see if we can make optics work in that totally different framework and use optical I/O instead of electrical I/O, get rid of electrical I/O entirely, get rid of copper.

“While there’s been attention on longer distance optical communication, we’re doing shorter distances, trying to replace copper cables that are used inside datacenters and even the copper traces on printed circuit boards. We’re very excited about the partnership with GlobalFoundries, having their backing and validation, because it will help open up a new customer base to us and this partnership will help us qualify our products and get them into the market faster,” said Wright-Gladstein.

Source: Ayar Labs

The problem that the technology is aiming to solve is well understood in HPC circles and in the semiconductor industry. Moore’s law has driven an exponential increase in the amount of computing power you can fit on a chip while the speed at which data moves in and out of chips has only made incremental gains. Over the past few decades, that has become a bottleneck so processors and servers can process huge amounts of data but spend a lot of time waiting to send and receive data.

The DARPA-backed research effort resulted in a chip with a bandwidth density of 300 gigabits per second per square millimeter, “about 10 to 50 times greater than packaged electrical-only microprocessors currently on the market.”

The technology is well described in the inventors’ December 2015 Nature paper, which HPCwire covered here. The first two authors, Chen Sun of UC Berkeley and Mark Wade of CU Boulder, are Ayar Labs full-time cofounders. The professors who co-authored the paper are part-time cofounders MIT’s Rajeev Ram, Vladimir Stojanovic at UC Berkeley, and Miloš Popovic from CU Boulder.

Improvements have been made in terms of data rates since the paper was published, Wright-Gladstein told HPCwire: “We’ve moved to standard wavelength ranges. We’re using O-band wavelengths rather than the non-standard 1,180nm that was described in that paper, but the fundamental architecture is still using that same approach so that micro-ring resonator based approach with dense wavelength division multiplexing (DWDM) at lots of wavelengths on a single fiber.”

Ayar’s first products will support 8 and 16 wavelengths of light on a single fiber and eventually they plan to go to 32.

Ayar is implementing the technology in multi-chip modules, where you have chiplets in the same package with very short electrical links connecting them. “So it doesn’t really matter where the processor is getting made or what node it’s in,” said the CEO. “It can be a 7nm CMOS node with ultra short reach links coming out of it that just go to our very close chiplet that’s integrated in-package with that processor chip or similarly with a switch ASIC, for example. And because those electrical links are so short they are also very very low power so you end up enabling a full kit package with multiple chiplets in it that is lower power than if you were to have beefier electrical SerDes driving the signal a longer distance.”

Ayar is targeting two spaces initially: high-performance computing and the traditional datacenter. “Until now there’s always been a tradeoff between going to go to higher bandwidth in your network versus having low latencies, and we are not forcing system architects to have to make those tradeoffs,” said Wright-Gladstein, who points to the potential for speeding machine/deep learning training by enabling highly parallel models.

Ayar also believes its technology has a role to play in enabling disaggregated architectures. “Having bigger pools of processing power and memory makes it so you can be more flexible about how you allocate your jobs across your datacenter and reduce the amount of time that your resources are idling and not being used,” said the CEO.

In the big picture, Wright-Gladstein is bullish about the success of on-chip optical I/O in the datacenter, expecting that it will replace CMOS SerDes within the next 5-10 years, clarifying that “it’s just the I/O portion that I think will be replaced, the power is we’re still using CMOS for everything else.”

“It’s pretty clear that we’re at the end of CMOS electrical I/O and being able to scale electrical SerDes as we have in the past,” she continued. “25 gigabits per second per pin is where we are today and people are working on 50 gigabits per second per pin. There’s a cadence of doubling that bandwidth every few years but folks are starting to struggle with the idea of moving to 100 gigabits per second per pin electrical SerDes and it’s tough to imagine going beyond that so something else needs to come along and [Ayar] technology which offers a 10x improvement rather than the standard 2x is really going to be an enabler for that.”

Alex Wright-Gladstein

While the Ayar CEO envisions light displacing electrical communications in the near future, over the longer term, she sees opportunities for silicon photonics beyond I/O, for example in quantum computing. She relates that teams looking to use optics for quantum computing face challenges with respect to manufacturability where it’s difficult to achieve the high consistency (low variation) that optical devices demand and high yield.

“At Ayar, we get to use this massive ecosystem of CMOS manufacturing, having GlobalFoundries, a standard CMOS fab, manufacturing our chips, something that is totally unique within optics,” she says. “Most optics manufacturing is much lower volume and much lower yield, but just the fact that billions of dollars have been poured into the CMOS manufacturing ecosystem means it’s a much more reliable manufacturing flow with much better controls. So in the longer term we want to make our platform available to a wide range of different applications for optics, such as quantum computing, LIDAR imaging for self-driving cars, and many healthcare applications.”

Anthony Yu, who leads the silicon photonics business within GF, told HPCwire that the company plans to move to its next-generation 45nm photonics process in 2019. That will be a follow on to the 90nm photonics process running currently in GF’s East Fishkill facility.

“That will be a process design entirely for photonics, entirely for things like optical tranceivers and we’ll be taking up Ayar Labs technology in the pure photonics process and bringing about even more performance for both Ayar Labs and our customers,” he said.

Of course, Ayar and GF aren’t the only companies pursuing the potential and promise of silicon photonics. Intel and IBM have demonstrated multiple breakthroughs already and hyperscalers have considerable motivation to develop their own technologies. Competition is sure to be fierce.

The post GlobalFoundries, Ayar Labs Team Up to Commercialize Optical I/O appeared first on HPCwire.

NVIDIA GPU Cloud Now Available to Hundreds of Thousands of AI Researchers Using NVIDIA Desktop GPUs

Mon, 12/04/2017 - 10:14

LONG BEACH, Calif., Dec. 4, 2017 — NVIDIA today announced that hundreds of thousands of AI researchers using desktop GPUs can now tap into the power of NVIDIA GPU Cloud (NGC) as the company has extended NGC support to NVIDIA TITAN.

NVIDIA also announced expanded NGC capabilities — adding new software and other key updates to the NGC container registry — to provide researchers a broader, more powerful set of tools to advance their AI and high performance computing research and development efforts.

Customers using NVIDIA Pascal architecture-powered TITAN GPUs can sign up immediately for a no-charge NGC account and gain full access to a comprehensive catalog of GPU-optimized deep learning and HPC software and tools. Other supported computing platforms include NVIDIA DGX-1DGX Station and NVIDIA Volta-enabled instances on Amazon EC2.

Software available through NGC’s rapidly expanding container registry includes NVIDIA optimized deep learning frameworks such as TensorFlow and PyTorch, third-party managed HPC applications, NVIDIA HPC visualization tools, and NVIDIA’s programmable inference accelerator, NVIDIA TensorRT 3.0.

“We built NVIDIA GPU Cloud to give AI developers easy access to the software they ned to do groundbreaking work,” said Jim McHugh, vice president and general manager of enterprise systems at NVIDIA. “With GPU-optimized software now available to hundreds of thousands of researchers using NVIDIA desktop GPUs, NGC will be a catalyst for AI breakthroughs and a go-to resource for developers worldwide.”

An early adopter of NGC is GE Healthcare. The first medical device maker to use NGC, the company is tapping the deep learning software in NGC’s container registry to accelerate bringing the most sophisticated AI to its 500,000 imaging devices globally with the goal of improving patient care.

New NGC Containers, Updates and Features

In addition to making NVIDIA TensorRT available on NGC’s container registry, NVIDIA announced the following NGC updates:

  • Open Neural Network Exchange (ONNX) support for TensorRT.
  • Immediate support and availability for the first release of MXNet 1.0
  • Availability of Baidu’s PaddlePaddle AI framework

ONNX is an open format originally created by Facebook and Microsoft through which developers can exchange models across different frameworks. In the TensorRT development container, NVIDIA created a converter to deploy ONNX models to the TensorRT inference engine. This makes it easier for application developers to deploy low-latency, high-throughput models to TensorRT.

Together, these additions give developers a one-stop shop for software that supports a full spectrum of AI computing needs — from research and application development to training and deployment.

Launched in October, NGC is also available free of charge to users of NVIDIA Volta GPUs on Amazon Web Services and all NVIDIA DGX-1 and DGX Station customers. NVIDIA will continue to expand the reach of NGC over time.

More information about NGC is available at www.nvidia.com/gpu-cloud.


NVIDIA‘s (NASDAQ: NVDA) invention of the GPU in 1999 sparked the growth of the PC gaming market, redefined modern computer graphics and revolutionized parallel computing. More recently, GPU deep learning ignited modern AI — the next era of computing — with the GPU acting as the brain of computers, robots and self-driving cars that can perceive and understand the world. More information at http://nvidianews.nvidia.com/.

Source: NVIDIA

The post NVIDIA GPU Cloud Now Available to Hundreds of Thousands of AI Researchers Using NVIDIA Desktop GPUs appeared first on HPCwire.

Mellanox and NEC Partner to Deliver Innovative High-Performance and Artificial Intelligence Platforms

Mon, 12/04/2017 - 09:24

SUNNYVALE, Calif. & YOKNEAM, Israel, Dec. 4, 2017 — Mellanox Technologies, Ltd. (NASDAQ: MLNX), a leading supplier of high-performance, end-to-end smart interconnect solutions for data center servers and storage systems, today announced in collaboration with NEC Corporation support for the newly announced SX-Aurora TSUBASA systems with Mellanox ConnectX InfiniBand adapters. A typical Aurora server platform includes from one to four InfiniBand adapters. The top-of-the-line Aurora platform utilizes 32 ConnectX adapters to support 64 vector engines in a single system. The NEC SX-Aurora TSUBASA systems leverage general-purpose vector-based NEC coprocessors and Mellanox in-network computing interconnect accelerators to achieve the highest application performance and scalability.

“We appreciate the performance, efficiency and scalability advantages that Mellanox interconnect solutions bring to our platform,” said Shigeyuki Aino, assistant general manager system platform business unit, IT platform division, NEC Corporation. “The in-network computing and PeerDirect capabilities of InfiniBand are the perfect complement to the unique vector processing engine architecture we have designed for our SX-Aurora TSUBASA platform.”

“Mellanox is proud to work with NEC to enable a next-generation computational platform for high-performance computing, machine learning, cloud and more,” said Gilad Shainer, vice president of marketing at Mellanox Technologies. “The combination of Mellanox ConnectX adapters, in-network computing, and acceleration engines, with NEC vector processing, provides our users with a world-leading compute platform that enables the highest application performance and the best return on investment.”

Mellanox InfiniBand solutions deliver the highest efficiency for high performance, artificial intelligence, cloud, storage and more infrastructures. InfiniBand accelerates all of the compute architectures – X86, Power, GPU, ARM, FPGA and Vector-based compute and storage platforms – delivering highest flexibility and best return on investment.

About Mellanox

Mellanox Technologies (NASDAQ: MLNX) is a leading supplier of end-to-end InfiniBand and Ethernet smart interconnect solutions and services for servers and storage. Mellanox interconnect solutions increase data center efficiency by providing the highest throughput and lowest latency, delivering data faster to applications and unlocking system performance capability. Mellanox offers a choice of fast interconnect products: adapters, switches, software and silicon that accelerate application runtime and maximize business results for a wide range of markets including high performance computing, enterprise data centers, Web 2.0, cloud, storage and financial services. More information is available at: www.mellanox.com.

Source: Mellanox

The post Mellanox and NEC Partner to Deliver Innovative High-Performance and Artificial Intelligence Platforms appeared first on HPCwire.

AMAX to Showcase HPC-Optimized DCIM Software for Heterogeneous Datacenters

Mon, 12/04/2017 - 09:17

FREMONT, Calif., Dec. 4, 2017 — AMAX, a leading manufacturer of High-Performance Computing solutions, Deep Learning platforms and dynamic Data Center solutions, today announced it is showcasing the [SMART]DC Data Center Manager Software and Appliance at Gartner IT Infrastructure, Operations Management & Data Center Conference on Dec. 4–7 in Las Vegas.

AMAX [SMART]DC Data Center Manager is the premier out-of-band DCIM solution for the modern-day heterogeneous data center. Built on a robust and field-proven software platform, the suite seamlessly manages legacy and white-box platforms using a single pane of glass, whether through a user-friendly GUI, command line or integration into existing software via API with zero code change needed.

As more GPU-integrated platforms move into data centers, [SMART]DC incorporates the capability of remote monitoring, managing and orchestrating GPU-based hardware alongside compute and storage hardware. For power-dense HPC and Deep Learning deployments, where real-time system health is critical to ensure uninterrupted operation, [SMART]DC not only provides insights of real-time temperature fluctuation, power consumption, capacity planning and server under- and over-utilization, but features fully automated policy-based server orchestration to maximize server health and data center power and efficiency management.

AMAX will feature [SMART]DC fully integrated at Booth #609 with the[SMART]Rack turnkey cluster. Designed plug-and-play and fully loaded, the [SMART]Rack, with compute, storage, backup power and management tiers all in one enclosure, is ready to drop into any data center. Variations of [SMART]Rack solutions include [SMART]Rack AI featuring NVIDIA GPUs for large-scale deep learning inference and data analytics and [SMART]Rack VM for virtual desktop infrastructure.

Mounted to the top of the rack, [SMART]DC manages all components inside the rack, including switches, CPU and GPU-based compute, storage and in-rack backup battery — to maintain server health, maximize hardware and rack space utilization and release stranded installed cooling and power capacity. Results are reduced CAPEX and OPEX, minimized TCO and maximized power savings to as much as 30 percent.

Other key benefits of the [SMART]DC include:

  • Out-Of-Band Management: Communication is agentless through the BMC management port, independent of server OS. Does not consume CPU resources or interfere with applications.
  • Virtual KVM: Administer, provision and diagnose servers from anywhere through remote KVM.
  • Real-Time Reporting: Track real-time server activity, power consumption and thermal trends. Gain insights needed for data center power and efficiency management and future capacity planning without having to rely on additional sensors or meters.
  • Integrated Health Monitoring: Detect, locate and identify server health issues. Receive alerts for over-cooling and hot spots of data room before they become incidents to improve uptime.
  • Security: Set user and group privileges for access control and rights management.
  • Density: Maximize server count per rack based on power consumption analytics to improve rack space efficiency.
  • Power Savings: Automatically flag idle and underutilized servers for consolidation or repurposing.
  • Call Home Ready: Supports add-on embedded service notification feature for AMAX hardware to expedite break/fix services.

“As computing systems continuously achieve higher performance, we need more powerful management tools to ensure workload and resource efficiency,” said Rene Meyer, VP of Technology, AMAX. “[SMART]DC Data Center Manager is designed to protect your investment through monitoring and managing power-dense infrastructures, especially those relying on compute-intensive solutions. Features like policy-based power and resource management are critical to keep infrastructures healthy, to ensure uptime, resource efficiency and maximum return on investment.”

[SMART]DC was developed to empower companies with the insight and control to achieve three major objectives — significant cost savings through more efficient use of IT, maximize density within usable data center space and to give facility operations an intelligent tool to operate and scale highly efficient data centers by leveraging advances in data analytics and software-defined policies.

To learn more about [SMART]DC Data Center Manager, please visit AMAX Gartner Data Center Infrastructure & Operation Management Conference Booth #609 or www.amax.com.

About AMAX 

AMAX is an award-winning global provider of cutting-edge Cloud, Data Center, HPC and next-generation computing technologies designed towards highest efficiency and optimal performance. As a full-service technology partner, AMAX tailors both solutions and programs to meet specific customer requirements. Whether you are a Fortune 1000 company seeking significant cost savings through better efficiency for your global data centers or a software startup seeking an experienced manufacturing partner to design and launch your flagship product, AMAX is your trusted solutions provider, delivering the results you need to meet your specific metrics for success. To learn more, please visit www.amax.com.

Source: AMAX

The post AMAX to Showcase HPC-Optimized DCIM Software for Heterogeneous Datacenters appeared first on HPCwire.

Supermicro Announces Filing of Revised Compliance Plan with NASDAQ Stock Market

Mon, 12/04/2017 - 08:53

SAN JOSE, Calif., Dec. 4, 2017 — Super Micro Computer, Inc. (NASDAQ:SMCI), a global leader in high-performance, high-efficiency server, storage technology and green computing, today announced that on November 29, 2017 it had submitted a revised compliance plan to The NASDAQ Stock Market (“NASDAQ”) to support its request for an extension of time to regain compliance with the NASDAQ continued listing requirements. Pursuant to NASDAQ rules, Super Micro’s securities will remain listed on the NASDAQ Global Select Market pending NASDAQ’s review of the revised compliance plan. Super Micro intends to take all necessary steps to achieve compliance with the NASDAQ continued listing requirements as soon as practicable.

Cautionary Statement Regarding Forward Looking Statements

Statements contained in this press release that are not historical fact may be forward-looking statements within the meaning of Section 27A of the Securities Act of 1933 and Section 21E of the Securities Exchange Act of 1934. Such forward-looking statements may relate, among other things, the preliminary financial information for the quarter ended September 30, 2017 and the guidance provided for the quarter ending December 31, 2017. Such forward-looking statements do not constitute guarantees of future performance and are subject to a variety of risks and uncertainties that could cause our actual results to differ materially from those anticipated, including any impact which may arise from the pending Audit Committee investigation. Additional factors that could cause actual results to differ materially from those projected or suggested in any forward-looking statements are contained in our filings with the Securities and Exchange Commission, including those factors discussed under the caption “Risk Factors” in such filings.

About Super Micro Computer, Inc.

Supermicro, a global leader in high-performance, high-efficiency server technology and innovation is a premier provider of end-to-end green computing solutions for Data Center, Cloud Computing, Enterprise IT, Hadoop/Big Data, HPC and Embedded Systems worldwide. Supermicro’s advanced Server Building Block Solutions offer a vast array of components for building energy-efficient, application-optimized, computing solutions. Architecture innovations include Twin, TwinPro, FatTwin, Ultra Series, MicroCloud, MicroBlade, SuperBlade, Double-sided Storage®, Battery Backup Power (BBP) modules and WIO/UIO.

Source: Super Micro Computer

The post Supermicro Announces Filing of Revised Compliance Plan with NASDAQ Stock Market appeared first on HPCwire.

Argonne Training Program on Extreme-Scale Computing Scheduled for July 29-August 10, 2018

Mon, 12/04/2017 - 08:37

Dec. 4, 2017 — Computational scientists now have the opportunity to apply for the upcoming Argonne Training Program on Extreme-Scale Computing (ATPESC), to take place from July 29-August 10, 2018.

With the challenges posed by the architecture and software environments of today’s most powerful supercomputers, and even greater complexity on the horizon from next-generation and exascale systems, there is a critical need for specialized, in-depth training for the computational scientists poised to facilitate breakthrough science and engineering using these amazing resources.

This program provides intensive hands-on training on the key skills, approaches and tools to design, implement, and execute computational science and engineering applications on current supercomputers and the HPC systems of the future. As a bridge to that future, this two-week program fills many gaps that exist in the training computational scientists typically receive through formal education or shorter courses. The ATPESC 2018 program will again be held at the Q Center, one of the largest conference facilities in the Midwest, located just outside Chicago.

Instructions for applying to the program can be found at http://extremecomputingtraining.anl.gov and the deadline for applicant submissions is Wednesday, Feb. 28, 2018.

Program Curriculum

Renowned scientists, HPC experts and leaders will serve as lecturers and will guide the hands-on laboratory sessions. The core curriculum will address:

  • Computer architectures and their predicted evolution
  • Programming methodologies effective across a variety of today’s supercomputers and that are expected to be applicable to exascale systems
  • Data intensive computing and I/O
  • Numerical algorithms and mathematical software
  • Performance measurement and debugging tools
  • Approaches to software productivity for HPC systems
  • Data analysis and visualization.

Eligibility and Application

Doctoral students, postdocs, and computational scientists interested in attending ATPESC can review eligibility and application details on the event website.


There are no fees to participate. Domestic airfare, meals, and lodging are provided.

ATPESC is supported by DOE’s Office of Science through the Exascale Computing Project, a joint project of the DOE Office of Science’s Advanced Scientific Computing Research Program and the National Nuclear Security Administration

Source: Argonne National Laboratory

The post Argonne Training Program on Extreme-Scale Computing Scheduled for July 29-August 10, 2018 appeared first on HPCwire.

From multicloud to one cloud for every workload

Mon, 12/04/2017 - 01:02

First there was the idea of cloud, followed by a tangle of cloud services piled up in every business around the globe. To manage such an eclectic pile of business user favorites, IT began to think of the piecemeal approach to doing business as the multicloud and set about to find ways to cobble it all together and make it work. But the reality is that business users still find themselves searching for an elusive mix of capabilities defined by individual workloads, and the services that make those capabilities more accessible and the work more manageable. The multicloud approach doesn’t solve those problems and so data silos persist, and economies of scale provided by cloud computing are rarely fully realized. To rectify those problems, there needs to be one cloud that can and does accommodate every workload. Now, such a cloud finally exists.

There are two main components in the one cloud strategy: hardware and software. Microsoft has addressed both with a firm commitment to expanding capabilities, improving efficiencies, and making all of it more accessible to more users. By expanding capabilities, Microsoft meant to take services to the cloud that before were impossible to move off the ground, including high-resource workloads such as artificial intelligence and supercomputing. Now those too are in the cloud as a service.

On the hardware side…

Microsoft partnered with Cray, NVIDIA, and Intel to complete the cloud infrastructure needed to meet the needs of any workload, no matter how complex.

The partnership with Cray brought supercomputing to Microsoft Azure datacenters for workloads in HPC, AI, and advanced analytics at scale. This exclusive partnership means organizations no longer need to compromise by either choosing the cloud for the large repositories or on-premises for the dedicated, tightly coupled architecture. Now organizations have both in the cloud with no compromises in capabilities, resources or performance.

New GPU offerings through the partnership with NVIDIA make it possible to train machine learning in a faster, more economical way and do it in the cloud. The NCv2 and ND series will be generally available by the end of 2017 to further those advancements in cloud computing for AI, machine learning and deep learning. The Azure NC-series enables users to run CUDA workloads on up to four Tesla K80 GPUs in a single virtual machine. The NC-series also provides RDMA and InfiniBand connectivity for extremely low-latency. This enables users to scale-up or scale-out any workload.

Linux and Windows virtual machines can be created in seconds for a wide range of computing scenarios using your choice of language, workload and/or operating system.

On the software side…

Microsoft makes it easy to run HPC in the cloud by offering an array of direct access services for end users, including Azure Batch, Batch AI, Batch Rendering, and CycleCloud.

Azure Batch enables batch computing in a cloud environment and handles resource provisioning, which enables the user to focus on their workload and not their infrastructure.  No capital investment is needed to gain access to a tremendous amount of scalable computing power.

Batch AI is a new cloud service that handles deep learning training and testing in parallel at scale. It frees researchers to focus solely on model development. Batch Rendering is Azure’s rendering platform, which also offers pay-per-use licensing for third-party rendering software. Both are domain-specific layers on top of Batch that further simplify access to compute resources for a given domain.

CycleCloud delivers cloud HPC and Big Compute environments to end users, and automates configurations too.

A strong ecosystem to complete the ‘one cloud for every workload’ build…

As any IT staff member knows, a strong ecosystem is essential to bridging and integrating applications, stabilizing and maturing a new technology or configuration, and generally adding to its capabilities and features. To that end, Microsoft has built a broad partner ecosystem around its one cloud concept that includes such heavy-hitters as Rescale, PBScloud, and Teradici.

Rescale on Azure is a high-performance computing (HPC) simulation platform which is fully integrated with all Azure data centers and over 200 simulation software applications.

PBScloud is also a HPC cloud management service with a wide range of tools to control and manage security, governance, costs, and poly-cloud environments.

Teradici provides the means for businesses to easily move Windows or Linux client applications to the public cloud.

Other partners will join the ecosystem over time as new applications such as those surrounding the still maturing AI industry arise and the need to be able to accommodate them and manage everything grows.

The first ‘one cloud’ on the market

Combined – hardware, software, and ecosystem – deliver the first ever ‘one cloud for every workload’ concept. It’s the result of an innovative and aggressive effort to overcome previous cloud limitations and reach high enough performance levels to meet the requirements of complex and large workloads. Until recently, such hasn’t been possible which is why bleeding edge workloads in HPC remained grounded in on-premises data centers and are the last to move to the cloud.

Some companies are using Microsoft cloud to move expenses from capex to opex and otherwise capitalize on economies of scale that cloud environs uniquely offer. End users prefer the ease of access and the one stop, silo busting capabilities rather than the typical span of cloud services.

Others prefer to test and benchmark new HPC and AI hardware and software before investing their capex dollars and associated labor costs. Still others are looking to this new one cloud idea to help them tame the coming avalanche of IoT data and the advanced analytics and data storage needs that are attached.

In any case this is the first cloud concept of its kind to hit the market. An array of exclusive partnerships, such as with Cray Supercomputing, makes it unlikely that a comparable competitor will arise any time soon. Obsolesce is also not likely to be an issue, given the innate nature of cloud computing structures (constant updates and upgrades), and the unique configuration of this specific cloud concept.

The post From multicloud to one cloud for every workload appeared first on HPCwire.

Linguists Use HPC to Develop Emergency-Response Translator

Sun, 12/03/2017 - 14:58

We live on a planet of more than seven billion people who speak more than 7,000 languages. Most of these are “low-resource” languages for which there are a dearth of human translators and no automated translation capability. This presents a big challenge in emergency situations where information must be collected and communicated rapidly across languages.

To address this problem, linguists at Ohio State University are using the Ohio Supercomputer Center’s Owens Cluster to develop a general grammar acquisition technology.

This graph displays an algorithm that explores the space of possible probabilistic grammars and maps out the regions of this space that have the highest probability of generating understandable sentences. (Source: OSC)

The research is part of an initiative called Low Resource Languages for Emergent Incidents (LORELEI) that is funded through the Defense Advanced Research Projects Agency (DARPA). LORELEI aims to support emergent missions, e.g., humanitarian assistance/disaster relief, peacekeeping or infectious disease response by “providing situational awareness by identifying elements of information in foreign language and English sources, such as topics, names, events, sentiment and relationships.”

The Ohio State group is using high-performance computing and Bayseian methods to develop a grammar acquisition algorithm that can discover the rules of lesser-known languages.

“We need to get resources to direct disaster relief and part of that is translating news text, knowing names of cities, what’s happening in those areas,” said William Schuler, Ph.D., a linguistics professor at The Ohio State University, who is leading the project. “It’s figuring out what has happened rapidly, and that can involve automatically processing incident language.”

Schuler’s team is using Bayseian methods to discover a given language’s grammar and build a model capable of generating grammatically valid output.

“The computational requirements for learning grammar from statistics are tremendous, which is why we need a supercomputer,” Schuler said. “And it seems to be yielding positive results, which is exciting.”

The team originally used CPU-only servers but is now using GPUs in order to model a larger number of grammar categories. The goal is to have a model that can be trained on a target language in an emergency response situation, so speed is critical. In August, the team ran two simulated disaster simulations in seven days using 60 GPU nodes (one Nvidia P100 GPU per node) but a real-world situation with more realistic configurations would demand even greater computational power, according to one of the researchers.

Read the full announcement here:

Owens Cluster technical specs here:

The post Linguists Use HPC to Develop Emergency-Response Translator appeared first on HPCwire.

HDF5-1.8.20 Release Now Available

Fri, 12/01/2017 - 08:54

Dec. 1, 2017 — The HDF5-1.8.20 release is now available. It can be obtained from The HDF Group Support Download page: https://support.hdfgroup.org/downloads/

It can also be obtained directly from: https://support.hdfgroup.org/HDF5/release/obtain518.html

HDF5-1.8.20 is a minor release with a few new features and changes. Important changes include:

  • An issue with H5Zfilter_avail was fixed where it was not finding available plugins.
  • Improvements were made to the h5repack, h5ls, h5dump, h5diff, and h5import utilities.

    In particular, please be aware that the behavior of the h5repack utility changed.

    A parameter was added to the “UD=” option of h5repack to allow the user defined filter flag to be changed to either H5Z_FLAG_MANDATORY (0) orH5Z_FLAG_OPTIONAL (1).

    An example of the command with the new parameter is shown here:
    h5repack -f UD=307,0,1,9 h5repack_layout.h5 out.h5repack_layout.h5

    Previously, the command would have been:
    h5repack -f UD=307,1,9 h5repack_layout.h5 out.h5repack_layout.h5

  • Support for the NAG compiler was added.
  • Many C++ APIs were added or modified to better reflect the HDF5 model.

This release contains many other changes that are not listed here. Please be sure to read the Release Notes for a comprehensive list of new features and changes.

Changes that affect maintainers of HDF5-dependent applications are listed on the HDF5 Software Changes from Release to Release page.

Future Changes to Supported Compilers

Please be aware that the minimum supported CMake version will be changing in the next release to CMake 3.8 or greater.

Source: The HDF Group

The post HDF5-1.8.20 Release Now Available appeared first on HPCwire.

UT Dallas Researcher Uses Supercomputers to Explore Nanopill Design

Fri, 12/01/2017 - 08:37

Dec. 1, 2017 — Imagine a microscopic gold pill that could travel to a specific location in your body and deliver a drug just where it is needed. This is the promise of plasmonic nanovesicles.

These minute capsules can navigate the bloodstream, and, when hit with a quick pulse of laser light, change shape to release their contents. It can then exit the body, leaving only the desired package.

This on-demand, light-triggered drug release method could transform medicine, especially the treatment of cancer. Clinicians are beginning to test plasmonic nanovesicles on head and neck tumors. They can also help efforts to study the nervous system in real-time and provide insights into how the brain works.

However, like many aspects of nanotechnology, the devil is in the details. Much remains unknown about the specific behavior of these nanoparticles – for instance, the wavelengths of light they respond to and how best to engineer them.

Writing in the October 2017 issue of Advanced Optical Materials, Zhenpeng Qin, an assistant professor of Mechanical Engineering and Bioengineering at the University of Texas at Dallas, his team, and collaborators from the University of Reims (Dr. Jaona Randrianalisoa), reported the results of computational investigations into the collective optical properties of complex plasmonic vesicles.

They used the Stampede and Lonestar supercomputers at the Texas Advanced Computing Center, as well as systems at the ROMEO Computing Center at the University of Reims Champagne-Ardenne and the San Diego Supercomputing Center (through the Extreme Science and Engineering Discovery Environment) to perform large-scale virtual experiments of light-struck vesicles.

“A lot of people make nanoparticles and observe them using electron microscopy,” Qin said. “But the computations give us a unique angle to the problem. They provide an improved understanding of the fundamental interactions and insights so we can better design these particles for specific applications.”

Striking Biomedical Gold

Gold nanoparticles are one promising example of a plasmonic nanomaterial. Unlike normal substances, plasmonic nanoparticles (typically made of noble metals) have unusual scattering, absorbance, and coupling properties due to their geometries and electromagnetic characteristics. One consequence of this is that they interact strongly with light and can be heated by visible and ultraviolet light, even at a distance, leading to structural changes in the particles, from melting to expansion to fragmentation.

Gold nanoparticle-coated liposomes — spherical sacs enclosing a watery core that can be used to carry drugs or other substances into the tissues — have been demonstrated as promising agents for light-induced content release. But these nanoparticles need to be able to clear the body through the renal system, which limits the size of the nanoparticles to less than few nanometers.

The specific shape of the nanoparticle — for instance, how close together the individual gold molecules are, how large the core is, and the size, shape, density and surface conditions of the nanoparticle — determines how, and how well, the nanoparticle functions and how it can be manipulated.

Qin has turned his attention in recent years to the dynamics of clusters of small gold nanoparticles with liposome cores, and their applications in both diagnostic and therapeutic areas.

“If we put the nanoparticles around a nano-vesicle, we can use laser light to pop open the vesicle and release molecules of interests,” he explained. “We have the capability to assemble a different number of particles around a vesicle by coating the vesicle in a layer of very small particles. How can we design this structure? It’s a quite interesting and complex problem. How do the nanoparticles interact with each other – how far are they apart, how many are there?”

Simulations Provide Fundamental and Practical Insights

To gain insights into the ways plasmonic nanoparticles work and how they can be optimally designed, Qin and colleagues use computer simulation in addition to laboratory experiments.

In their recent study, Qin and his team simulated various liposome core sizes, gold nanoparticle coating sizes, a wide range of coating densities, and random versus uniform coating organizations. The coatings include several hundred individual gold particles, which behave collectively.

“It is very simple to simulate one particle. You can do it on an ordinary computer, but we’re one of the first to looking into a complex vesicle,” Randrianalisoa said. “It is really exciting to observe how aggregates of nanoparticles surrounding the lipid core modify collectively the optical response of the system.”

The team used the discrete dipole approximation (DDA) computation method in order to make predictions of the optical absorption features of the gold-coated liposome systems. DDA allows one to compute the scattering of radiation by particles of arbitrary shape and organization. The method has the advantage of allowing the team to design new complex shapes and structures and to determine quantitatively what their optical absorption features will be.

The researchers found that the gold nanoparticles that make up the outer surface have to be sufficiently close together, or even overlapping, to absorb sufficient light for the delivery system to be effective. They identified an intermediate range of optical conditions referred to as the “black gold regime,” where the tightly packed gold nanoparticles respond to light at all wavelengths, which can be highly useful for a range of applications.

“We’d like to develop particles that interact with light in the near-infrared range – with wavelengths of around 700 to 900 nanometers — so they have a deeper penetration into the tissue,” Qin explained.

They anticipate that this study will provide design guidelines for nano-engineers and will have a significant impact on the further development of complex plasmonic nanostructures and vesicles for biomedical applications.

[In a separate study published in ACS Sensors in October 2017, Qin and collaborators showed the effectiveness of gold nanoparticles for assays that detect infectious diseases and other biological and chemical targets.]

Inspired by recent developments in optogenetics, which uses light to control cells (typically neurons) in living tissues, Qin and his team plan to use the technology to develop a versatile optically-triggered system to perform real-time studies of brain activity and behavior.

He hopes the fast release feature of the new technique will provide sufficient speed to study neuronal communication in neuroscience research.

“There are a lot of opportunities for using computations to understand fundamental interactions and mechanisms that we can’t measure,” Qin said. “That can feed back into our experimental research so we can better advance these different techniques to help people.”

Source: Aaron Dubrow, TACC

The post UT Dallas Researcher Uses Supercomputers to Explore Nanopill Design appeared first on HPCwire.

NCSA SPIN Intern Daniel Johnson Develops Open Source HPC Python Package

Thu, 11/30/2017 - 19:05

Nov. 30 — At the National Center for Supercomputing Applications (NCSA), undergraduate SPIN (Students Pushing INnovation) intern Daniel Johnson joined NCSA’s Gravity Group to study Albert Einstein’s theory of general relativity, specifically numerical relativity. Daniel has used the open source, numerical relativity software, the Einstein Toolkit on the Blue Waters supercomputer to numerically solve Einstein’s general relativity equations to study the collision of black holes, and the emission of gravitational waves from these astrophysical events. During his SPIN internship, Daniel developed an open source, Python package to streamline these numerical analyses in high performance computing (HPC) environments.

This image shows the collision of two black holes with masses of 28 to 36 and 21 to 28 times the mass of the sun, respectively. The spheres at the center represent the event horizons of the black holes. The size of the black holes has been increased by a factor of 4 to enhance visibility. Elevation and color of the surface gives an indicating of the strength of the gravitational field at that point. Orange is strongest, dark blue is weakest. The collision happened between 1.1 to 2.2 billion light years from Earth, and was observed from a direction near the Eridanus constellation. The mass equivalent of 3 suns was converted to gravitational radiation and radiated into space. Authors: Numerical simulation: R. Haas, E. Huerta (University of Illinois); Visualization: R. Haas (University of Illinois)

Just this month, Johnson’s paper “Python Open Source Waveform Extractor (POWER): An open source, Python package to monitor and post-process numerical relativity simulations” was accepted by Classical and Quantum Gravity, a remarkable feat for an undergraduate student, and one that will benefit the numerical relativity community.

“With a long history of developing scientific software and tools, NCSA provides a rich environment where HPC experts, faculty, and students can work together to solve some of the most challenging problems facing us today,” said NCSA Director Bill Gropp. “This is a great example of what young researchers can accomplish when immersed in an exciting and supportive research program.”

“It’s very gratifying to have an undergraduate’s work published in a world-leading journal in relativity,” said Eliu Huerta, research scientist at NCSA who mentored Johnson. “Given the lack of open source tools to post-process the data products of these simulations in HPC environments, there was an opportunity for Daniel to create something that could be very useful to the numerical relativity community at large.”

When Albert Einstein published his theory of general relativity in 1915, he probably couldn’t have imagined its transformative impact across science domains, from mathematics to computer science and cosmology. Einstein would have been pleased to foresee that the advent of HPC would eventually allow detailed numerical studies of his theory, providing key insights into the physics of colliding neutron stars and black holes—the sources of gravitational waves that the LIGO (Laser Interferometer Gravitational-Wave Observatory) detectors observed for the first time on September 14, 2015, and which are now becoming routinely detected by ground-based gravitational wave detectors.

Johnson developed an open-source Python package that seeks to streamline the process of monitoring and post-processing the data products of large scale numerical relativity campaigns in HPC environments. This, in turn, will allow researchers an end-to-end infrastructure within the Einstein Toolkit where they can submit, monitor, and post-process numerical relativity simulations.

“This whole project came to be when we were trying to compute numerical relativity waveforms from a large dataset we had created with the Einstein Toolkit in Blue Waters,” said Johnson. “We realized it would be much more efficient if we could post-process numerical relativity simulations directly on Blue Waters without having to move the massive amounts of data to another environment. I was able to adapt existing code and functions to Python, and now we can post-process huge amounts of data more efficiently than before.”

“It’s become clear that open source software is a key component in most research today, and increasingly, researchers are coming to realize that it can be an intellectual contribution on its own, separate from any one research result, as can be seen in the fact that Classical and Quantum Gravity accepted this software paper,” said Daniel S. Katz, assistant director of Science Software and Applications at NCSA. “Young researchers like Daniel are pushing the limits of what open source software can do and then getting credit for their software, which is a victory for the entire science community.”

The software Johnson developed, POWER, is an open-source Python package of great use to the larger numerical relativity community. It will lay the groundwork for future research and simulations on high performance computing systems across the globe.

“The SPIN program allowed me to combine my interests in physics and computer science,” said author and SPIN Intern Daniel Johnson. “The SPIN program provided me an avenue to learn about computational astrophysics, which is what I now plan to study in graduate school.”

About the National Center for Supercomputing Applications

The National Center for Supercomputing Applications (NCSA) at the University of Illinois at Urbana-Champaign provides supercomputing and advanced digital resources for the nation’s science enterprise. At NCSA, University of Illinois faculty, staff, students, and collaborators from around the globe use advanced digital resources to address research grand challenges for the benefit of science and society. NCSA has been advancing one third of the Fortune 50® for more than 30 years by bringing industry, researchers, and students together to solve grand challenges at rapid speed and scale.

About NCSA’S Students Pushing INnovation (SPIN) Program

NCSA has a history of nurturing innovative concepts, and some of the best ideas have come from highly motivated, creative undergraduate students. The SPIN (Students Pushing INnovation) internship program was launched to provide University of Illinois undergraduates with the opportunity to apply their skills to real challenges in high-performance computing, software, data analysis and visualization, cybersecurity, and other areas of interest to NCSA.

Source: NCSA

The post NCSA SPIN Intern Daniel Johnson Develops Open Source HPC Python Package appeared first on HPCwire.

HPE In-Memory Platform Comes to COSMOS

Thu, 11/30/2017 - 17:04

Hewlett Packard Enterprise is on a mission to accelerate space research. In August, it sent the first commercial-off-the-shelf HPC system into space for testing aboard the International Space Station (ISS) and this week the company revealed Stephen Hawking’s Centre for Theoretical Cosmology (COSMOS) would be using one of its systems to probe deeper into the mysteries of the space and time.

On Tuesday, HPE and the Faculty of Mathematics at the University of Cambridge announced COSMOS will leverage the new HPE Superdome Flex in-memory computing platform to process massive data sets that represent 14 billion years of history. The cosmologists are “searching for tiny signatures in huge datasets to find clues that will unlock the secrets of the early Universe and of black holes,” said Professor Hawking.

The HPE Superdome Flex, which HPE bills as the world’s most scalable and modular in-memory computing platform,” incorporates in-memory technology gained in the SGI acquisition. The Intel Skylake-based system is designed to scale from 4 to 32 sockets and supports 768 GB to 48 TB of shared memory in a single system. Although the exact configuration of the new supercomputer wasn’t disclosed, project leads say the shared-memory platform will enable the COSMOS group to process massive data sets much faster and will also ease the programming burden.

“In a fast-moving field we have the twofold challenge of analyzing larger data sets while matching their increasing precision with our theoretical models,” said Professor Paul Shellard, director of the Centre for Theoretical Cosmology and head of the COSMOS group in an official statement. “In-memory computing allows us to ingest all of this data and act on it immediately, trying out new ideas, new algorithms. It accelerates time to solution and equips us with a powerful tool to probe the big questions about the origin of our Universe.”

In order to expand and hone their scientific understanding of the cosmos, the team first forms theories about the nature of the universe, then creates precise simulations that are used to make predictions, and then tests those predictions against data from new sources, such as gravitational waves, the cosmic microwave background, and the distribution of stars and galaxies. A large in-memory computing system makes it possible to analyze the data through visualization and in real-time while the simulation is running.

More than 50 researchers will leverage the Cosmos system for a diverse range of fields. In addition to the COSMOS work, the Faculty of Mathematics at the University of Cambridge will use the in-memory computer to solve problems ranging from environmental issues to medical imaging and even experimental biophysics, for example, using light-sheet microscopy to study the dynamics of early embryonic development.

“Curiosity is essential to being human,” said Hawking in a video describing the collaboration (see below). “From the dawn of humanity we’ve looked up at the stars and wondered about the Universe around us. My COSMOS group is working to understand how space and time work, from before the first trillion trillionth of a second after the big bang up to today, fourteen billion years later.”

Professor Hawking, who led the founding of the COSMOS supercomputing facility in 1997, was on the forefront of transforming cosmology from a largely speculative endeavor to a quantitative and predictive science. He concludes, “Without supercomputers, we would just be philosophers.”

The post HPE In-Memory Platform Comes to COSMOS appeared first on HPCwire.