Supercomputing for the Exascale Era

Supercomputing for the Exascale Era

Cray, an HPE company, held a panel discussion webinar on October 18 to discuss Exascale (10^18, get it?) supercomputing. This is definitely not in my area of expertise, but it is certainly interesting.

Following is information I gleaned from links they sent to me. Basically, it is Why Supercomputing. And not only computers, but also networking to support them.

Today’s science, technology, and big data questions are bigger, more complex, and more urgent than ever. Answering them demands an entirely new approach to computing. Meet the next era of supercomputing. Code-named Shasta, this system is our most significant technology advancement in decades. With it, we’re introducing revolutionary capabilities for revolutionary questions. Shasta is the next era of supercomputing for your next era of science, discovery, and achievement.

WHY SUPERCOMPUTING IS CHANGING

The kinds of questions being asked today have created a sea-change in supercomputing. Increasingly, high-performance computing systems need to be able to handle massive converged modeling, simulation, AI, and analytics workloads.

With these needs driving science and technology, the next generation of supercomputing will be characterized by exascale performance, data-centric workloads and diversification of processor architectures.

SUPERCOMPUTING REDESIGNED

Shasta is that entirely new design. We’ve created it from the ground up to address today’s diversifying needs.

Built to be data-centric, it runs diverse workloads all at the same time. Hardware and software innovations tackle system bottlenecks, manageability, and job completion issues that emerge or grow when core counts increase, compute node architectures proliferate, and workflows expand to incorporate AI at scale.

It eliminates the distinction between clusters and supercomputers with a single new system architecture, enabling a choice of computational infrastructure without tradeoffs. And it allows for mixing and matching multiple processor and accelerator architectures with support for our
new Cray-designed and developed interconnect we call Slingshot.

EXASCALE-ERA NETWORKING

Slingshot is our new high-speed, purpose-built supercomputing interconnect. It’s our eighth generation of scalable HPC network. In earlier Cray designs, we pioneered the use of adaptive routing, pioneered the design of high-radix switch architectures, and invented a new low-diameter system topology, the dragonfly.

Slingshot breaks new ground again. It features Ethernet capability, advanced adaptive routing, first-of-a-kind congestion control, and sophisticated quality-of-service capabilities. Support for both IP-routed and remote memory operations broadens the range of applications beyond traditional modeling and simulation.

Quality-of-service and novel congestion management features limit the impact to critical workloads from other applications, system services, I/O traffic, or co-tenant workloads. Reduction in the network diameter from five hops (in the current Cray XCTM generation) to three reduces cost, latency, and power while improving sustained bandwidth and reliability.

FLEXIBILITY AND TCO

As your workloads rapidly evolve, the ability to choose your architecture becomes critical. With Shasta, you can incorporate any silicon processing choice — or a heterogenous mix — with a single management and application development infrastructure. Flex from single to multi-socket nodes, GPUs, FPGAs, and other processing options that may emerge, such as AI-specialized accelerators.

Designed for a decade or more of work, Shasta also eliminates the need for frequent, expensive upgrades, giving you exceptionally low total
cost of ownership. With its software architecture you can deploy a workflow and management environment in a single system, regardless of packaging.

Shasta packaging comes in two options: a 19” air- or liquid-cooled, standard datacenter rack and a high-density, liquid-cooled rack designed to take 64 compute blades with multiple processors per blade.

Additionally, Shasta supports processors well over 500 watts, eliminating the need to do forklift upgrades of system infrastructure to accommodate higher-power processors.

Predictive Tool to Improve Human-machine Interactions in Digital Manufacturing

Predictive Tool to Improve Human-machine Interactions in Digital Manufacturing

As manufacturing shifts towards smart factories, with interconnected production systems and automation, engineers at the University of Nottingham are leading a £1.9m project to develop a predictive toolkit to optimise productivity and communication between human workers and robots.

This research fits in with much other reporting I’ve done including the work of Dell Technologies on “human-machine partnerships.”

DigiTOP is one of seven national projects to create novel digital tools, techniques and processes to support the translation of digital capabilities into the manufacturing sector, funded by the Engineering and Physical Sciences Research Council (EPSRC).

It comes following the industry-led Made Smarter review, chaired by Siemens Chief Executive Juergen Maier, which stated that industrial digitalisation could be worth as much as £455bn to UK manufacturing over the next decade.

DigiTOP officially started on 1st July with the first month dedicated to project set up activities culminating in our internal kick off meeting at the end of the month, after which we should have a more outward focus. The project will take 39 months and complete on 30 September 2021. The twitter account @DigiTOP_Project will be regularly updated, and they are in the process of setting up a website to aid dissemination of progress.

A digital toolkit for the optimisation of operators and technology in manufacturing partnerships, DigiTOP will be led by Professor Sarah Sharples at the University of Nottingham in collaboration with Loughborough University, Cranfield University, University of the West of England, BAE Systems, Babcock International, Synertial Labs Ltd, Artinis Medical Systems B.V., High Value Manufacturing (HVM) Catapult and Jaguar Land Rover Ltd.

The toolkit will focus on using human factor theories and data to digitally capture and predict the impact of digital manufacturing on future working practices. Demonstrators will be used to test the implementation of sensing technologies that will capture and evaluate performance change and build predictive models of system performance.

The project will also provide an understanding of the ethical, organisational and social impact of the introduction of digital manufacturing tools and digital sensor-based tools to evaluate work performance in the future workplace.

DigiTOP’s findings will help companies that are planning to implement digital manufacturing technologies to understand how it will alter working practices, and how to optimise workplace designs to take these changes into account.

The tools developed within DigiTOP will help industry to design future work which might take place with a human and robot working in collaboration to complete a task or help with understanding how to design a data visualisation which shows how current parts of the factory are performing, and where maintenance or systems change might be needed in the short or long-term future.

Professor Sharples said: “The manufacturing industry, with the drive towards ‘Industrie 4.0’, is experiencing a significant shift towards digital manufacturing. This increased digitisation and interconnectivity of manufacturing processes is inevitably going to bring substantial change to worker roles and manual tasks by introducing new digital manufacturing technologies to shop floor processes.

“It may not be enough to simply assume that workers will adopt new roles bestowed upon them; to ensure successful worker acceptance and operational performance of a new system it is important to incorporate user requirements into digital manufacturing technologies design.

“New approaches to capture and predict the impact of the changes that these new types of technologies, such as robotics, rapidly evolvable workspaces, and data-driven systems are required,” adds Professor Sharples, who is Associate Pro-Vice Chancellor for Research and Knowledge Exchange for Engineering at Nottingham.

“These approaches consist of embedded sensor technologies for capture of workplace performance, machine learning and data analytics to synthesise and analyse these data, and new methods of visualisation to support decisions made, potentially in real-time, as to how digital manufacturing workplaces should function.”

The EPSRC investment arose out of work conducted by the Connected Everything Network Plus, which was established to create a multidisciplinary community focussed on industrial systems in the digital age.

EPSRC’s Executive Chair, Professor Philip Nelson, said: “The adoption of advanced ICT techniques in manufacturing provides an enormous opportunity to improve growth and productivity within the UK.

“The effective implementation of these new technologies requires a multidisciplinary approach and these projects will see academic researchers working with a large number of industrial partners to fully harness their potential, which could generate impact across many sectors.”

Manufacturing Professionals – Think You Don’t Matter to the Bottom Line

Manufacturing Professionals – Think You Don’t Matter to the Bottom Line

My latest email from The Information highlighted the woes and tribulations of Tesla. There are headlines in all the major media outlets—manufacturing problems at Tesla impacting stock price, profitability, and cash flow.

How would you like to be the engineers who “over automated” the factory according to the boss (Elon Musk)? Want to be the Director of Manufacturing hung out to dry in the Wall Street Journal or The New York Times?

Just consider all this and see how you matter to the company—the employees, stockholders, customers.

From The Information quoting Reveal:

Tesla’s 2018 is starting to look like Uber’s 2017: Every week there is a new allegation or setback about workplace culture or business performance or the quality of its products. In this case, an investigative report by Reveal says that Tesla consistently under-reported ailments suffered by workers at its main production plant. “Everything took a back seat to production,” said a former safety manager, Justine White, who left at the start of 2017. “It’s just a matter of time before somebody gets killed.” Tesla, as is its custom, fired back by calling the report by Reveal, which is part of the nonprofit Center for Investigative Reporting, a tool of an “extremist organization” that is trying to unionize Tesla’s workers and that reporters misunderstood how injuries are reviewed. We suggest reading the Reveal report and Tesla’s response, and coming to your own conclusion. (the Reveal)

And another quote from The Information about a class action lawsuit where the former director of manufacturing is giving information to the plaintiffs.

It’s not common for a shareholder class-action lawsuit, typically filed after a stock’s value has fallen precipitously, to get buzz among reporters. But this one against Tesla and its CEO Elon Musk seems unique: No fewer than 11 former workers at Tesla, including an ex-director of manufacturing at the company’s main car-production plant, provided information to the plaintiffs’ lawyers who filed the suit, according to an amended filing from March 23. It alleges Musk knowingly made false statements to investors that Tesla would be able to make 5,000 Model 3 sedans per week by the end of 2017, despite being told by his subordinates that that would never happen and continued to do so in the face of mounting evidence. Tesla’s stock dropped in price by 20% between May 2017 and November of that year, after it became clear that production target would not be met—not by a long shot. Five months later, the production pace is about 2,000 per week, Tesla has said. A spokesman for the company didn’t immediately respond to a request for comment about the suit, which is worth reading.

We have an important role within our companies. We must always consider that. Sometimes even being required to tell overoptimistic executives the reality of manufacturing.

Companies Discover IoT Applications Boost Performance

Companies Discover IoT Applications Boost Performance

I am still stung by a comment and ensuing discussion made by a maintenance manager during a talk I gave a few years ago. The talk was an early IoT description of networks, data, information, and the like. The guy raised his hand and said, “The engineers in my plant tell me that this stuff doesn’t work. So just forget about it!”

Emerson Automation developed a strategy called Top Quartile Performance and a service plan called Operational Certainty in order to operationalize Industrial IoT to benefit customers. This report comes from Covestro, one of the world’s largest polymer companies, which has selected Emerson to provide Industrial Internet of Things (IoT) technologies to help achieve its goals of minimizing risk and improving uptime at nine high-utilization plants.

As part of the $14 million, five-year contract, Emerson will provide remote monitoring and predictive maintenance to help Covestro optimize these manufacturing facilities for improved production, safety and reliability.

The Emerson program is a tenet of Covestro’s comprehensive digitization program called [email protected] that considers and implements new Industrial IoT strategies and operating procedures to deliver improved performance and meet defined financial targets. Covestro’s reliability program will leverage strategies, solutions and technologies in Emerson’s Operational Certainty program designed to help manufacturers achieve Top Quartile performance. Emerson data shows that Top Quartile companies spend half as much on maintenance compared to average performers and operate with an additional 15 days of available production each year.

Emerson will remotely monitor and maintain 40 of its DeltaV distributed control systems at Covestro plants in China, the United States and Germany. Remote teams at Emerson’s Innovation Center in Austin, Texas, will monitor and provide best practices-based maintenance strategies for local Emerson teams to implement at each Covestro plant. “By collaborating with Emerson to stay proactive about plant availability, we can drive toward always-on production and continue to satisfy customers in our high-demand market,” said Klaus Schaefer, chief technical officer, Covestro.

The Emerson-Covestro agreement reflects an emerging business model in industry, where manufacturers rely on a strategic supplier’s software solutions and deep automation expertise to monitor and execute maintenance, equipment health or energy management programs, allowing customers to focus their attention on critical operating functions that drive plant performance.

“Covestro and Emerson have a shared focus on driving Top Quartile operational performance,” said Jamie Froedge, president of Emerson’s Process Systems and Solutions business. “Connecting Covestro’s global product manufacturing expertise with our remote and local service capabilities allows the right expert to be available, real time, to ensure reliable operations.”

Manufacturing Professionals – Think You Don’t Matter to the Bottom Line

People Are The Key to Digital Transformation According to Emerson

The era of improving plant performance and profitability through efficiency—that is by cutting costs—is over. So stated Emerson Automation Solutions executive president Mike Train while kicking off the 2017 edition of Emerson Global Users Exchange in Minneapolis.

“The past 30 years have brought us fantastic advances in the manufacturing sector, including greater operating efficiencies enabled by automation,” said Train. “But the incremental benefits gained are diminishing. The pressure is on industry leaders to take the next step to the game-changing performance made possible by digitally empowering the workforce.”

Emerson has researched industry performance and drew a profile of Top Quartile industry performers – those in the top 25 percent of performance among their peers – Emerson has identified five essential competencies as critical to realize the value of “digital transformation”:

• Automated Workflow: Eliminate repetitive tasks and streamline standard operations to focus personnel on exceptions and other opportunities that require human intervention

• Decision Support: Leverage analytics and embedded expertise to provide actionable insights that reduce complexity and enable higher quality, faster decision-making

• Workforce Upskilling: Identify approaches that empower workers to acquire knowledge or experience faster and more effectively, to support higher-level and collaborative decision-making

• Mobility: Provide secure, on-demand access to information and expertise regardless of location, enabling collaborative workflows

• Change Management: Combine strategies, processes, tools and expertise that, in the right combination, simplify and accelerate the institutionalization of operational best practices

As always, this is a huge customer conference. There is abundant energy. Informal networking occurred all over the place. At this time, Emerson is the most vibrant of the companies in this area. It’ll be interesting to watch how, or if, business continues to grow from the company’s continued vision of industry.

More coming. Gotta listen to the next speaker.