Computer, Information Theory, and Robotics Society Journal
Edited by Steven Peliotis, IEEE Computer Society Journalist
Past Presentation: May 9, 2019
Denver IEEE Computer Society Guest Lecturer: Christopher Chang, PhD, HPC User Program Lead
Presentation: High Performance Computing at the National Renewable Energy Laboratory
Christopher Chang, a Senior IEEE and Computer Society member, focuses his research at NREL at the intersection of HPC software and workflows, electronic structure, artificial intelligence, and dynamics of connected systems.
As scaling laws and the development of information technologies have driven increasing fidelity and scope of simulations, computational science, engineering, and analytics (CSEA) take a greater role in engineering and investigation. CSEA is actively applied to materials discovery, energy-efficient design, Smart Grid technologies, and more. The newest enabler of technological advancement in these areas is Eagle, NREL’s 8-petaflop, 2114-node compute cluster. Dr. Chang discussed the architecture of this machine, NREL’s Computational Science Center, research computing, and the trends that might shape future systems.
In addition to his research role, as the User Program lead of the High Performance Computing (HPC) program at NREL, Dr. Chang manages applications, resource requests, and related topics that enable productive use of NREL’s HPC systems for science and engineering. The HPC program is integral to renewable energy research, which is NREL’s primary focus and an important part of the work sponsored by the Department of Energy.
Renewable energy is important first because of its economic impact. Approximately 6.4 million Americans work in the energy sector. Of the workforce dedicated to electric power, 43% are involved in solar energy, with the median wage for a solar installer at $26 per hour. Solar jobs include installation, manufacturing, sales and distributions, project development, R&D, and finance, and have increased at least 20% annually over the past four years. During that period, one in 50 new U.S. jobs was in the solar industry. The American solar industry employs more than twice as many people as the entire coal industry and as many people as the natural gas industry. In addition, there are about 2.4 million jobs in the motor vehicle sector.
Second, trends in energy supply are shifting. While coal use in the United States is declining as the use of natural gas increases, and the growth rate in natural gas use has been about double that of renewable energy sources over the past ten years, renewable energy is nevertheless growing fast. In the short-term in fact, wind and solar are expected to be the fastest growing sources of electricity generation in the United States for at least the next two years. Currently, renewables produce 10% of the total U.S. electricity generation, and renewable generation—not including hydropower—is expected to grow to 13% over the next two years. The number of utility-scale solar generating units is expected to grow by 10% in 2019 and by 17% in 2020. Wind energy is expected to grow by 12% in 2019 and 14% in 2020.
A key driver of renewables growth is their declining cost. Since 2009, the cost of utility-scale solar has gone from one of the most expensive methods of generating power to one of the least. By comparison, building solar power generation plants has become more economical in many ways than natural gas. Wind-powered generation has added another element to long-term analyses in regard to planning future infrastructure development to meet the needs of the growing U.S. population.
NREL focuses on applied research and analysis, and serves as an intermediary between universities and national laboratories on the one hand (which host more basic research), and the corporate setting on the other (where product development and production techniques are the primary focus). By shepherding technologies and products through intermediate technical readiness levels, NREL helps industry deliver them in robust, manufacturable states to market. Accelerated time to market provides significant advantages to American businesses and ultimately to consumers. NREL also performs techno-economic analyses of renewable and non-renewable energy sources to establish comparisons that enable decision makers to plan cost-effective power generation from all available sources.
NREL is one of a number of Department of Energy national laboratories. The Office of Science in the DOE focuses on basic science, and sponsors well known laboratories that include Oak Ridge National Laboratory, Argonne National Laboratory, and Lawrence Berkeley National Laboratory. The National Nuclear Security Administration (NNSA) conducts classified work at laboratories that include Lawrence Livermore, Los Alamos, and Sandia. Finally, other DOE laboratories carry out mission-oriented work related to their primary sponsors, which include NREL (DOE/EERE), Savannah River National Laboratory (DOE/EM), the Idaho National Laboratory (DOE/NE), and the National Energy Technology Laboratory (DOE/FE).
NREL’s primary sponsor is the Office of Energy Efficiency and Renewable Energy, or EERE. NREL hosts a diverse set of disciplines that serve the top-level concerns of Transportation, Renewable Power, and Efficiency, including molecular dynamics, quantum chemistry, materials science, computational fluid dynamics, multiphysics, and economic and technical analyses including production cost studies, capacity expansion, grid optimization, and economic dispatch (the short-term determination of the optimal mix of power generation facilities to meet system load at minimum possible cost).
HPC @ NREL
Within NREL, the Computational Science Center is part of the Scientific Computing and Analysis Directorate, and serves as the main repository of large-scale computing expertise at NREL and DOE/EERE. The Datacenter and staff are located in the ESIF (Energy Systems Integration Facility), an EERE User Facility to address challenges of interfacing different energy systems at meaningful scales. The datacenter is the most energy-efficient datacenter in the world devoted to energy efficiency and renewable energy research, with a recent PUE (Power Usage Effectiveness–the ratio of total power consumed and power devoted to computing per se) of 1.03. It can achieve this, versus typical values of 1.8 and industry records near 1.2, through a variety of concepts including warm water and evaporative cooling that can in turn enable waste heat recovery for secondary utilization.
HPC is distinguished from other computing categories by its focus on floating-point operations, parallelism, and high-performance networking. In 2018, NREL acquired its latest flagship system, called Eagle. Weighing in at 8 petaflops (8×10^15 64-bit floating-point operations per second) total Rpeak, Eagle provides about 3.5 times the computing capability as its predecessor, Peregrine. This relative increase was verified by a throughput benchmark suite of tests in addition to the traditional High Performance Linpack metric. The cluster is organized as 2114 dual-socket 36-core Intel® Skylake nodes arranged in an 8-dimensional hypercube topology, with an attached 14-PB Lustre parallel filesystem and 360-TB NFS-mounted utility filesystem. While operating in a different class from DOE’s Leadership Class machines, capable of hundreds of petaflops, it is interesting to note that a single Eagle node with an Rmax of 2.4 teraflops would have been a world-class supercomputer only 25 years ago (on the “Top 500” list, which documents the largest computing systems in the world, the June 1994 #1 system achieved 2.3 teraflops; this performance level was at the bottom of the list by June 2006).
Beyond pure high-performance numerical computing, the ESIF-HPC supports more general technical computing, which can include high-throughput computing, data-centric computing (e.g., Hadoop/Spark frameworks), small-scale calculations with commercial frameworks like MATLAB or ANSYS, and emerging paradigms like microservices and containers. Much of this work is not about coding or architecture, but rather how best to use the computing resources given the code.
The Evolving HPC Landscape
DOE’s Exascale Computing Initiative is intended to achieve an exaflop (1018 flops) by the early 2020s. As much as Eagle represents a huge amount of computing power, and as much as the coming exascale machines will be on the order of 100-fold more, the nature of future HPC is not clear. Several scaling laws that dominated the semiconductor and computing industries, and enabled relentless growth of computing power, are now ending. With the limits of transistor size and power density being approached, performance must be derived from other sources.
New computing models are multiplying, as are CPU architectures. Accelerators like FPGAs and GPUs are existing technology, and can serve certain workloads well. Ideas like photonic, neuromorphic, and quantum computing are still in research stages, although DOE has released reports on the latter two. NREL-CSC currently is exploring the implications of quantum computing, primarily as Noisy Intermediate-Scale Quantum or NISQ computing, on various combinatorial optimization problems relevant to the DOE/EERE mission. Combined with emerging paradigms like containers and cloud computing, the field of high-performance computing is rapidly gaining new dimensions. The near future will undoubtedly represent interesting times!
Past Presentation: April 25, 2019
Denver IEEE Computer Society Guest Lecturer: Richard George
Presentation: Costs/Benefits of Changing Human Capital, Financial, and Supply Chain Management Systems
Richard George is Vice President of Strategic Services at Sierra-Cedar, a provider of management consulting services across multiple technologies and industries, generally categorized into industry-based consulting services and industry-agnostic shared services. Service and Solution areas include Application Services, Business Intelligence, Host and Managed Services, Infrastructure Services, Integration and Cloud Solutions, Research, and Strategy.
Many organizations have moved or are considering moving from on-premise back office systems such as Human Capital Management (HCM), Financial Management Systems (FMS), and Supply Chain Management systems (SCM) to systems that are available in the Cloud such as Software as a Service (SaaS). Sierra-Cedar performs a systems survey of Human Resource operations. Surveys have been performed for more than 21 years. About 1600 current survey respondents provide detailed information about the HR systems that they use and how they use them.
The Sierra-Cedar 2018–2019 HR Systems Survey represents more than 20 years of continuous data gathering and is the most comprehensive survey in the industry. The survey describes Strategy, Process, and Structure; Administrative and Service Delivery Applications; Workforce Management Applications; Talent Management Applications; BI/Analytics/Workforce Planning Applications; Integration and Implementation; Emerging Technologies and Innovations; Vendor Landscape; Workforce and HR Expenditures; and Workforce Usage and Perception.
Survey results enable Sierra-Cedar to benchmark systems, assess the systems in current use, and to provide the basis for recommendations for changes (configurations, integrations, reporting) to systems used by client companies and adjustments in how those systems are used. Business cases are developed to identify solutions to issues in efficiency, innovation, and change management. Process improvements and cost effectiveness are two primary areas where Sierra-Cedar focuses its efforts in supporting its clients. The management consulting approach enables client companies to make purchasing and integration decisions to maximize quality of service while managing costs.
Time and money issues drive many decisions in Human Resources operations, as in most business settings. The transition to the Cloud (a remote data center that is not owned by a client company) for an on-premise system is motivated by a variety of considerations including improved user experience, mobile experience, reduced IT management and expenses, more consistent update scheduling, growth and scale requirements, and improved security and data storage.
The value of the survey includes long-term planning, enabling client companies to clarify and implement strategies. Developing and maintaining enterprise systems strategies are of critical importance because typically there are 25% higher business outcomes in organizations that have implemented functional enterprise systems strategies. Business outcomes that are quantified in the survey include innovation, market share, profitability, customer satisfaction, and competitiveness.
Because about half of organizations in business have no standard approach to enterprise technology integration, developing an integration strategy gains importance in effecting cogent, cost-effective policies that ultimately affect bottom-line performance. The analyses performed at the survey level provide the data that enables client companies to evaluate adopting Cloud-based systems for finance and HCM applications. Reasons for moving to the Cloud include improved performance, functionality, cost, security and data privacy, improved access to systems and data, improved user experience, more effective integration, configurability, and increased choices among vendors.
Methods of capturing employee data have become more complex, greater in number, and deliver more data than ever before. In finance applications, security and data privacy are the greatest concerns about Cloud adoption. Other concerns about moving to the Cloud include loss of control, integration complexities, cost, performance, and availability, the inability to customize, vendor lock-in, functionality, and decline in user experience.
Vendor satisfaction is an issue that is highly relevant because of the role of vendor systems in use in many companies. Benefits that correlate to high vendor satisfaction include best practice functionality, the ability to customize, integrated solutions, vendor relationship, and service and support. Low vendor satisfaction relates to issues of functionality not specific for the industry, poor service and support, the inability to customize, poor user experience, and high costs.
Decision making is facilitated by survey results that detail total HR technology costs per deployment and average costs by deployment model: All on Premise, Hybrid, All Cloud, and other implementations. Client companies are able to quantify how effectively the technologies that they use meet their current needs. Being able to determine how well systems in use meet company business needs enables business leaders to plan and make cost-effective decisions in a changing technological environment.
The transformation of HR technologies is facilitated through multiple pathways that provide the means to replace legacy systems, take a hybrid approach and move only selected applications to the Cloud, parallel and patchwork methods that combine licensed and Cloud solutions, and the move to hosted and outsourced single-tenant or BPO systems.
Adopting emerging technology applications require companies to adapt to the pervasive transition to mobile devices and social media. IaaS, PaaS, benchmarking databases, predictive analytics, sentiment analysis, machine learning, RPA, and Blockchain all must be considered in planning future integration and implementation to meet customer requirements while adapting to new developments in all areas of technology.
Mobile-enabled technologies require the adoption of processes to manage the transition to devices used by HR personnel and company employees. Policies are being written to control personal devices, their use, and the security of company data. Cyber security strategies are gaining importance as systems diversify, become more accessible to more people, and access more data than in previous generations of HR systems.
In terms of business success, having a system strategy is the number one priority regardless of the system being used. Second, a coherent integration strategy facilitates consistency and higher outcomes in the business. Reducing the number of integration tools contributes to efficiency in integrating solutions. Competitive advantage, as always, is facilitated by good user experience, effective customer service and support, positive vendor relationships, integrated solutions that are functional internally and in the customer-facing scenario, management of costs, and system and data security.
Past Presentation: March 14, 2019
Denver IEEE Computer Society Guest Lecturer: Ernest Worthman
Presentation: A Flyover of the Wireless World of Tomorrow
Ernest Worthman, in his presentation to the IEEE Computer Society, A Flyover of the Wireless World of Tomorrow, discussed the transition from existing voice and data communication technologies—the hardware and software systems currently in use—to wireless technologies that promise faster and wider pipes, more data at higher rates, ubiquitous coverage, and be critical to every segment of industry and their operations.
A major change in IT settings from the relatively familiar scenario will be where applications in use and conventional connections evolve into the model of the future. Tomorrow, disparate hardware systems will have to interface with a global platform of companies, their programs, and their business models.
The Cloud is becoming the universal platform being used by companies to manage an increasing number of applications, reduce on-site hardware, handle growing data requirements, and enable faster communication. IT professionals are growing into roles where they must manage and service an increasing range of systems outside of the traditional technologies. The shift from wired to wireless—3G, 4G, and emerging 5G—includes unlicensed wireless technologies such as Wi-Fi and Bluetooth.
This technology shift brings to companies the necessity of integrating into their operations a growing collection of devices that include smartphones, laptops, desktop PCs, tablets, and products yet to reach wide acceptance.
The most important issue for IT personnel, and the computer industry, will be the need to understand the emerging 5G wireless ecosystem, the umbrella of the next generation of wireless that will encompass all wireless technologies. The greatest challenge on the horizon will be interfacing current systems with the new 5G model. Networks—smart networks, virtualized networks, the Internet of Everything (IoX), Smart-interface Buildings, Autonomous Vehicles—are adding complexity to the IT world.
Wider channels of communication and new technologies are being added to the mix as product developers create new functionalities. The demand on industry is to develop technological opportunities while creating cohesive networks and serving a growing global population of technology users.
Creativity in technology is becoming a necessity for professionals in the IT field. Rapid changes in service requirements require analytical acumen, the ability to learn from system interaction, and the aptitude to create solutions where hardware, software, and networks interact in ways not necessarily foreseen by their designers.
5G promises greater interoperability among systems. Cloud and wireless technologies will connect existing systems to newly developed technologies. Traditional network topologies will evolve and become more efficient as the new 5G umbrella connects current systems, older legacy systems, and the new products and schemas being created. Technological advancements are arriving more quickly than ever and the benefits to society are increasing at faster rates.
Scientists, engineers, and other technical professionals are making their contributions to the future of our society by creating applications that are becoming ubiquitous in their use, accepted as normal in everyday life, and contribute increasingly to quality of life. The future is now.