Mission Need Statement Nersc

Summary

Mission Need Statement
NERSC-8
Mission Need Statement
for the
Next Generation High Performance
Production Computing System Project
(NERSC-8)
(Non-major acquisition project)
Office of Advanced Scientific Computing Research
Office of Science
U.S. Department of Energy
Date Approved:
Month / Year
Page 1
Mission Need Statement
NERSC-8
Submitted by:
David Goodwin, Program Manager Date
Advanced Scientific Computing Research, Office of Science, DOE
Concurrence:
Daniel Lehman, Director, Date
Office of Project Assessment, Office of Science, DOE
Approval:
Daniel Hitchcock, Acquisition Executive, Associate Director, Date
Advanced Scientific Computing Research, Office of Science, DOE
Page 2
Mission Need Statement
NERSC-8
Table of Contents
1 Statement of Mission Need .................................................................................................... 4
2 Capability Gap/Mission Need................................................................................................ 4
2.1 Scientific Demand on Computing Resources ................................................................ 6
2.1.1 Science at Scale ........................................................................................................... 6
2.1.2 Science Through Volume............................................................................................ 7
2.1.3 Science in Data............................................................................................................ 7
2.2 Strategic Risk to DOE Office of Science if Not Approved ........................................... 8
3 Potential Approach................................................................................................................. 8
3.1 Constraints and Limitations ........................................................................................... 9
4 Resource and Schedule Forecast ........................................................................................... 9
4.1 Cost Forecast ................................................................................................................... 9
4.2 Schedule Forecast............................................................................................................ 9
4.3 Funding Forecast ............................................................................................................ 9
Page 3
Mission Need Statement
NERSC-8
1 Statement of Mission Need
The U.S. Department of Energy (DOE) Office of Science is the lead federal agency supporting
basic and applied research programs that accomplish DOE missions in efficient energy use,
reliable energy sources, improved environmental quality, and fundamental understanding of
matter and energy. The research and facilities funded by the Office of Science are critical to
enhancing U.S. competitiveness and maintaining U.S. leadership in science and technology. One
of two principal thrusts within SC is the direct support of the development, construction, and
operation of unique, open-access High Performance Computing (HPC) scientific user facilities

For computing within SC, the Office of Advanced Scientific Computing Research (ASCR) has a
mission to discover, develop, and deploy computational and networking capabilities to analyze,
model, simulate, and predict complex phenomena important to the DOE. Enabling extreme-scale
science is a major ASCR priority. In support of this mission ASCR operates the National Energy
Research Scientific Computing Center (NERSC) at Lawrence Berkeley National Laboratory

NERSC serves as SC’s High Performance Production Computing (HPPC) Facility, supporting
the entire spectrum of SC research, and its mission is to accelerate the pace of scientific
discovery by providing high performance computing, information, data, communications, and
support services for Office of Science sponsored research

The U.S. Department of Energy (DOE) Office of Science (SC) requires a high performance
production computing system in the 2015/2016 timeframe to support the rapidly increasing
computational demands of the entire spectrum of DOE SC computational research. The system
needs to provide a significant upgrade in computational capabilities, with at least a ten-times
increase in sustained performance over the NERSC-6 Hopper system on a set of representative
DOE benchmarks

In addition to increasing the computational capability available to DOE computational scientists,
the system also needs to be a platform that will begin to transition DOE scientific applications to
more energy-efficient, manycore architectures required for exascale computing. This need is
closely aligned with the US Department of Energy’s 2011 strategic plan, which states an
imperative to continue to advance the frontiers of energy-efficient computing and
supercomputing to enable greater computational capacity with lower energy needs. Energy-
efficient computing is a cornerstone technology of what has been called exascale computing and
represents the only way of continuing NERSC’s historic performance growth in response to
science needs. In the most recent DOE strategic plan, development and deployment of high-
performance computing hardware and software systems through exascale is a targeted outcome
within the Science and Engineering Enterprise goal of maintaining a vibrant U.S. effort in
science and engineering with clear leadership in strategic areas

2 Capability Gap/Mission Need
During 2009-2011, ASCR commissioned a series of workshops to characterize the computational
resource requirements each program office will need to reach its research objectives in 2014

(For full workshop reports see: http://www.nersc.gov/science/requirements-workshops/.) A
careful analysis of these requirements demonstrates a critical mission need for over 14 times the
Page 4
Mission Need Statement
NERSC-8
current High Performance Production Computing platform capability by 2014 to address
computational needs of projects sponsored by Office of Science Program Offices and to avoid
creating an unacceptable gap between needs and available computing resources. By 2016, it is
estimated that there will be a need for 47 times the current HPPC capability. Due to budget
constraints, NERSC’s 2013 supercomputer acquisition (“NERSC-7”) will satisfy only a small
portion of the need identified in the workshops

Dr. William F. Brinkman, Director of the Office of Science Department of Energy, stated in
March 20, 2012 testimony before the House Subcommittee on Energy & Water Appropriations,
that, “In the course of our regular assessment of the needs of the scientific community, it is clear
that in several areas DOE’s simulation and data analysis needs exceed petascale capabilities.”
Brinkman, as well as others, noted, however, that making the transition to exascale poses
numerous unavoidable scientific, algorithmic, mathematical, software, and technological
challenges, all of which result from the need to reduce power consumption of computing
hardware

It is now widely recognized that computing technology is undergoing radical change and that
exascale systems will pose a wide variety of technology challenges. These challenges have
arisen because hardware chip technologies are reaching physical scaling limits imposed by
power constraints. Although transistor density continues to increase, chips will no longer yield
faster clock speeds. Instead, vendors are increasing the number of cores on a chip such that
within this decade, chips are expected to have over 100 times more processing elements than
today. These innovations in architecture and vast increases in chip parallelism (referred to as
“manycore”) will affect computers of all sizes. Many of the challenges we anticipate in the
exascale computing era are confronting us even today. According to the recent National
Research Council report The Future of Computing Performance: Game Over or Next Level?,
“We do not have new software approaches that can exploit the innovative architectures, and so
sustaining performance growth—and its attendant benefits—presents a major challenge.” Users
of such systems will require new algorithms that run more efficiently, exploit parallelism at a
much deeper level, accommodate far less memory per process space, and take much greater care
in considering data placement and movement and machine resilience. The recent report from the
ASCR Workshop on Extreme-Scale Solvers: Transition to Future Architectures outlined these
challenges in detail for the mathematical solver technology that is at the heart of the application
codes that enable DOE scientific discovery

According to the summary report of the Advanced Scientific Computing Advisory Committee in
the fall of 2010 entitled, “The Opportunities and Challenges of Exascale Computing”, the
transition required to adapt to these new architectures is expected to be as disruptive as that from
vector machines to massively parallel systems in the early 1990s. While this transition was not
easy, NERSC was able to work with the scientific user community to transform applications
from a vector model to an entirely different programming model based on the Message Passing
Interface (MPI). A high level of user engagement will again be necessary to help users transition
to more energy efficient architectures due to the large number of applications running at NERSC

Because of NERSC’s successful history of engaging with users, locating a production, low-
power, manycore machine at NERSC will take DOE scientists through the technology transition
to manycore architectures required for exascale. Not all Office of Science computing problems
require computing at the exascale level but all SC computing problems must be aware of the all-
encompassing issues of energy efficient computing, manycore architectures and programming
Page 5
Mission Need Statement
NERSC-8
models, communication reduction, and fault tolerance. In other words, those challenges that
have been identified for reaching exascale will be necessary for achieving performance growth
on systems of any size but especially at the level required for the next HPC system. Therefore,
there is a need to move the entire DOE workload to a more energy-efficient manycore machine;
that is, a need to provide, for thousands of users, a machine based on the innovative architectures
that are emerging for the next generation extreme scale computing systems. The consequence of
the trends noted in the National Research Council report is that computing performance required
to achieve DOE science goals can be attained only if parallel computing systems, and the
software that runs on them, can become more power-efficient

2.1 Scientific Demand on Computing Resources
The workshops that NERSC held with each of the SC Offices reinforced the notion that DOE
users support projects addressing a wide range of essential scientific issues of pressing national
and international concern. The collective nature of the workshops, the selection of participants
spanning the full range of Office of Science research areas, and the participation of DOE
program managers, enabled derivation of consensus computing resource requirements for
science goals that are well aligned with the mission of all six DOE program offices. The
workshop reports show that computational requirements for SC Program Offices are expected to
dramatically increase and without a significant increase in resources, progress on key DOE SC
initiatives are in danger. In total, over 47 times 2011 capability are needed in 2016. This is
approximately equivalent to a 75 peak peta-flop/second system. Quite simply, the demand for
high performance computing resources by the DOE SC scientists and researchers far outpaces
the ability of the DOE Office of Science production facility to provide it

The scientific goals driving the need for additional computational capability and capacity are
clear. Well-established fields that already rely on large-scale simulation, such as fusion and
climate modeling, are moving to incorporate additional physical processes and higher resolution

Additional physics to allow more faithful representations of real-world systems is needed, as is
the need to model larger systems in more realistic geometries and in finer detail. Acutely
constrained resources mean that scientists must compromise between increasing spatial
resolution, performing parameter studies, and running higher volumes of simulations

In order to support the diverse research and development needs of DOE mission-driven science
the NERSC-8 system will be expected to support three essential classes of computational inquiry
that transcend any one mission category: Science at Scale, Science through Volume, and Science
in Data

2.1.1 Science at Scale
There are many compelling examples within the NERSC workload of science areas for which an
extreme scale system is needed. Many of the projects that run at NERSC also have a current
INCITE allocation and INCITE users across a broad range of science areas run many of the same
codes at NERSC as they do at the Leadership Computing Facilities. There are also many
projects at NERSC with codes that run at scale which do not have INCITE allocations. Today,
numerous scientists running applications at NERSC have codes capable of routinely using
100,000 cores or more. To illustrate, NERSC workshop attendees reported that requirements for
fusion research using particle-based codes will need to use more than a million cores for the
largest runs and 10,000 to 100,000 cores for hundreds of routine runs, while integrated transport-
Page 6
Mission Need Statement
NERSC-8
MHD modeling will likely require HPC resources at the exascale to address the problem of
electromagnetic turbulence relevant to ITER

2.1.2 Science Through Volume
A large and significant portion of the scientific discovery of importance to DOE consists of
computational science not performed at the largest scales, but rather, performed using a very
large number of individual, mutually-independent compute tasks, either for the purpose of
screening or to reduce and/or quantify uncertainty in the results

High-throughput computational screening, in which a database of results from a large number of
similar or identical simulations on mixtures of related materials, chemicals, or proteins is
compiled and then scanned, has rapidly become a key computational tool at NERSC. One
approach is to carry out low-resolution studies on huge numbers of materials to narrow the focus
and then carry out more detailed studies on only those identified as promising candidates. This
method is being used in the Materials Project to cut in half the typical 18 years required from
design to manufacturing, in, for example, the 20,000 potential materials suitable for lithium ion
battery construction. Another similar example is the search through approximately three million
metal organic frameworks to reveal those most suitable for storing (or sequestering) carbon
dioxide; there are currently two successful projects in this area at NERSC. Another example
uses massive numbers of molecular dynamics simulations to catalogue protein conformations for
better understanding of biological pathways relevant to biofuels, bioremediation, and disease

This method of scientific inquiry is expected to grow at NERSC in the coming years

A key finding from the NERSC/Basic Energy Sciences (BES) workshop was that poor job
turnaround time due to inadequate compute resources profoundly limits the pace of scientific
progress in volume-based computational science. The simple availability of additional resources
immediately improves the pace of scientific discovery

Additionally, computational time dedicated to Uncertainty Quantification (UQ) is expected to
continue to grow as more scientific disciplines include verification, validation and UQ as
standard simulation analysis to provide more defensible and quantifiable understanding of model
parameters and simulation results. Already UQ is an important component in simulation areas
such as fusion energy, climate modeling, high-energy physics accelerator design, combustion,
subsurface environmental modeling, turbulence, and nuclear physics. Many UQ studies require
large ensemble calculations. The size of the ensemble runs depends on the number of parameters,
model nonlinearities, and the number of models under simultaneous study. In the report High
Performance Computing and Storage Requirements for Advanced Scientific Computing
Research, experts predict a need for an additional 10 percent increase in computer time across all
projects for scientists to perform UQ studies

2.1.3 Science in Data
A next generation HPC system is required to meet the rapidly growing computational and
storage requirements to support key DOE user facilities and experiments such as the Joint
Genome Institute (JGI), Planck, and the Advanced Light Source (ALS). As an example, the
Office of Basic Energy Sciences (BES) recently created the Scientific User Facilities (SUF)
Division, which encompasses 17 user facilities and research centers. In recognition of the
growing need for computing resources to support these facilities, BES has created a new pool of
computational hours for SUF researchers for the 2012 allocation year. Data set sizes are growing
Page 7
Mission Need Statement
NERSC-8
exponentially, because detectors in sensors, telescopes, and sequencers are improving at rates
that match Moore’s Law. The JGI Microbial Genome and Metagenome Data Processing and
Analysis Project is a prime example of the explosive growth in computing needed to support data
analysis and also a good example of overlap between Science in Data and Science at Scale

Rapid advances in DNA sequencing platforms driven by comparative analysis and bioenergy
needs are fueling a substantial increase in the number and size of genome and metagenome
sequence datasets. The next generation HPPC system is required to meet the demand from these
facilities

2.2 Strategic Risk to DOE Office of Science if Not Approved
If the next generation HPPC system is not acquired according to plan and made available to the
DOE scientific community on the proposed schedule, there will be numerous, direct, and
consequential impacts to the nation’s energy assurance and to the DOE mission, such as:
● Approximately 4,500 research scientists will see no increase in computational resources,
as required to further scientific discovery, engineering innovation, and maintain U.S

global competitiveness

● DOE Office of Science HPPC users will not have a system to transition applications and
algorithms to the next generation of energy efficient, massive on-chip parallelism systems
required for exascale computing

● DOE Office of Science community will be unprepared to take advantage of energy
efficient computing platforms

● New science initiatives will not be accommodated, thus increasing the number of U.S

scientists and engineers who do not have access to world-class tools, adversely affecting
American long-term competitiveness and prestige

● Other DOE user facilities will be unable to process and analyze data sufficiently or
sufficiently quickly and some will be left unanalyzed

3 Potential Approach
The NERSC-8 project will conduct an Alternatives Analysis prior to CD-1. Potential approaches
to meet the mission need include constructing a new facility, utilizing an existing DOE facility
such as NERSC, or utilizing a commercial or academic facility. Technical alternatives will
include upgrading or replacing an existing computing system, acquiring a new high performance
computer system, or utilizing shared resources such as cloud computing. The NERSC-8 project
will also consider partnerships with other National Laboratories such as ACES (Alliance for
Computing at the Extreme Scale, a collaboration between Los Alamos National Laboratory and
Sandia National Laboratory)

NERSC is in a unique position to help transition the broad scientific community to new energy
efficient architectures because of its exemplary record of science engagement with users and its
support of the broad DOE production computing user base. An extensive ‘application readiness’
effort will be key to assuring that the architected system can perform well on the DOE workload

The intention of this effort will be to help application developers make changes to their codes
that will be sustained to future systems. Doing so is the only way to achieve additional
computational performance required to meet science goals

Page 8
Mission Need Statement
NERSC-8
3.1 Constraints and Limitations
In order to satisfy the mission need, the next generation HPC system must satisfy the following
constraints and limitations:
● Maximize application performance while transitioning the DOE Office of Science user
community to energy-efficient computing platforms with massive on-chip parallelism

● Provide high bandwidth access to existing data stored by continuing research projects

● Delivery of the system in the 2015/2016 timeframe

● Fit within the budget profile of the NERSC Center

4 Resource and Schedule Forecast
4.1 Cost Forecast
The estimated Total Project Cost (TPC) range for the considered alternatives is $5 to $19.8M

For this type of project, there is no Other Project Costs (OPC), so TEC is equal to TPC

4.2 Schedule Forecast
Critical Decision (CD) Fiscal Year
CD-0 Approve Mission Need FY 2013
CD-1 Approve Alternate Selection and Cost Range FY 2013
CD-3A Approve Long Lead-time Procurements
CD-2 Approve Project Baseline FY 2014
CD-3B Approve Start of Execution (Acquisition)
CD-4 Approve Project Completion FY 2016
4.3 Funding Forecast
The following funding profile projection is based on the high end of the project cost estimate

Fiscal Year FY13 FY14 FY15 FY16 Total ($M)
OPC 0 0 0 0 0
TEC $2 $3 $8 $6.8 $19.8
TPC $2 $3 $8 $6.8 $19.8

 
The Total System Cost is estimated to be between $50M-$100M.
 
Page 9

Mission Need Statement NERSC-8 Page 5 current High Performance Production Computing platform capability by 2014 to address computational needs of projects sponsored by Office of …

Download Now

Documemt Updated

Popular Download

• › Multi Year Planning Template Mypt Job Aid
• › Martial Arts Studio Business Plan Template
• › Marketing With Firm Brochures General Information Brochures
• › Microsoft Teams Use Case Catalog Fhubspotusercontent40net
• › Mental Health Treatment Plans
• › Making Effective Health Promotion Materials University Of
• › Monthly And Quarterly Financial Reporting
• › Model Making As A Career
• › Microsoft Office Publisher 2013
• › Microsoft Word 2010 For Dummies Download Save Print
• › March 2022 Printable Calendar
• › Multifunction Color Ricoh Mp C3004 Mp C3504
• › Monti Senior News
• › Minneapolis Business College
• › Ms Stahls Classroom Newsletter
• › Medical Equipment Management Plan Duke University
• › Missouri Real Estate Purchase Agreement
• › M E S S A G E Free Forms Online
• › Ms Word Business Brochure Template Software Activator
• › Mindful Eating Handouts

Frequently Asked Questions

What is a mission need statement?

Mission Need Statement. Jump to navigation Jump to search. A Mission Needs Statement (MNS) is a U.S. Department of Defense type of document which identifies capability needs for a program to satisfy by a combination of solutions (DOTMLPF) to resolve a mission deficiency or to enhance operational capability.

How to prepare a needs statement?

In preparation for making a needs statement, the following should be considered: The needs statement should contain the what of the free proposal examples and should focus on the issue at hand. The needs statement should not in any way mention of the solution but of the issues being addressed by the needs statement analysis.

What is the coast guards mission needs statement?

Not later than July 1, 2015, the Commandant shall submit to the Committees a new Mission Needs Statement (MNS), which will be used to inform the out-year CIP. The MNS should assume that the Coast Guard requires the capability to continue to carry out all of its eleven statutory missions. House Report 113-481 includes the following:

What does the nsarc do?

DOT: As a member of the NSARC, the Federal Aviation Administration provides emergency service to aircraft in distress, assuring that search and rescue procedures will be initiated if an aircraft becomes overdue or unreported, and cooperating in the physical search by making all possible facilities available for use by the searching agencies.