September 2011 - Allocations-admins - mailman3.cels.anl.gov

New LCRC report submitted for ApsRenewalLattice
by support＠lcrc.anl.gov 06 Sep '11

06 Sep '11

Full project report at https://accounts.lcrc.anl.gov/reporting/project/2011/?project=ApsRenewalLat…

1 0

New LCRC report submitted for ApsRenewalLattice
by support＠lcrc.anl.gov 06 Sep '11

06 Sep '11

Full project report at https://accounts.lcrc.anl.gov/reporting/project/2011/?project=ApsRenewalLat…

1 0

[LCRC Accounts] Yearly Allocation Request from Nek5000
by accounts＠lcrc.anl.gov 06 Sep '11

06 Sep '11

Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Shashi Aithal Project Name: Nek5000 Division: MCS Project title: Scalable CFD Algorithms Associated funding: Applied Math Research Program, Nuclear Energy Advanced Modeling and Simulation Other Systems: ALCF Intrepid (BG/P), 128-core Linux cluster, Cosmea Science: Our mission is to develop and deploy scalable algorithms for simulation-based science applications that rely on numerical solution of partial differential equations. Our primary vehicle for algorithm development is the fluid/thermal simulation code, Nek5000, and its off-shoots, NekCEM (for computational electromagnetics) and NekLBM (for lattice Boltzmann methods). There are approximately 30 research groups worldwide currently using Nek5000. We use Nek5000---and the science applications brought to us through our user community---as vehicles for the development and test of new numerical algorithms that include stabilized high-order methods, scalable linear solvers, advanced timestepping methods, and so forth. We emphasize that these algorithms cannot be developed in isolation from applications, as many of the relevant issues do not arise at small scale or in simple model problems devoid of complex physics and/or geometry. Thus, to address algorithmic issues at the forefront of science-based simulation we require a code that is capable of solving problems _at_scale_ and that is sufficiently flexible to allow adaptation to new physics. Project description: Nek5000 is a petascale fluid/thermal simulation code used for a host of applications including reactor thermal hydraulics, combustion, MHD, oceanography, spatiotemporal chaos, and vascular flow modeling. Nek5000 is a grid-based PDE (partial differential equation) solver in three space dimensions and time. The code has long been at the forefront of high-order methods and scalable algorithms. It was one of the first comprehensive software tools developed for distributed memory parallel platforms. It pioneered the development of high-order discretizations that, because of their ability to accurately capture interacations over a broad range of scales, are now deemed an essential ingredient of petascale simulation science and that are being adapted across virtually all disciplines (e.g., fusion, aeronautics, biofluids and geophysics). We have demonstrated > 80% parallel efficiency on BG/P (strong scaling) out to P=131,072 processors, using only 7300 points/processor at the largest scale. We note that we have access to a dedicated 128-core linux cluster (under a nuclear engineering project) and to resources on BG/P through INCITE and discretionary allocations. This request is to address mid-range capabilities -- particularly for those applications that would not fit on the 128-core cluster as discussed below. Our goal is to use Fusion to test the performance of mid-size problems (with 10,000 to 100000) elements with various compilers and compiler options. We also intend to set up a suite of large problems to ensure that various builds are working correctly. We also intend to run production cases for problems in nuclear engineering and conjugate heat-transfer. Project URL: http://www.mcs.anl.gov/~fischer/sem1b Current FY Hours Used: undetermined amount New FY Requested allocation: 95000 Q1: 25000 Q2: 25000 Q3: 25000 Q4: 20000 Justification: Thank You, The LCRC Accounts System

1 0

[LCRC Accounts] Yearly Allocation Request from uncertainty-climate
by accounts＠lcrc.anl.gov 31 Aug '11

31 Aug '11

Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Emil M Constantinescu Project Name: uncertainty-climate Division: MCS Project title: Uncertainty in weather forecasting for energy systems Associated funding: Other Systems: NFS @ MCS Science: The theoretical part of the project consists in developing uncertainty models for numerical weather and climate prediction systems for use in the operation and planning of energy systems. Fusion is used to run several scenarios and obtain forecasts and ensemble-based uncertainty information about ambient variables by running weather/climate simulation codes. We also perform adjoint sensitivity analyses that indicate suitable locations for meteorological sensor siting. This information is used in energy management applications such as stochastic optimization control loops that provides optimal operational point for energy systems. A secondary objective is to develop robust data assimilation methods (i.e., correct model predictions with observations) in order to calibrate weather and climate forecasts for effective operation and planning of energy systems. Project description: The aim is to create very accurate uncertainty estimates of the weather and climate forecasts to be used for optimal management of energy systems such as wind- and solar-power generation, building management and power grid planning. The purpose of using Fusion is to execute several scenarios and obtain forecast and uncertainty information about ambient variables by running high resolution weather/climate simulation codes (such as WRF) as well as adjoint models that provide sensitivities with quantities of interest. This information is used in a stochastic optimization control loop that provides an optimal operational point for energy systems. For the practical part of the project, we developed (in FY11) and will enhance (in FY12) an uncertainty quantification framework (C or FORTRAN, MPI, XML, Python) to run Weather Research and Forecasting (WRF) automatically and in a realistic fashion in order to illustrate theoretical findings and provide realistic results. We consider running several WRF instances (each on several node) at the same time. Partial results are collected on a separate node and processed before they are sent back to each specific instance. The instances will continue to run and produce new results. In terms of parallelism, this application framework parallelizes easily since each instance run independently of each other. Moreover, each instance evolves in time using a local spatial stencil and explicit-in-time algorithms, which yields very good scalability. In FY12, we will to run sensitivity analyses for economic cost functions and perform data assimilation experiments with high resolution and obtain more samples in order to achieve a better level of accuracy. The improved weather forecast accuracy simulations allow us to find a more robust operational point for the operation of energy systems (wind power for electricity generation and building heating and cooling). Project URL: Current FY Hours Used: undetermined amount New FY Requested allocation: 325000 Q1: 75000 Q2: 75000 Q3: 75000 Q4: 100000 Justification: Thank You, The LCRC Accounts System

1 0