Allocations-admins

Download

[email protected]

August 2012

5 participants
152 discussions

[LCRC Accounts] Project Allocation Request
by accounts＠lcrc.anl.gov 04 Aug '12

04 Aug '12

Hello, A change in allocation has been requested: Requester: jiechen (Jie Chen) Project: ScalaGAUSS Title: Scalable Gaussian Process Data Analysis Description: THIS PROJECT WAS STARTED CLOSE TO THE END OF FY2011 AND ONLY A SMALL NUMBER OF CORE-HOURS WERE ALLOCATED. WE ARE REQUESTING AN ADDITIONAL AMOUNT OF 500000 CORE-HOURS FOR FY2012. We model data as Gaussian process model using several covariance models, including a Matern covariance kernel. The goal of the computation is to estimate the parameters of the covariance kernel. We employ a maximum likelihood method, which is reformulated by using a sample average approximation approach. The novelty of the approach is that it allows for scalable analysis without compromise in the accuracy of the correlation. This reduces to solving linear systems that have the same size of the grid from which the data are sampled. The linear systems have multiple right-hand sides, and in the case of a regular grid, they are multilevel Toeplitz matrices. We employ a block CG solver with multilevel circulant preconditioners. Previous experiments indicate a linear scalability of computational costs for 2D grids of size up to $10^3times10^3$ on a single desktop machine in a Matlab environment. We have demonstrated that up to 1 million data points our approach scales virtually pe rfectly. We have developed parallel C++ codes based on the Trilinos project and will investigate the scalability of the algorithm for 4D data set with on the order of $10^9-10^{12}$ data points; beyond the $10^6$ data sets we can manipulate at the moment. For being able to compile our code, we need (in addition to standard compilers) Trilinos, and ideally also PETSC. The ScalaGauss code can be used in numerous other applications (chemistry, power grids, materials science) that require a Gaussian process statistical models. Current: undetermined amount Justification: Requested: 100000 A specific reason has been given: This request is because of the coding development during the past two months. New codes have been tested and we expect that additional core-hours are needed for the month of August and September. Roughly 100000 core hours were used in the past two months, we therefore request the same amount of resources for the remaining two months. This needs to be approved and the final allocation amount decided upon. Thank You, The LCRC Accounts System

1 0

LCRC Project Allocation Request - Sumer - Multicomp_Nanopart
by Bair, Raymond A. 01 Aug '12

01 Aug '12

Dear LCRC Allocations Committee, Below is a request for 200K core-hours for the current quarter. Since it is >150K we need your approval, or your recommendation for an alternate allocation amount. A prompt response would be appreciated. Regards, Ray On 7/31/12 11:13 AM, "Aslihan Sumer" <asumer(a)anl.gov> wrote: >Dear Ray, > >Thank you for prompt reply. We are currently using 100,000 cpu-hours per >month and we plan to use all the 200,000 cpu-hours till the end of >September, 2012. > >Regards, > >-Aslihan > >On Jul 31, 2012, at 11:05 AM, Bair, Raymond A. wrote: > >> Dear Aslihan Sumer, >> >> Thank you for the detailed request. Do you plan to use the 200K core >>hours all in this quarter (July-Sept)? >> >> Regards, >> >> Ray Bair >> >> ----- Original Message ----- >> From: accounts(a)lcrc.anl.gov [mailto:[email protected]] >> Sent: Tuesday, July 31, 2012 11:01 AM >> To: allocations-admins(a)lcrc.anl.gov <allocations-admins(a)lcrc.anl.gov> >> Subject: [allocations-admins] [LCRC Accounts] Project Allocation Request >> >> Hello, >> >> A change in allocation has been requested: >> >> Requester: asumer (Aslihan Sumer) >> Project: Multicomp_Nanopart >> Title: Next Generation of Multicomponent Nanocatalysts: >>Computational Studies >> Description: The planned computations will explore the structural >>(e.g., mixed, core-shell, dumbbell-like, and other), energetic, >>dynamical/thermal, electronic, and chemical/catalytic properties and >>characteristics of multicomponent metal nanoparticles, both gas phase >>and deposited on supports. Among the targeted nanocatalysts are Pt-Co >>(the work on this system has been initiated already), Pt-Ni, Pt-Fe, >>CoPt3/Au, FePt/Au, Au/Fe2O3, CoPd2/Au, and others. These are the one >>currently investigated or planned for investigation by the experiments. >>The candidate supports include iron-oxide and cobalt-oxide. >> >> Among the issues that will be addressed are: 1) Use of size, structure >>and composition of the nanocatalysts as "knobs" for tuning their >>activity and selectivity, and 2) Use of functionalizing ligands such as >>amines of different "sizes" (different lengths of the carbon chains) as >>an additional means of affecting the catalytic functionality >>(experiments on the effects of functionalization of various >>multicomponent nanocatalysts by amines are currently underway). The >>computational studies of the structural, energetic, and electronic >>aspects will be performed using density functional theory. The packages >>that will be utilized include NWChem and Crystal (we have recently >>purchased the license for Crystal and installed the software on the >>Fusion cluster). Naturally, we will take full advantage of the >>parallelized versions of these packages as we intend to explore systems >>with an ever increasing number of atoms: 1) The properties of the >>nanocatalysts will be explored over the size r >> ange from a few tens to a few hundreds of atoms, and 2) The effects of >>the supports will later be incorporated as well. In addition to the DFT >>studies, we also plan to perform large scale molecular dynamics >>simulations of the dynamical and thermal behavior of the multicomponent >>catalytic nanosytems. The expected number of users on this project is 3 >>or more. >> Current: undetermined amount >> Justification: The allocation request is based on estimates made with >>NWChem on Fusion and on NERSC (Carver and Hopper) machines. Tests on a a >>relatively small Pt14 cluster with the new Stuttgart pseudopotential and >>a basis set of 38 contracted Gaussians per atom with a 2-butyne or >>2-butene molecule on it performed with 64, 128 and 192 processors >>indicate a substantial increase in performance with the increase in the >>number of processors up to about 128. Tests with Crystal are in >>progress, however literature data indicates high efficiency of the >>parallelization of this code. >> >> 1) Scalability of NWChem on Fusion as evaluated by us for the C4H6-Pt14 >>system (total of 736 basis functions and with charge density fitting): >> Cores Average SCF cycle (s) >> 64 19.4 >> 128 12.5 >> 192 10.8 >> >> 2) Scalability of MPP_Crystal on Hector (UK) as evaluated for the >>Ga432Sb426N6 system (19836 basis functions) [1]: >> Cores Average SCF cycle (s) >> 856 2503 >> 1712 1308 >> 3424 711 >> >> [1]I. J. Bush et al., Proc. Royal. Soc. 467, 2112 (2011) >> >> >> Requested: 200000 >> >> A specific reason has been given: >> One the issues that is addressed in this project is the relationship >>between the size of metal nanoparticles and their efficiency in >>selected model catalytic transformations, which includes the selective >>hydrogenation of unsaturated aldehydes, for example. The computational >>studies (within the framework of density functionals) performed to date >>include nanoparticles ranging from a few atoms to a few tenths of atoms. >>Our results provide important information on trends in catalytic >>activity and selectivity at this subnanometer size range. In order to >>provide a complete picture of the size effect, the ongoing studies have >>recently been expanded to include nanoparticles of a few hundreds of >>atoms. The increased complexity of systems at this size (i.e., the >>larger number of possible adsorption sites to be tested in order to >>reach the minimum energy adsorption configurations of the reactants on >>the nanoparticles), along with the computational efficiency requirements >>that increase >> in proportion with the number of atoms in the system (our scaling >>studies showed that maximum efficiency is obtained with 64-128 >>processors for the systems with approximately 300 atoms that are >>currently under investigation), make it necessary to request this >>additional allocation for the current fiscal year. >> >> This needs to be approved and the final allocation amount decided upon. >> >> Thank You, >> The LCRC Accounts System >> _______________________________________________ >> allocations-admins mailing list >> allocations-admins(a)lists.lcrc.anl.gov >> https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins >

1 0

Jump to page: