What kinds of problems are suitable to be solved by desktop grid environment? Alexander Afanasyev The Institute for Information Transmission Problems (Kharkevich.

Slides:



Advertisements
Similar presentations
Neural and Evolutionary Computing - Lecture 4 1 Random Search Algorithms. Simulated Annealing Motivation Simple Random Search Algorithms Simulated Annealing.
Advertisements

Computational Modeling for Engineering MECN 6040
From Quark to Jet: A Beautiful Journey Lecture 1 1 iCSC2014, Tyler Dorland, DESY From Quark to Jet: A Beautiful Journey Lecture 1 Beauty Physics, Tracking,
Tunneling Phenomena Potential Barriers.
Fundamentals of Data Analysis Lecture 12 Methods of parametric estimation.
Chapter 4: Network Layer
Time Domain Analysis of the Multiple Wires Above a Dielectric Half-Space Poljak [1], E.K.Miller [2], C. Y. Tham [3], S. Antonijevic [1], V. Doric [1] [1]
Chapter 3 Steady-State Conduction Multiple Dimensions
Chapter 1 Introduction The solutions of engineering problems can be obtained using analytical methods or numerical methods. Analytical differentiation.
The Advanced Chemical Engineering Thermodynamics The retrospect of the science and the thermodynamics Q&A -1- 9/16/2005(1) Ji-Sheng Chang.
Force Directed Algorithm Adel Alshayji 4/28/2005.
Ant Colony Optimization Optimisation Methods. Overview.
1/36 Gridless Method for Solving Moving Boundary Problems Wang Hong Department of Mathematical Information Technology University of Jyväskyklä
Force Directed Algorithm Adel Alshayji 4/28/2005.
ON MULTIVARIATE POLYNOMIAL INTERPOLATION
Joo Chul Yoon with Prof. Scott T. Dunham Electrical Engineering University of Washington Molecular Dynamics Simulations.
Component Reliability Analysis
Tennessee Technological University1 The Scientific Importance of Big Data Xia Li Tennessee Technological University.
Javier Junquera Molecular dynamics in the microcanonical (NVE) ensemble: the Verlet algorithm.
Kurochkin I.I., Prun A.I. Institute for systems analysis of RAS Centre for grid-technologies and distributed computing GRID-2012, Dubna, Russia july.
ORGANIZING AND ADMINISTERING OF VOLUNTEER DISTRIBUTED COMPUTING PROJECT Oleg Zaikin, Nikolay Khrapov Institute for System Dynamics and Control.
Molecular Dynamics Simulation Solid-Liquid Phase Diagram of Argon ZCE 111 Computational Physics Semester Project by Gan Sik Hong (105513) Hwang Hsien Shiung.
Scheduling Many-Body Short Range MD Simulations on a Cluster of Workstations and Custom VLSI Hardware Sumanth J.V, David R. Swanson and Hong Jiang University.
A Grid fusion code for the Drift Kinetic Equation solver A.J. Rubio-Montero, E. Montes, M.Rodríguez, F.Castejón, R.Mayo CIEMAT. Avda Complutense, 22. Madrid.
Lecture 8 Numerical Analysis. Solution of Non-Linear Equations Chapter 2.
LECTURE 13. Course: “Design of Systems: Structural Approach” Dept. “Communication Networks &Systems”, Faculty of Radioengineering & Cybernetics Moscow.
1 A method of successive elimination of spurious arguments for effective solution the search- based modelling tasks Oleksandr Samoilenko, Volodymyr Stepashko.
Bashkir State Univerity The Chair of Mathematical Modeling , Ufa, Zaki Validi str. 32 Phone: ,
Chemical Reactions in Ideal Gases. Non-reacting ideal gas mixture Consider a binary mixture of molecules of types A and B. The canonical partition function.
1 Complex Images k’k’ k”k” k0k0 -k0-k0 branch cut   k 0 pole C1C1 C0C0 from the Sommerfeld identity, the complex exponentials must be a function.
Serge Andrianov Theory of Symplectic Formalism for Spin-Orbit Tracking Institute for Nuclear Physics Forschungszentrum Juelich Saint-Petersburg State University,
Mobile Agent Migration Problem Yingyue Xu. Energy efficiency requirement of sensor networks Mobile agent computing paradigm Data fusion, distributed processing.
Time-dependent Schrodinger Equation Numerical solution of the time-independent equation is straightforward constant energy solutions do not require us.
Silesian University of Technology in Gliwice Inverse approach for identification of the shrinkage gap thermal resistance in continuous casting of metals.
BOINC Projects promotion – experience Oleg Zaikin Institute for System Dynamics and Control Theory SB RAS, Irkutsk, Russia.
Gaussian Mixture Models and Expectation-Maximization Algorithm.
Ant Algorithm and its Applications for Solving Large Scale Optimization Problems on Parallel Computers Stefka Fidanova Institute for Information and Communication.
1 IDGF International Desktop Grid Federation PORTING APPLICATION TO THE BOINC DESKTOP GRID: ISA RAS EXPERIENCE Mikhail Posypkin Oleg Zaikin Nikolay Khrapov.
Chapter 2-OPTIMIZATION G.Anuradha. Contents Derivative-based Optimization –Descent Methods –The Method of Steepest Descent –Classical Newton’s Method.
Nonlinear differential equation model for quantification of transcriptional regulation applied to microarray data of Saccharomyces cerevisiae Vu, T. T.,
Structural alignment methods Like in sequence alignment, try to find best correspondence: –Look at atoms –A 3-dimensional problem –No a priori knowledge.
Lecture 39 Numerical Analysis. Chapter 7 Ordinary Differential Equations.
Optimal Relay Placement for Indoor Sensor Networks Cuiyao Xue †, Yanmin Zhu †, Lei Ni †, Minglu Li †, Bo Li ‡ † Shanghai Jiao Tong University ‡ HK University.
The Search for Systems of Diagonal Latin Squares Using the Project Oleg Zaikin, Stepan Kochemazov Matrosov Institute for System Dynamics and Control.
Solving Weakened Cryptanalysis Problems for the Bivium Keystream Generator in the Volunteer Computing Project Oleg Zaikin, Alexander Semenov,
CluBORun: tool for utilizing idle resources of computing clusters in volunteer computing. Maxim Manzyuk, Oleg Zaikin, Mikhail Posypkin Volgograd state.
the project of the voluntary distributed computing ver.4.06 Ilya Kurochkin Institute for information transmission problem Russian academy of.
Grid computing simulation of superconducting vortex lattice in superconducting magnetic nanostructures M. Rodríguez-Pascual 1, D. Pérez de Lara 2, E.M.
Nizhni Novgorod State University Faculty of Computational Mathematics and Cybernetics Searching Globally-Optimal Decisions for Multidimensional Optimization.
LabVIEW Real Time for High Performance Control Applications
Our task is to estimate the axial displacement u at any section x
Design and Analysis of Algorithms (09 Credits / 5 hours per week)
Chapter 7. Classification and Prediction
Parallel Plasma Equilibrium Reconstruction Using GPU
Joint Institute for Nuclear Research, Dubna, Russia
Scalable Load-Distance Balancing
Oleg Zaikin, Alexander Semenov, Mikhail Posypkin
Virtual laboratories in cloud infrastructure of educational institutions Evgeniy Pluzhnik, Evgeniy Nikulchev, Moscow Technological Institute
Boundary Element Analysis of Systems Using Interval Methods
International Workshop
Ordinary Differential Equations
Chapter 3 Component Reliability Analysis of Structures.
Xin-She Yang, Nature-Inspired Optimization Algorithms, Elsevier, 2014.
1 Department of Engineering, 2 Department of Mathematics,
Chapter 27.
1 Department of Engineering, 2 Department of Mathematics,
1 Department of Engineering, 2 Department of Mathematics,
High Performance Computing & Bioinformatics Part 2 Dr. Imad Mahgoub
Load forecasting Prepared by N.CHATHRU.
Quantum One.
Presentation transcript:

What kinds of problems are suitable to be solved by desktop grid environment? Alexander Afanasyev The Institute for Information Transmission Problems (Kharkevich Institute) RAS Moscow BOINC:FAST’2015 Petrozavodsk, Russia, September 14-18, 2015

Supercoputers Grids and Clouds Desktop grid

DEGISCO WP4 29/09/ For which tasks are GSPC? A large amount of calculations (a few days of calculations on a PC and more). Weakly-related tasks (problem is decomposed into independent parts that do not require interaction in the process of calculations ).

Tasks combinatorial and global optimization

Applications Combinatorics Cryptanalysis of ciphers Search Latin squares Test mathematical hypotheses Verification Optimization Docking proteins The analysis of the spatial configuration of chemical compounds Identification of model parameters

DEGISCO WP4 29/09/ the Most successful project of GSPC in Russia The project is a research project aimed at solving complex scientific problems reducible to the problem of inversion of Boolean functions Peak performance of 10 TFlops ESTO RAS IITP

Search configuration of atoms with a minimum of interaction energy - the distance between atoms i and j - the pair potential ; - the potential of a Lennard Jones; - the Morse potential.

The organization of calculations in the BOINC environment Selection variants and generation of new jobs The server clients

The results of the experiment The problem of finding the minimum of the cluster with the Morse potential (rho = 14), 150 atoms has been solved. There were able to get the minimum value in 10 days calculations. The result corresponds to the best found so far value using a geometrically-based methods. The improved results obtained previously using the methods of General use.

Potential Tersoff Cohesive energy E coh of a system of atoms - the potential energy of interaction between two particles i and j, where - cutoff-fnction, Members are responsible for the repulsion and attraction: Multi-particle interaction parameter

The objective function for the optimization problem For regular repeating lattice of atoms to measure a number of parameters with the potential Tersoff: -cohesive energy, elastic constants The objective function is written as follows, where - free potential parameters Tersoff

The results of the experiment 3 days calculation

The method of branches and borders TREE BRANCHBRANCHING SUBPROBLEM OUT SUBPROBLEM: 1. HAS NO SOLUTIONS 2. THE OPTIMUM WAS FOUND 3. THE OPTIMUM IS NOT BETTER THAN PREVIOUSLY FOUND (RECORD )

The scope of application of the method of branches and borders Method of branches and borders lies at the heart of many deterministic optimization algorithms Deterministic methods guarantee the accuracy of the solutions and are used in the solution of problems where this guarantee is essential

Method of branches and boundaries for GSPC The main problem is different, unknown in advance, the complexity of the subproblems. Therefore, an efficient distributed implementation is more complicated than in the case of iterative algorithms. Require load balancing, which in the case of GSPC is reduced to the proper formation of computational units, taking into account a priori estimates of the complexity of the subproblems.

The problem arises in the analysis of carbon nanostructures that arise in the form of deposits on the walls of tokamaks. The tokamak (toroidal chamber with magnetic coils) is the toroidal unit for magnetic confinement of plasma to achieve the conditions necessary for the flow of controlled thermonuclear fusion. The analysis of these structures is a typical big data problem.

17 L2 optimization R=1 nm, Rx=0.3 nm, Ry=0.5 nm R=1.05 nm, Rx=0.3 nm, Ry=0.5 nm R=1.1 nm, Rx=0.3 nm, Ry=0.5 nm R=0.95 nm, Rx=0.3 nm, Ry=0.6 nm R=1 nm, Rx=0.3 nm, Ry=0.6 nm R=1.2 nm, Rx=0.3 nm, Ry=0.65 nm Atoms: CHCr, Fe, NiNb, MoTl Chemical contents of amorphous medium taken from experiment Structural analysis of carbon films. The mixture of structures.

18 Визуализация результатов Суммарный график

Supercoputers Grids and Clouds Desktop grid

MathCloud: Architecture

Cauchy problem for differential equations with polynomial right-hand sides Where - n-dimensional polynomial m-th order with constant coefficients.

Quasi-symbolic representation of the approximate solution of the Cauchy problem for systems of ordinary differential equations with polynomial right-hand sides in the form of a segment of a Taylor series n - the order of system control m - degree polynomial right-hand side of the multidimensional control i - the serial number of the polynomial T x0 in the Taylor series expansion (Ti (x0)) im-i +1 - the degree of Ti (x0) nim-i +2 - the number of coefficients in the i-th term of the Taylor expansion i - number of expansion terms

Lorenz System under standart parameters σ =10, r=28,b =8/3

crunchers corporative computing computers crowdsourcing

THANK YOU!

Технологии ГСПК BOINC (Berkeley Open Infrastructure for Network Computing) OurGrid XtremWeb-HEP (XWHEP)

BOINC (Berkeley Open Infrastructure for Network Computing) Самая популярная платформа для добровольных вычислений Поддерживается всеми видами OS и Android Средняя суммарная реальная производительность всех BOINC-проектов на примерно 18 PetaFLOPS (по данным

Collatz’ Conjecture Tests the Collatz conjecture the Collatz Conjecture (hypothesis 3n+1, the hypothesis 3x+1, of Collatz problem, the problem of 3n+1, the problem 3x+1, Syracuse problem) is one of the unsolved problems of mathematics, named after the German mathematician Lothar Collatz (English), who proposed it in The average performance of 2.5 PFlops (according to

The solution of combinatorial problems in project is a research project that uses connected via the Internet of computers to solve hard and practically important problems (discrete functions, discrete optimization, bioinformatics, etc.) that can be effectively reduced to the problem of satisfiability of Boolean formulas. At this point in the project is analyzed cipher Bivium (a weakened version of the cipher Trivium). Peak performance of 10 TFlops. An average of 1,000 active hosts from connected to the project.

Алгоритм решения: псевдо-локальный поиск

Алгоритм решения: глобализация Улучшение Отбор лучших

Компоненты приложения Сервер Клиент Скрипт, осуществляющий генерацию новых заданий и обработку полученных (periodic task) Программа, осуществляющая полу-локальный поиск, интегрированная с boinc- wrapper.

Задачи идентификации параметров модели