Symposium on
Z
HPC Applications
Y
X
March 12 - 14, 2012
Indian Institute of Technology Kanpur
ABSTRACT BOOK
[email protected]
Z
Y
X
Symposium on
March 12-14, 2012
IIT, Kanpur
http://www.iitk.ac.in/cc/symphpc
Symposium on HPC Applications @IITK
Simulation of hypersonic flows in real-life configurations
Krishnendu Sinha
IIT Mumbai
Hypersonic flows are characterized by strong shock and expansion fans, with large changes in fluid properties in
these regions. The high temperature associated with strong shock waves excites internal energy modes of the gas
molecules. Gas density in some regions of the flow can be too low for the thermo-chemical processes to reach
equilibrium. Finite rate model are therefore required, along with the Navier-Stokes equations, to simulate the
flow physics. In addition presence of unsteady, transitional or turbulent fluctuations increases the range of length
and time scales drastically. Computing such flows in geometrically complex real life configuration is a
challenging task. In house CFD codes based on robust numerical algorithm and three-dimensional multi-block
structured meshes are being routinely used at IIT Bombay to simulate flows of practical relevance. A dedicated
200 processor-core linux cluster is set up for this purpose. Two test cases will be presented, namely, wake flow
behind a re-entry capsule at Mach 16 and three-dimensional shock-boundary layer interaction in a scramjet
intake.
The metrics of high performance computing
Tapan Sengupta
IIT Kanpur
In this talk we identify the numerical aspects of scientific computing which render these aspects, the sobriquet of
high performance computing. Additionally, HPC also implies the usage of parallel computers, which demands
new class of methods. Some of these methods developed at HPCL, IIT Kanpur will be discussed. We will discuss
all of the above with respect to the direct numerical simulation of turbulent flows from disturbance onset stage.
We finally focus on the emerging issues of HPC, with respect to methods, efficient implementation and error
control, all of which contribute to the eventual success of high performance computing.
Numerical dynamo models: progress and challenges
Binod Sreenivasan
IIT Kanpur
Advances in computational ability have enabled us to perform three-dimensional computer simulations of
planetary dynamos. Geodynamo simulations provide realizations of geomagnetic field features such as the largescale dipolar structure, historical time-changes of the magnetic field, lower-mantle effects and polarity reversals.
In this talk I will summarize the progress made in numerical modelling of the Earth's dynamo and some of the
challenges that lie ahead.
Use of a Hybrid WENO-Compact Scheme for Study of Laminar Shock
Wave Boundary Layer Interactions at Hypersonic Speeds
Manoj Nair, Ramakrishnan R., Arun Kumar K.T.
CSIT-NAL Bangalore
Accurate prediction of loads in the presence of shock wave boundarylayer interaction (SWBLI) is one of the
major challenges in theuse Computational Fluid Dynamics (CFD) as a tool for hypersonic vehicledesign. These
interactions can lead to presence of massive flowseparation, drag rise and shock unsteadiness leading to
deterioration ofper formance of the vehicle. High wall heating in regions of theseinteractions adversely effect the
structure and needs additional protection.The TVD-MUSCL reconstruction used traditionally are secondorderaccurate and known to reduce to first-order in the presence ofshocks due to limiters. Prediction of loads in
the presence of theseinteractions with the traditional second-order shock capturing methodshas not been very
successful - particularly in predicting the wall heattransfer rate. Higher-order schemes are necessary for
accurateprediction of physical phenomena in such flows. The high order methodsused must have good shock
capturing ability and fine scale resolution.Compact schemes have been successfully used in DNS/LES of
turbulencesimulations. They are more appropriate in resolving short waves and arenot suitable to capture shocks
as they induce large oscillations neardiscontinuities. Weighted Essentially Non-Oscillatory (WENO) scheme is a
shockcapturing scheme which has third-order accuracy near the shock and higheraway from it. However, WENO
is too dissipative for small length scalesand therefore for DNS/LES. Hybrid methods, which use a WENO scheme
neardiscontinuities and low dissipative compact scheme away fromdiscontinuities are good candidates for
DNS/LES of flows with shocks. Aneffective shock detector is to be used which would avoid use of
compactschemes across the shock and mix of both compact and WENO near the shock.In the present work such a
hybrid method is used to study the phenomena oflaminar shock wave boundary layer interaction in ramps. A fifthorderWENO scheme along with a fourth-order compact scheme is used for thesimulations. Four different ramp
angles have been used in theinvestigation.
CFD of Human Airways
Anuj Jain
Professor, Department of Applied Mechanics
Motilal Nehru National Institute of Technology, Allahabad-211004, India.
With the increase of air pollution level in various Indian cities, the breathing troubles are on the rise, especially
among urban population. Deposition of inhaled allergens causes acute broncho-constriction for asthma sufferers.
Also elevated deposition of some particulate pollutant causes tissue injuries due to peak values of wall shear
stress at these locations. Over a period of time, this condition leads to various chronic obstructive pulmonary
diseases (COPD) diseases. One of the popular clinical treatment for these diseases is administration of noncongestive drugs (in aerosol form) using inhalers/rotahalers into the respiratory system via mouth. However, it is
found that the drug may not reach to the infected/injured areas of the respiratory tract in many cases. As a result,
the healing of the disease gets delayed. In this context, there is a need to understand airflow in the realistic human
airways and obtain the deposition of toxic/therapeutic micro-particles under different (viz. normal and heavy)
breathing conditions and identify the local 'hot-spots' where the particles are concentrated/deposited in the human
bronchial airways. The correlation of the flow parameters with the medical aspects of pulmonary diseases will
help to understand the causes and remedies of pulmonary diseases which may be potentially applicable for better
diagnosis, prognosis, treatment and prevention of these diseases. Therefore, it proposed to analyze the human
airways as a bifurcating airflow problem and obtain the flow parameters (velocity, wall shear stress and pressure
distributions) computationally using Computational Fluid Dynamics (CFD) simulation.
A human airways consists nasal cavity, trachea and bronchus which appears like a combination of Y-shaped
bifurcated ducts. Human airways are irregular, asymmetrical in shape and are complex due to the presence of
cartilaginous rings. The function of the human airways is to supply oxygen and dispose the carbon dioxide
between atmosphere and lungs. The flows through this passage are cyclic in nature. In normal breathing
condition, airflow is laminar, but during heavy breathing condition (exercise, running etc.), it is turbulent.
Generation of 3D model from CT scan data and computational mesh require significant amount of computational
time and efforts. The variation in Hounsfield number is used to identify the respiratory tract in the CT scan image
of transverse plane of the body in MIMCS software. Inner wall of the respiratory tract is picked up as a curved
fitted ring for each slice. The 3D model of the airways is reconstructed by generating surface along these curvefitted rings. In order to make the 3D model of the airways realistic, the slices are required at the closest possible
distance. If two consecutive slices are used at a gap of 0.625 mm, 200 slices of CT scan images would be required
for the reconstruction of realistic airways from trachea to third generation of bronchus of about 13 cm length.
The unstructured mesh is required to capture the flow distribution near the irregular shaped wall. Fine mesh is
required near the wall and the bifurcation to capture the flow physics at increased resolution at these locations of
interest. It was found that minimum 2 million elements are required up to first generation bronchus for good
computational results. However, as it moves from 2nd to 6th generation about 6 million
or more elements are needed.
Two-phase simulation involves solution of additional equations related to
toxic/therapeutic micro-particles. Respiration is pulsatile flow in which about 500 ml
of vital volume is inhaled and exhaled in each cycle. In normal breathing condition, a
single cycle takes 4-5 seconds corresponding to a flow rate of 6-8 l/min.
The typical run time of first generation model for the single phase simulation (with
inspiratory flow) on Xeon quad-core processor with 8GB memory is approximately 16
hours. The dynamic simulation of particle transport for one complete breathing cycle is
envisaged to take several days on the facility available with the department. A HPC
system will be helpful in handling the problem.
Research publications:
1. Vivek Kumar Srivastav, Anuj Jain. Wall Shear Stress Study of Airflow in Realistic Human Airways
Model: A CFD Analysis. Advances in Modeling, Optimization and Computing (AMOC-2011). Dec. 5-7,
2011, IIT Roorkee, India. pp. 362-367, ISBN: 9788186224748.
2. Vivek Kumar Srivastav, Anuj Jain, Akshoy Ranjan Paul, Shrey Joshi. CFD Modelling of Airflow in Human
Respiratory System. International Conference on Mechanical Engineering (ICME-2011), Dec.18-20, 2011,
Dhaka, Bangladesh. Paper no. ICME-11-FL-009.
3. Vivek Kumar Srivastav, Amit Kumar, Anuj Jain, Akshoy Ranjan Paul. CFD Modelling of Bifurcated
Human Airways. 38th National Conference on Fluid Mechanics and Fluid Power (NCFMFP-2011), Dec. 1517, 2011, MANIT, Bhopal. Paper no. FMFP-11- CFD-07.
4. Vivek Kumar Srivastav, Amit Kumar, Akshoy Ranjan Paul, Amba D Bhatt, Anuj Jain. CFD Study of Airflow
in Human Respiratory Tract with Tumor. International Congress on Condition Monitoring and Diagnostic
Engineering Management (COMADEM-2012), June 18-20, 2012. University of Huddersfield, UK.
(Accepted).
Dynamics of a polyelectrolyte in simple shear flow: A comparison
between Brownian Dynamics and Dissipative Particle Dynamics
simulations.
Jayasree Kandiledath, Rajkuma Manna, P. B. Sunil Kumar
IIT Madras
The configurational dynamics of a polyelectrolyte (PE), subjected to a simple shear flow is studied using
Brownian dynamics (BD) and dissipative particle dynamics (DPD) simulations of a bead-spring model with
explicit counterions. We explore the tumbling and extension of PEs by varying the shear rates for a range of values
of the electrostatic coupling parameter $A$ (which is defined as the ratio of the Bjerrum length to the size of the
monomer). The PE exhibits extensional and tumbling dynamics. We observe that the dynamic behavior of the PE
at lower values of $A$ is very similar to that of a semiflexible polymer, with the tumbling frequency being
dependent on the hydrodynamic interaction (HI). For $A>15$, we observe a critical shear rate below which the
PE remains in the globular state with a structure akin to that of a neutral polymer in poor solvent. The behavior of
the PE above the critical shear rate is dependent on the HI. In the absence of HI, the PE extends as a whole and then
collapses by the formation of folds with no observable periodicity in tumbling. When the HI is included, the
counterion fluctuations are suppressed, and the polymer now extends through a protruding end and rolls back in a
manner similar to that of a neutral polymer. Simulation artifacts resulting from implicit nature of solvent in BD
and that due to boundary conditions, are discussed by comparing the BD results with that from a dissipative
particle dynamics simulations.
Investigations on Protein Membrane Interactions
K. G. Ayappa
IISc Banglore
Understanding the interaction of proteins with membranes is important in the development of gene and drug
delivery systems, assessing toxicity and developing a molecular understanding of interaction pathways. In this
talk I will discuss our recent efforts to study the structure of membranes and the interaction of a special class of
proteins, known as pore forming toxins with biological membranes using atomistic as well as coarse grained
simulation methods. Pore forming toxins are expressed by many strains of bacteria and mount their attack on
mammalian cells by forming stable membrane pores leading ultimately to cell death. The process of pore
formation consists of the transformation from a water soluble protein to a membrane bound form. At the cell
membrane, the protein oligomerizes to form a stable transmembrane pore complex which eventually destroys the
cell. The crystal structure of only a few pore forming toxins are known and the mechanism of pore formation is
only partially understood. We have carried out fully atomistic (100 ns) and coarse grained simulations (1
microsecond) of the dodecamer pore complex of the ClyA pore forming toxin. I will illustrate the need to carry
out long time molecular dynamics simulations as well as the issues associated with such simulations.
Topological Insulators: A New Class of Materials
R. Prasad
IIT Kanpur
Understanding the interaction of proteins with membranes is important in the development of gene and drug
delivery systems, assessing toxicity and developing a molecular understanding of interaction pathways. In this
talk I will discuss our recent efforts to study the structure of membranes and the interaction of a special class of
proteins, known as pore forming toxins with biological membranes using atomistic as well as coarse grained
simulation methods. Pore forming toxins are expressed by many strains of bacteria and mount their attack on
mammalian cells by forming stable membrane pores leading ultimately to cell death. The process of pore
formation consists of the transformation from a water soluble protein to a membrane bound form. At the cell
membrane, the protein oligomerizes to form a stable transmembrane pore complex which eventually destroys the
cell. The crystal structure of only a few pore forming toxins are known and the mechanism of pore formation is
only partially understood. We have carried out fully atomistic (100 ns) and coarse grained simulations (1
microsecond) of the dodecamer pore complex of the ClyA pore forming toxin. I will illustrate the need to carry
out long time molecular dynamics simulations as well as the issues associated with such simulations.
Python: a swiss-army-knife for HPC
Prabhu Ramachandran
IIT Mumbai
Python is an easy to learn high-level, open source programming language. Itinterfaces well with low-level HPC
languages like C/C++ and FORTRAN. Python'sstrong set of data analysis and visualization libraries coupled
with itsconvenient interactive interpreter allows it to be used as a data analysisenvironment. Unlike Matlab and
similar tools, Python is a general purposeprogramming language with a wide variety of libraries to perform all
kinds oftasks including writing shell-scripts, building Graphical User Interfaces, Webframeworks, etc. This
makes Python a powerful tool in the armory of a HPC-user.Python by itself can be rather slow for pure numerical
looping and it is oftennecessary to write the high-performance parts in a low-level language. Inthis talk, I will
show I have interfaced a fairly large C++ library using SWIG(http://swig.org) to Python and effectively use
Python for my own dataanalysis needs.Over the last few years, we have embarked on creating a Python-based
opensource framework for Smoothed Particle Hydrodynamics (SPH) called
PySPH(http://pysph.googlecode.com). We use Cython (http://cython.org) to obtainhigh-performance in PySPH.
The mpi4py library provides MPI bindings in Pythonand is used for parallel computing. I will also discuss some
of our attemptsat using PyOpenCL to perform both multi-core computation and GPU computationin this
context.Finally, as a taste of what is to come, we will take a look at PyPy(http://pypy.org). PyPy is an alternative
implementation of Python andfeatures a just-in-time compiler. On some of our recent benchmarks itperforms
fairly competitively with C++. This is a huge advantage since itmeans that one can write pure Python code and get
close to the performance ofC.Using the above, I hope to show that Python is really a swiss-army-knife forthe HPC
community.
Hybrid Computing: Exploiting All Resources on Your Desktop
P. J. Narayanan
IIIT Hyderabad
Modern computer systems have a few cores on its GPU, severalcores on accelerators like the GPU, and
perhaps others.Parallel and high-performance computing is within the reach ofalmost everyone today, as a
result. The challenge is inengaging all available resources profitably to maximize theperformance of the
system. In this talk, we will look at someattempts at exploiting the cores of the CPU and the cores of theGPU
cooperatively to solve a problem. We will also look at thegeneral issues involved and speculate on where this
may go inthe near future.
Mathematical model for dealing with some HPC Applications.
P. K. Baruah
Sri Sathya Sai Institute of Higher Learning
Computational capability of GPUs is harnessed to expedite the growth of scientific computing areas, for example,
in the area of image processing, GPUs are being used for Image registration and Image Impainting. Scientific
computation involves the operations on huge amounts of floating point data. The application areas such asclimate
modeling, earth science ...... etc, deal with analyzing huge data. Here the rate at which data is read from the storage
devices (IO bandwidth) could be a significant bottleneck. We propose to address this issue using fast floating
point compression algorithms. We are also looking do some mathematical model for prefetching the data using
some machine learning techniques, reduced order modelling etc which can deal with IO bottleneck. We are also
exploring innovations and creative strategy designs to enhance the performance of existing "outsourcing"
methods. We describe recent work on Data security and Privacy on outsourced data to substantiate this point.
A review of power and energy consumption optimization in HPC
Rishi Pathak, Sandeep Joshi
CDAC, Pune
Power and energy consumption have become important challenges in the field of HPC. In case of large HPC
cluster installations the energy costs over the full period of operation are as high as the cost of building the cluster.
In case of petascale and exascale computing the energy costs would also be economically non-viable. This has led
to considerable efforts being put in the direction of greening of HPC. The efforts are directed towards new
architectures, power efficient processors, voltage and frequency scaling capabilities and new system softwares
which can leverage these hardware capabilities. In this talk we present a short review of the work carried out in
this field over the recent past by many researchers.
HPC @ Intel
Rama Malladi
Intel India
In this talk, we would discuss and highlight some of the architectural innovations in Intel processors and hardware
solutions with specific focus towards HPC applications and user community. Some of the details include
increased vector length of the registers (SSE 128-bit, AVX 256-bit and MIC 512-bit), increased processor core
counts and new instructions/ capabilities. Also, we would go into the details of software tools that Intel provides
for HPC users. These include compilers, libraries and performance analysis tools for single/ multi-node clusters.
100 Gigabit Ethernet and Software Defined Networking: Fast and faster
Ananda Rajagopal
Brocade
100 Gigabit Ethernet was standardized as part of IEEE 802.3ba in 2010. Commercial products have now matured
leading to applications in high-performance research networks. With the size of flows increasingly growing in
research networks, the time is now ripe to consider speeds beyond 10 Gigabit Ethernet. This talk will explore how
100 Gigabit Ethernet be used to accelerate innovation in research networks.An exciting area of networking
innovation that is attracting interest from both the academic community and industry is the area of software
defined networking (SDN), where technologies like OpenFlow promise to bring modularity and dramatically
accelerate the pace at which new protocols can be validated and rolled out to a network.Together, the combination
of 100GbE and SDN can significantly increase the efficiency of HPC networks.
Accelerating High Performance Computing using GPUs
Manish Bali
Nvidia, Pune
Over the last decade, commodity graphics processors (GPUs) have evolved from fixed-function graphics units
into powerful, programmable data-parallel processors. Today's GPU is capable of sustaining computation rates
substantially greater than today's modern CPUs, with technology trends indicating a widening gap in the future.
Researchers in the rapidly evolving field of GPU computing have demonstrated mappings to these processors for
a wide range of computationally intensive tasks, and new programming environments offer the promise of a
wider role for GPU computing in the coming years. In this talk I will begin by discussing the background for GPU
computing and describe some of the recent advances & adoption in this field. The field of GPU computing has
substantially changed over its short lifetime due to new applications, techniques, programming models, and
hardware. As parallel computing has decidedly moved into the mainstream, the lessons of GPU computing are
applicable to both today's systems and HPC applications as well as to the designers of tomorrow's systems and
HPC applications as we look at Petascale and Exascale computing challenges.
Case-studies in Optimizing HPC applications
Yogish Sabharwal
IBM India
We consider two case-studied in optimization of HPC applications:(i) Weather modeling with nested domains;
and (ii) Analytics on graphsIn the first case-study, we consider weather models with high spatial and temporal
resolutions that are required for accurate prediction of meso-micro scale weather phenomena. Using these models
for operational purposes requires forecasts with sufficient lead time, which in turn calls for large computational
power. There exists a lot of prior studies on the performance of weather models on single domain simulations with
a uniform horizontal resolution. However, there has not been much work on high resolution nested domains that
are essential for high-fidelity weather forecasts. We identify and discuss opportunities for improving the
scalability of such configurations.In the second case-study, we consider algorithms for data analytics on graph
data structures. Data intensive supercomputer applications are increasingly becoming more important for HPC
workloads. We discuss data layout strategies that are more suitable for scalability of such applications on HPC
systems. We discuss how these data layouts are useful in reducing the communication and increasing the
performance of these applications on modern day supercomputers.
HPC: A Useful Tool to Study and Understand Monsoons
Ravi Nanjundiah
IISc Bangalore
Monsoon is a coupled multi-component non-linear dynamical system. Incorporation of diverse effects such as the
coupling between ocean and atmosphere, the feedbacks between land and atmosphere and interactions between
clouds and radiation are some of the challenges faced by climate modellers. Long-distance teleconnections such
as the El-Nino-Monsoon teleconnection make the problem more complex. For forecasts on the scale of a season,
initial conditions both for ocean and atmosphere play a critical role. The resolution of models used also has an
impact on the quality of forecasts/simulations. Computing the complex interactions in a realistic fashion in a
reasonabletime-frame is a major challenge. Processes such as radiation, cloud formation and boundary-layer
interactions which are computationally intensive need to be included in a realistic fashion. For predictions we also
need assimilate data from diverse sources such as satellites,ships, weather balloons, etc to create good initial
conditions and these are also very computationally intensive. High Performance computing has always been used
for this purpose. Scalability of climate models remains a major issue. Efforts at our centre and elsewhere to
address the scientific and computational issues will be discussed.
Parallel HiFUN Experience
N. Balakrishnan
IISc Bangalore
Historically, the Aerospace CFD has been the prime mover in thedevelopments in HPC. The talk will introduce a
typical CFD process andhigh light the development of the parallel CFD solver, HIgh resolutionFlow solver on
UNstrauctured meshes (HiFUN). The parallel experiencewith HiFUN, from a modest IISc-PACE128 (128 nodes
of P3) to 125 Terra floprated CRL-EKA with 14,000 Xeon cores will be brought out. The issuesbehind achieving
good parallel scalability without compromising on theAlgorithmic scalability and how HPC can directly impact
the aerodynamicdesign process, will also be discussed. A more recent study on thescalability of HiFUN, carried
out in collaboration with intel will bepresented. The talk will conclude by summarizing the expectations of an
average CFD code developer from the HPC community.
Molecular Simulations for Engineering Applications
Jayant Singh
IIT Kanpur
Molecular simulation of complex fluids, metals and biological systems combined with contemporary
experiments has potentials to be an extremely powerful tool in the development of novel engineering
applications. For example, bottom up approach is now been considered a promising mean for the development of
new materials. From electronic structure to coarse grain simulations various methods now exist and play an
important role in bridging the gap between different length and time scales. However, have these methods
become the mainstream tool for engineers in the way that process simulations or finite element modeling are?In
this talk, a short history of molecular simulation for engineering applications will be presented. Subsequently,
various applications will be considered where molecular simulation has played/or playing an important role.
Some aspect of its usefulness, practicality and accessibility will be discussed. Specific applications of molecular
simulation in this talk will be covered in details namely on adsorption, wettability, nano tribology, and phase
transition of simple and complex fluids. In the end, I will present some of the current challenges in the area of
molecular simulation and its wider use for on the fly calculations at an industrial scale.
Groove binding mechanism of Ionic liquids: A key factor in long-term
stability of DNA in hydrated Ionic liquids?
Sanjib Senapati
IIT Madras
Nucleic acid sample storage is of paramount importance in biotechnology and forensic sciences. Very recently,
hydrated ionic liquids (ILs) have been identified as ideal media for long-term DNA storage. Hence,
understanding the binding characteristics and molecular mechanism of interactions of ILs with DNA is of both
practical and fundamental interest. We employ molecular dynamics (MD) simulations and spectroscopic
experiments to unravel the key factors that stabilize DNA in hydrated ILs. Both simulation and experimental
results show that DNA maintains the native B-conformation in ILs. Simulation results further suggest that, apart
from the electrostatic association of IL cations to DNA backbone, groove binding of IL cations through
hydrophobic and polar interactions contribute significantly to DNA stability. CD spectral measurements and
fluorescent dye displacement assay confirm the intrusion of IL molecules into the DNA minor groove. Very
interestingly, the IL cations were seen to disrupt the water cage around DNA, particularly the spine of hydration in
the minor groove. We believe this partial dehydration of DNA by IL cations contributes to the long-term DNA
stability, as it prevents the hydrolytic reactions that denature DNA. The detailed understanding of IL-DNA
interactions provided here could guide the future development of novel ILs, specific for nucleic acid solutes.
In silico analysis of structural properties of genomic DNA sequences
and promoter prediction
Manju Bansal
IISc Bangalore
Promoters play an important role in transcription initiation and gene regulation, so it is necessary for these DNA
regions to be identified in the various genome sequences being reported. Sequence motif based computational
methods have not been able to identify these regions with high degree of sensitivity and precision. On the other
hand several experimental and computational studies have shown that promoter sequences possess some special
properties, such as stability, bendability, nucleosome positioning preference and curvature, that are common
across many organisms. The core promoter regions show distinctly different structural properties like lower
stability, lower bendability and slightly higher curvature compared to other neighbouring regions. Moreover, in
yeast, we also observed that TATA-containing promoters are less stable, more flexible and more curved compared
to TATA-less promoters, but both have similar nucleosome positioning preferences. The difference in average
free energy or relative stability of promoter regions has been incorporated into a software ‘PromPredict’ for
detailed analysis of predicted promoters and results for bacterial and fungal genomes will be presented.
Computer simulation studies of Bcl-2 family of proteins: HPC
application in cancer drug discovery
R. Sankararamakrishnan
IIT Kanpur
B Cell lymphoma 2 (Bcl-2) family of proteins plays a significant role in the intrinsic pathway of apoptosis and
regulates the mitochondrial outer-membrane permeabilization. Pro-apoptotic and pro-survival Bcl-2 members
interact with each other that decides the fate of the cell. Many anti-apoptotic Bcl-2 members have been shown to
be highly expressed in several types of cancers. Molecules which antagonize these anti-apoptotic Bcl-2 proteins
are likely to be the candidates for cancer drugs. Hence it is important to understand how the pro- and antiapoptotic Bcl-2 proteins with similar helical fold [1] recognize each otherIt has been shown that the peptides from
the conserved BH3 region of different pro-apoptotic Bcl-2 proteins bind to the hydrophobic groove of antiapoptotic Bcl-2 proteins. However, their affinity differs for a specific pro-survival protein. We have carried out
several molecular dynamics simulation studies to understand the phenomena of the molecular recognition in Bcl2 family [2]. Simulations of anti-apoptotic Bcl-2 members in free form and in complex with several pro-apototic
BH3 peptides have revealed the flexibility of hydrophobic binding grooves and specific interactions that can give
rise to high affinity. A large number of pro-apoptotic BH3 peptides has also been simulated in isolation to find the
relationship between the helix stability and binding affinity [3,4].References[1] D. Lama and R.
Sankararamakrishnan, Biochemistry 49, 2574 (2010)[2] D. Lama and R. Sankararamakrishnan, Proteins 73, 492
(2008)[3] D. Lama and R. Sankararamakrishnan, J. Comp. Aided Mol. Des. 25, 413 (2011)[4] V. K. Modi, D.
Lama and R. Sankararamakrishnan, J. Biomol. Struct. Dyn. (2012) in press.
Density of states calculation of peptides using the Wang-Landau
method: Challenges in High Performance Computing
Priya Singh, Subir Sarkar, Pradipta Bandyopadhyay
JNU, New Delhi
Density of states (DOS) is a fundamental quantity for any molecular system. Any thermodynamic quantity for a
system can be calculated with the knowledge of the DOS. The Wang-Landau (WL) method estimates the DOS
using an iterative procedure, which is based on a random walk in energy space [1]. Although the WL is method is
more than ten years old, its applications to biological systems are only a few, presumably because of high
computational cost. We have investigated the performance and accuracy of the WL method for two small
peptides, met-enkephalin and Trp-cage [2, 3]. We have found that the convergence of the WL method for Trpcage, a peptide with 20 amino acids, takes more than 3 weeks in a 8 processor 2.7 GHz machine. This is more than
100 times than the time taken for the other peptide, met-enkephalin which has 5 amino acids. To apply this method
for larger peptides and proteins, substantial decrease in convergence time is required. Different approaches to
decrease the convergence time of the WL method will be discussed in the talk.References:[1] F. Wang, D. P.
Landau, Phys. Rev. Lett. 86 (10): 2050–2053.[2] Priya Singh, Subir K. Sarkar, Pradipta Bandyopadhyay, Chem.
Phys. Lett. 514, (2011) 357–361.[3] Priya Singh, Pradipta Bandyopadhyay, J. Atom. Mol. Opt. Phys. (under
revision).
Kinetic Monte Carlo simulations of thin film growth
Madhav Ranganathan, Pinku Nath
IIT Kanpur
In this work, we show how Kinetic Monte Carlo (KMC) simulations are used in the study of heteroepitaxial thin
films, such as Germanium (Ge) grown on top of Silicon(Si). KMC allows us to study the interplay between two
processes in crystal growth - material deposition and surface diffusion. In a typical experimental set-up, there is a
flux of Ge atoms on an Si substrate in ultra-high vaccuum conditions. Under these conditions, the film shows
morphologies that are dependent on various paramaters like the growth temperature, flux and the thickness of the
film.The length scales involved in these simulations are of the order of hundreds of nanometers and the
corresponding time-scales of these simulations are of the order of seconds to minutes. For these reasons, KMC is
the method of choice for studying these problems. We demonstrate how KMC can be implemented in these
systems using a simple model, known as the Solid-on-Solid (SOS) model. An atomic model of elasticity is used to
model the strain due to lattice mismatch between the substrate and the film. HPC plays a role in these simulations
in two places.The first is parallelization of the elastic energy calculation, the most time-consuming part of the
program. The other role of HPC is to run multiple trajectories simultaneously. We discuss results demonstrating
the effect of elasticity on the early stages of Ge growth on Si.
Issues in effective large scale simulation of the mechanical properties
of amorphous polymers
Sumit Basu
IIT Kanpur
Design of effective polymers for specific structural functions requires computational tools that can connect the
macromolecular architecture to the mechanical properties in a reliable way. This need has given rise to a large
number of research efforts with the aim of predicting mechanical properties, especially uniaxial stress strain
responses over a large range of strain, through Molecular Dynamics (MD) simulations of amorphous, semicrystalline and cross linked polymeric systems. Such simulations are supposed to provide valuable inputs to
downstream continuum simulations (especially Finite Element models) in the form of constitutive models that
are micromechanically motivated rather than phenomenologically obtained from a limited set of experiments.
However, the wide differences in time and length scales between MD and continuum simulations pose a
significant challenge. In this talk we will try to highlight the aspects of mechanical behavior of polymers that are
sensitive to time and length scales and demonstrate that by designing large scale MD simulations carefully, it is
possible, in spite of computational limitations, to enrich continuum models effectively.
Chemical Reactions in Virtual Laboratory
Nisanth N. Nair
IIT Kanpur
In silico study of complex chemical reactions in condensed matter systems and thereby accessing molecular level
details of reaction mechanisms and associated free energies remains as a great challenge in the field of molecular
simulation. Employing the latest developments of ab initio molecular dynamics techniques and the state-of-theart computer technology, we are able to tackle the time-scale and the length-scale bottlenecks in computer
simulations. These enable us to address some of the challenging problems in chemistry and biology. Some
examples will be presented where large scale simulations have unraveled the detailed mechanisms of enzymatic
reactions, homogeneous catalytic reactions in aqueous solutions and heterogeneous catalytic reactions. Finally, I
will be addressing some of the current challenges in the field of ab initio molecular dynamics.
List of Posters:
1. venkatesh shenoi, "Soft Matter Simulations on HPC platform at C-DAC, Pune" , Scientific Computing
Group, C-DAC, Pune (email: [email protected])
2. Prathyusha K. R., "A parallel algorithm for Lowe- Andersen Thermostat",Prathyusha K. R.
Research Scholar,Complex Fluids and Biological Physics Laboratory,Department of Physics, IIT Madras,
Chennai, INDIA 600036 (email: [email protected])
3. Tushar K. Ghosh, "Activation of Oxygen and CO Oxidation at Rhodium/Alumina Interface",PhD student,
C/O-Dr. Nisanth N. Nair, IIT Kanpur (email: [email protected])
4. Siddharth Paliwal,"Thermodynamic stability and phase transitions in metastable systems",Dr. Pankaj Apte,
Dept. of Chemical Engineering, IIT Kanpur(email: [email protected] )
5. Rajat Srivastava, "Fully Atomistic Molecular Simulation for Understanding Nanoscale Phenomena using
High Performance Computing", Rajat Srivastava and Jayant K. Singh, IIT Kanpur. (email: [email protected],
[email protected])
6. S Dinda, "Acetate-Assisted Sequential Metal-NHC Complexation and 1, 5-cyclooctadiene (COD) Activation
at the Iridium Center: A Computational Study", Shrabani Dinda, Department of Chemistry, Indian Institute of
Technology, Kanpur, Kanpur 208016, Tel: +9151 2259 7558, +9184004 01922. (email:
[email protected])
7. Ambrish,” Turbulent convection using HPC”, Ambrish Pandey.,Roll No. 10109872,Supervisor - Prof.
Mahendra Kumar Verma, Department of Physics, IIT Kanpur. (email: [email protected])
8. Rakesh Yadav, “Scaling of Tarang: A pseudospectral solver for turbulence simulations",IIT Kanpur. (email:
[email protected],[email protected])
9. Sandeep Reddy, Simulation of MHD Turbulence and Dynamo, IIT Kanpur. (email:
[email protected],[email protected]).
10. Ravi TRipathi, “Insights into the Protonation States of Class C-beta Lactamase", PhD Scholar
c/o. Dr. Nishanth Nair, Dept. of Chemistry, IIT Kanpur. (Email: [email protected]).
11. Ashish Bhateja, Applications of high performance computing in studying granular segregation process”, IIT
Kanpur. (email: [email protected], [email protected], [email protected])
12. Pranati Rath (will submit soon) [email protected]
13. Bhumkar, "High Accuracy Computing of Transitional and Turbulent Flows", IIT Kanpur (email:
[email protected]).
14. Vivek, “A Uniform Convergent Third-order Domain-Decomposition Method for Singularly Perturbed PDE's
on Parallel Computers”, IIT Kanpur (email:[email protected], [email protected])
15. V M Krushnarao K, Navrose, Suresh Behara, Sidharth G S, Sanjay Mittal, Finite element simulations of fluid
flows
Event Sponsors :