Educating and Training the HPC Community at KAU. Iyad... · Iyad Katib Educating the HPC Community...
Transcript of Educating and Training the HPC Community at KAU. Iyad... · Iyad Katib Educating the HPC Community...
Educating and Training the
HPC Community at KAU
Dr. Iyad Katib
King Abdulaziz University
@ Saudi HPC Conference
KAUST, 14 March 2017
Outline
• HPC Centre Vision and Mission
• Infrastructure
• Services
• Trainings
Iyad Katib Educating the HPC Community 2
Our Drivers
The level of expertise and infrastructure in
HPC will determine the competitiveness of a
nation and its global standing/leadership
Iyad Katib Educating the HPC Community 3
Mission Statement…
• Vision– To acquire for the university a world-class capacity in
high performance computing to create next generation expertise and workforce for the application of HPC in nurturing a knowledge based (smart) society and economy.
• Mission– To develop both soft and hard infrastructure
(knowledge, science, technologies and infrastructure) for research and development of high performance computing in support for, and collaboration with, other stakeholders, including university divisions, national and international organisations.
Iyad Katib 4Educating the HPC Community
Aziz
• The supercomputer at the King AbdulAziz University (KAU)
• Aziz supercomputer is Fujitsu made and is able to deliver peak performance of ~230 teraflops
• It has a total of 11,904 cores in 496 nodes– each node comprises dual socket Intel Xeon E5-2695v2 12-core processor running at
2.4GHz
– 380 of these nodes contain 96 GB memory each
– the rest of the 112 nodes contain 256 GB each
– making up a total of 66 TB memory in the system.
• The system also contains 2 NVidia Tesla K20 GPU equipped compute nodes with 48 cores and 2 Intel Phi 5110P co-processor equipped compute nodes with 48 cores
• Aziz was ranked number 360 in the June 2015 Top500 competition, currently it is at number 491 (November 2015)
Iyad Katib Educating the HPC Community 5
Aziz: Compute Nodes
– CPU: Intel Xeon, 12000 Core
– NVIDIA GPGPU : 5000 CUDA cores
– Xeon-Phi: 120 cores (MIC)
Iyad Katib 6Educating the HPC Community
7
380 Nodes
9120 Cores
96 GB
256 GB
96 GB
96 GB
Iyad Katib Educating the HPC Community
8Iyad Katib Educating the HPC Community
Services
• Access to Infrastructure (HPC and Big Data Resources)– The Aziz supercomputer
– Spark and Hadoop on clusters and Aziz
– SAP HANA Platform
– Continuing to build additional capacity in HPC and big data
• Expertise– Parallelisation of sequential codes
– Optimisation of parallel codes
– Debugging and configuration of high performance computing codes
– Same as above for big data
– Research support in big data and HPC
• Education– Bachelors, Masters and PhD courses and projects
• Training– Awareness Sessions
– Introductory, intermediate, advanced and on-demand training for students, faculty and external clients and stakeholders
Iyad Katib Educating the HPC Community 9
HPC Roadmap
Y1 Y2 Y3
Establishing the HPC Facility
Establishing Operational Model
Building Capacity
Establishing HPC CoE
Iyad Katib 10Educating the HPC Community
Aziz Users
Iyad Katib Educating the HPC Community 11
College/Center Research Topics
1Center of Excellence for Climate Change
Research
• KAU Climate Model
• Running Applications such as COSMO, WRF,
RegCM
2Faculty of Computing and Information
Technology
• Smart Cities (Healthcare, Transport,
Logistics, Networks)
• Sparse Linear Algebra
• Big data and HPC Integration
• Spectrum Assignment Problem
• Motif Finding Problem
• 3D Object Categorization
• DNA Sequence Alignment
• Real-time Face Detection and Recognition
• Image Processing Tool (for students)
• Password Cracking
• Parallel File Indexing
• DNA Protein Folding Structure (Brute Force)
Aziz Users
Iyad Katib Educating the HPC Community 12
College/Center Research Topics
3Faculty of
Engineering
• Turbulent flow around wings near stall conditions.
• High angles of attack separated flow around slender bodies.
• Shock-Wave/Turbulence-Boundary-Layer interaction.
• Running Applications such as OpenFoam, HiFiles Solver, Anaconda,
GSL, and FFTW.
4Faculty of
Science
• Developing and improving quantum chemical methods to facilitate
the reliable prediction of molecular properties for a broad spectrum
of different chemical and physical situations.
• The modeling of all states of matter.
• Merging methods from molecular quantum chemistry and solid state
physics, facilitating computational research with regard to materials
science and nanotechnology problems
• Offering materials science, nanotechnology, surface chemistry, and
catalysis courses at the graduate or undergraduate level.
• Running Applications such as: Gaussian09, VASP, MolCAS, and
GroMacs.
• Analysis of MRI Images using specific applications such as FSL and
FreeSurfer.
Aziz Users
Iyad Katib Educating the HPC Community 13
College/Center Research Topics
5
Center of Excellence
in Genomic Medicine
Research
• Many of biological problems like:
• genome assembly,
• protein structure prediction,
• molecular dynamics,
• simulation of biological networks.
• Analyzing next generation sequencing data using AZIZ platform.
• Running Applications such as: GATK, SamTools, VcfTools, and Picard.
6Faculty of Maritime
Studies
• Using the Hydrodynamics Model (COHERENCE) to simulate the
hydrodynamics/circulation and the sediment transport along Jizan
Coastline.
7
Center of Excellence
in Environmental
Studies
• The BBVN laser dye to understand its optical signature and maximize its
laser performance.
• Compute the vibrational resolved absorption and emission spectra of the
dye.
Number of Jobs Submitted: 2015
Iyad Katib 14Educating the HPC Community
Number of Jobs Submitted: 2016-17
Iyad Katib 15Educating the HPC Community
Cluster Utilization
Iyad Katib 16Educating the HPC Community
Cluster Utilization: 2015
Iyad Katib 17Educating the HPC Community
Cluster Utilization: 2016
Iyad Katib 18Educating the HPC Community
Cluster Utilization - 2017
Iyad Katib 19Educating the HPC Community
Education
• Bachelors, Masters and PhD Courses– Introduction to HPC
– Performance analysis
– Dependence analysis
– Parallel programming models
– OpenMP
– MPI
– OpenMP
– CUDA
– MIC
– Hybrid CPU/GPU/MIC
– General purpose optimization
Iyad Katib Educating the HPC Community 21
Education…
• Programming and Systems Models /
Architecture
– Distributed Systems
– HPC
• Message Passing Interface
– Big Data
• Bulk synchronous Parallel (BSP)
– Emerging Integration of Big Data and HPC
Iyad Katib Educating the HPC Community 22
Education
• Scientific Computing– Numerical Algebra
– Classic Algorithms• Sorting, Addition etc
• Emerging Applications– Social networks Analysis
– Graph algorithms
– Smart City Applications• Operations Research
• Mobile Computing
• Logistics / Supply Chain Management
– Healthcare / Transportation Analytics
Iyad Katib Educating the HPC Community 23
Training
• Awareness Sessions
• Introductory, intermediate, advanced and
on-demand training for
– students, faculty and external clients and
stakeholders
Iyad Katib Educating the HPC Community 24
Awareness Sessions, Introductory /
Advanced Lectures• We provide basic and advanced lectures on high performance
computing and big data to all faculties.
• The basic lectures are designed for those who are beginners or have zero knowledge of HPC and big data.
• Advanced lectures are delivered based on the specific requirements of the faculties. – E.g. suited to physics or genomics audience
• These lectures and brief workshops provide faculty staff opportunities to discuss and explore new ways for accelerating their research by using data and computing intensive simulations and studies.
• Trainings are tailored for both staff and students.
Iyad Katib Educating the HPC Community 25
Trainings
• Cuda training
– Cuda Architecture
– Cuda in C/C++
– Based on Nvidia material
• MIC training
– MIC architecture and programming
– Emerging developments, Knightslanding etc
Iyad Katib Educating the HPC Community 26
Trainings
• HPC Introduction
• Hardware platform– HPC Cluster: Processors, Memory, Inter-connect
• Programming Paradigms– Vectorization
– Multi-Threading
– Message Passing
– Data Parallel
– Hybrid Parallelization
• How to build a parallel application– Things to take in account
– The inhibitors to parallelism
– Parallel efficiency
• Make and run serial or parallel application– Fortran, C and C++ Compilers usage
– Basic and advanced compilation flags
– Programming examples
Iyad Katib Educating the HPC Community 27
Big Data Analytics (KAU-SAP)
• SAP HANA In-Memory platform that could be used for high speed big data analytics and visualization
• Introduction to technologies including Cloud Solutions, Mobile Applications, EdTech Innovation, Blockchain, IoT, Machine Learning, Deep Learning, Artificial Intelligence, Industry 4.0, and Robotics
• opportunity for KAU faculty and student researchers to discuss their big data requirements and solutions
• Aim at the HPC applications community to explore the possibilities to explore new ways for accelerating data intensive tasks
Iyad Katib Educating the HPC Community 28
Debugging and Profiling (Allinea
/ ARM)• DDT
– Introduction to debugging
– Prepare your application for debugging
– Get started debugging jobs with Allinea DDT
– Discover Allinea DDT interface and its features to debug your software problems
• Allinea / ARM MAP
– Introduction to profiling
– Get started with Allinea MAP in a cluster environment
– Improve memory access performance issues
– Fix MPI load imbalance
Iyad Katib Educating the HPC Community 29
Conclusions and Future
• The Aziz supercomputer, housed at the King Abdulaziz University (KAU), has been operational since June 2015.
• #361 in the June 2015 Top500 supercomputing benchmark.
• It is being used to support a wide range of research areas at the University and other collaborating institutions
– climate change research, meteorology, computer science, engineering, aeronautics, chemistry, physics, nanotechnology, materials science, biosciences, genomics, smart cities, web science, maritime and environmental studies.
• The sheer number of research areas, as well as the significance and diversity of these areas, demonstrates the impact that Aziz is making, and will continue to make, on the scientific, economic and social standing of KAU and Saudi Arabia.
• We have a nascent HPC community at KAU except a few HPC application areas where the ongoing research is considered world leading.
• There is a need to cultivate the KAU community to prepare the next generation HPC and big data scientists and engineers.
• I have discussed the various initiatives and efforts that the HPC Center has been involved in for education and training of the HPC community at KAU.
• We are growing at a fast pace!
Iyad Katib Educating the HPC Community 30