Podcast
Questions and Answers
What is the primary goal of HPC?
What is the primary goal of HPC?
- To simplify software development processes.
- To make previously infeasible computations possible. (correct)
- To reduce the cost of computing infrastructure.
- To improve the energy efficiency of data centers.
Which of the following is a key component of HPC?
Which of the following is a key component of HPC?
- Low-power processors.
- Single-core processing units.
- Standard desktop operating systems.
- Hardware systems and architectures designed for high performance. (correct)
What is a significant motivating factor for using High Performance Computing (HPC) in weather forecasting?
What is a significant motivating factor for using High Performance Computing (HPC) in weather forecasting?
- To simplify the user interface of weather applications.
- To reduce the amount of data collected by weather sensors.
- To provide forecasts in a timely manner, despite complex calculations. (correct)
- To decrease the complexity of weather models.
What is the relationship between HPC and computational intelligence?
What is the relationship between HPC and computational intelligence?
What is the primary reason for utilizing parallel processing within High Performance Computing (HPC)?
What is the primary reason for utilizing parallel processing within High Performance Computing (HPC)?
Which of the following best describes a 'Grand Challenge' problem in the context of HPC?
Which of the following best describes a 'Grand Challenge' problem in the context of HPC?
Why is High Performance Computing (HPC) frequently used in the automotive industry?
Why is High Performance Computing (HPC) frequently used in the automotive industry?
What is a primary reason why airlines leverage High Performance Computing (HPC) systems?
What is a primary reason why airlines leverage High Performance Computing (HPC) systems?
How can HPC contribute to minimizing the impact of natural disasters?
How can HPC contribute to minimizing the impact of natural disasters?
What is the relationship of total render hours and movie production?
What is the relationship of total render hours and movie production?
Which type of HPC architecture involves machines where each processor has its own dedicated memory?
Which type of HPC architecture involves machines where each processor has its own dedicated memory?
Which of the following is essential for developing and running programs in an HPC environment, according to the content?
Which of the following is essential for developing and running programs in an HPC environment, according to the content?
What programming model is commonly used for parallel programming in HPC environments?
What programming model is commonly used for parallel programming in HPC environments?
In weather forecasting, what calculation approximates the number of grid points when predicting weather for two days over the USA and Canada?
In weather forecasting, what calculation approximates the number of grid points when predicting weather for two days over the USA and Canada?
Considering that for predicting the weather of the world at 20 km altitude for the next 2 days results in $4.8 \times 10^{16}$ calculations. How much time would it take to perform such calculation at $10^{12}$ calculations per second?
Considering that for predicting the weather of the world at 20 km altitude for the next 2 days results in $4.8 \times 10^{16}$ calculations. How much time would it take to perform such calculation at $10^{12}$ calculations per second?
For HPC, what does the diagram detailing Instruction [31-26] and components like RegDst, Branch, and ALUOP represent?
For HPC, what does the diagram detailing Instruction [31-26] and components like RegDst, Branch, and ALUOP represent?
In the context of protein folding simulation with HPC, what does '1 petaflop/s machine for 1 year' indicate?
In the context of protein folding simulation with HPC, what does '1 petaflop/s machine for 1 year' indicate?
According to the material, what is the average over-warning cost associated with evacuations due to natural disasters?
According to the material, what is the average over-warning cost associated with evacuations due to natural disasters?
Which of the following is a primary application of HPC in stem cell research?
Which of the following is a primary application of HPC in stem cell research?
How does the use of HPC in nuclear energy research primarily contribute to safety and efficiency?
How does the use of HPC in nuclear energy research primarily contribute to safety and efficiency?
Besides faster processing, what other aspect makes HPC beneficial?
Besides faster processing, what other aspect makes HPC beneficial?
What key metric measures the performance of supercomputers?
What key metric measures the performance of supercomputers?
Where was the K Computer installed during its operation?
Where was the K Computer installed during its operation?
As of the presented data, which supercomputer is recognized as the first to break the exascale barrier?
As of the presented data, which supercomputer is recognized as the first to break the exascale barrier?
What component is crucial for the Frontier supercomputer to move water through the system per minute.
What component is crucial for the Frontier supercomputer to move water through the system per minute.
Approximately how much data can the Orion storage system Frontier hold.
Approximately how much data can the Orion storage system Frontier hold.
What is the name of the supercomputer being developed with IBM's help for ASTRON's Square Kilometer Array project?
What is the name of the supercomputer being developed with IBM's help for ASTRON's Square Kilometer Array project?
When was the executive order creating a National Strategic Computing Initiative signed?
When was the executive order creating a National Strategic Computing Initiative signed?
According to the material provided what are the HPC Architectures?
According to the material provided what are the HPC Architectures?
Which of the following is a necessary component for this Cloud and High Performance Computing course?
Which of the following is a necessary component for this Cloud and High Performance Computing course?
Which of the following programming tools is explicitly mentioned as needed for the course?
Which of the following programming tools is explicitly mentioned as needed for the course?
What is NOT a facility that HiPer-FC provides access to?
What is NOT a facility that HiPer-FC provides access to?
What metric is being measure to demonstrate that computation across multiple cores is faster than a single core in the Matrix Multiplication test on HiPer-FC?
What metric is being measure to demonstrate that computation across multiple cores is faster than a single core in the Matrix Multiplication test on HiPer-FC?
What type of nodes are located in the AIC – HPC (Imhotep)
What type of nodes are located in the AIC – HPC (Imhotep)
From the AIC – HPC (Imhotep) which rack types are related to computation.
From the AIC – HPC (Imhotep) which rack types are related to computation.
Which of the following problems is best suited for HPC, concerning limitations?
Which of the following problems is best suited for HPC, concerning limitations?
A research group wants to use supercomputers to analyse the interactions of millions of molecules to find a new drug. Which limitation does High Performance Computing (HPC) help to overcome in this scenario?
A research group wants to use supercomputers to analyse the interactions of millions of molecules to find a new drug. Which limitation does High Performance Computing (HPC) help to overcome in this scenario?
What is the name of the first computer to reach exaflop speeds?
What is the name of the first computer to reach exaflop speeds?
According to the course material, how can High Performance Computing (HPC) change scientific methodology?
According to the course material, how can High Performance Computing (HPC) change scientific methodology?
Flashcards
What is HPC?
What is HPC?
High-Performance Computing involves powerful hardware, specialized software, programming languages, and parallel paradigms to solve complex problems.
Why use HPC?
Why use HPC?
HPC is needed to solve problems that are computationally intensive, too expensive, too slow, or too dangerous for regular computers.
Grand Challenge problem
Grand Challenge problem
A problem unsolvable in a reasonable time frame with regular computers is called a 'Grand Challenge' problem.
HPC in disaster management
HPC in disaster management
Signup and view all the flashcards
HPC architectures
HPC architectures
Signup and view all the flashcards
Essential HPC tools
Essential HPC tools
Signup and view all the flashcards
HPC in the airline industry
HPC in the airline industry
Signup and view all the flashcards
Study Notes
Learning outcomes
- The goal is to understand HPC, its application and its impact
- The goal is to appreciate the need for HPC through examples
- HPC's difference and affinity to parallel computing should be understood
- A basic knowledge of HPC architectures is the goal
- Learn what you need, to learn HPC
Outline
- Introduction
- What is HPC
- Why is HPC important
- HPC's impact on Science and Engineering
- How to use HPC
- Build local HPC systems
- Course requirements
Intro
- The Cray Jaguar supercomputer could perform more than a million billion operations per second
- The Cray Jaguar took up more than 5,000 square feet
- The Cray Jaguar resided at Oak Ridge National Laboratory in the United States
- In 2009 the Cray Jaguar became the fastest computer in the world
- The K Computer was installed at the RIKEN Advanced Institute for Computational Science, in Kobe, Japan
- The K Computer contains 88,128 processors
- The K computer was shut down in August 2019 after 7 years of operation
- The last rack of Supercomputer "Fugaku" was delivered in May 2020
- Software adjustment is ongoing for general use scheduled in FY2021
- Fugaku tech FX1000 & FX700 are GA already, and A64FX CPU is supplied to HPE
- Frontier - HPE Cray EX235a, AMD Optimized 3rd Generation EPYC 64C 2GHz, AMD Instinct M1250X, Slingshot-11, HPE
- Frontier has Cores: 8,730,112
- Frontier has Rmax (PFlop/s) of 1,102.00
- Frontier has Rpeak (PFlop/s) of 1,685.65
- Frontier has Power (kW) of 21,100
- The El Capitan - HPE Cray EX255a, AMD 4th Gen EPYC 24C 1.8GHz, AMD Instinct M1300A, Slingshot-11, TOSS, HPE
- El Capitan is located at DOE/NNSA/LLNL in the United States
- El Capitan has Cores of 11,039,616
- El Capitan has Rmax (PFlop/s) of 1,742.00
- El Capitan has Rpeak (PFlop/s) of 2,746.38
- El Capitan has Power (kW) of 29,581
Top supercomputers (ranked)
- El Capitan - HPE Cray EX255a, with 11,039,616 cores, Rmax of 1,742.00 Pflop/s, Rpeak of 2,746.38 Pflop/s and 29,581 kW power
- Frontier - HPE Cray EX235a, with 9,066,176 cores, Rmax of 1,353.00 Pflop/s, Rpeak of 2,055.72 Pflop/s and 24,607 kW power
- Aurora - HPE Cray EX - Intel Exascale Compute Blade, with 9,264,128 cores, Rmax of 1,012.00 Pflop/s, Rpeak of 1,980.01 Pflop/s and 38,698 kW power
- Eagle - Microsoft NDv5, with 2,073,600 cores, Rmax of 561.20 Pflop/s, and Rpeak of 846.84 Pflop/s
- HPC6 - HPE Cray EX235a, with 3,143,520 cores, Rmax of 477.90 Pflop/s, Rpeak of 606.97 Pflop/s and 8,461 kW power
- Supercomputer Fugaku - Supercomputer Fugaku, with 7,630,848 cores, Rmax of 442.01 Pflop/s, Rpeak of 537.21 Pflop/s and 29,899 kW power
- Alps - HPE Cray EX254n, with 2,121,600 cores, Rmax of 434.90 Pflop/s, Rpeak of 574.84 Pflop/s and 7,124 kW power
- LUMI - HPE Cray EX235a, with 2,752,704 cores, Rmax of 379.70 Pflop/s, Rpeak of 531.51 Pflop/s and 7,107 kW power
- Leonardo - BullSequana XH2000, with 1,824,768 cores, Rmax of 241.20 Pflop/s, Rpeak of 306.31 Pflop/s and 7,494 kW power
- Tuolumne - HPE Cray EX255a, with 1,161,216 cores, Rmax of 208.10 Pflop/s, Rpeak of 288.88 Pflop/s and 3,387 kW power
What is HPC?
- High Performance Computing (HPC) is a collection of POWERFUL hardware systems and architectures, software tools, programming languages, and parallel programming paradigms,
- HPC provides Computational Intelligence
- HPC makes unfeasible computations possible
- HPC's purpose is to apply to both How Powerful and Why Parallel computing
Why HPC
- HPC allows us to predict the weather of USA and Canada
- Weather prediction is done for the next 2 days at 20 km Altitude
- Weather prediction requires 20 million square km
- Weather prediction requires 2.0×107 km² × 20km×10³ cubes per km³ = 4×10¹¹ grid points
- Weather prediction requires 4×10¹³ calculations × 48hours ≈ 2×10¹⁵ calculations
- Using 10¹² calculations per second: 2×10¹⁵ calculations / 10¹² calculations per second = ~ half hour
- World weather prediction requires 50 million square km
- World weather prediction requires 5.0×10⁸ km² × 20km×10³ cubes per km³ = 1×10¹³ grid points
- World weather prediction requires 1×10¹⁵ calculations × 48hours ≈ 4.8×10¹⁶ calculations
- Using 10¹² calculations per second World weather prediction equals 4.8×10¹⁶ calculations / 10¹² calculations per second = ~ 13 hours
Why Parallel
- Because a standard for statement requires many trips
- for (i = 0; i < TRILLION; i++)
- z[i] = x[i] + y[i];
- This is at least 3×10¹² trips
- Assuming trips at the speed of light calculation takes literally zero time
- r = (3×10⁸ meter / sec×1 sec) / (3×10¹² meters) = 10⁻⁴ meters
- 2×10⁻⁴ / √(3×10⁶) ≈ x×10⁻¹⁰ meters
Science
- Science types include Theoretical, Data Driven, Experimental and Computational
Grand Challenges for HPC
- Examples are problems that cannot be solved in a reasonable amount of time with today's computers
- Applied Fluid Dynamics is a grand challenge
- Meso- to Macro-Scale Environmental Modeling is a grand challenge
- Ecosystem Simulations is a grand challenge
- Biomedical Imaging and Biomechanics is a grand challenge
- Molecular Biology is a grand challenge
- Molecular Design and Process Optimization is a grand challenge
- Fundamental Computational Sciences is a grand challenge
- Nuclear power and weapons simulations is a grand challenge
Computational Science and Engineering Problems
- Too Hard: building large wind tunnels
- Too Expensive: Throw away a passenger Jet
- Too Slow: Climate or galactic evolutions
- Too Dangerous: Nuclear Energy and stem cell research
- Airlines see system-wide logistics optimizations evaluated on HPC systems
- Airlines save approx. US$100 million per airline per year
- The Automotive Industry saves over US$ 1billion per company per year in crash testing, structural integrity, and aerodynamics, with typical use of 500+ cpus
- HPC usage can Minimize Disasters’ Impact
- Cost of evacuation: $1M per Mile
- Typical over-warning factor is 3
- Average over-warning is 200 miles = $200M
- Protein Folding Simulation of 300 amino acid protein in water, 32000 atoms requires 1 millisecond folding time
- Protein Folding Simulation requires 1 petaflop/s machine for 1 year
- Single GPU (RTX 3090/A100) runs this simulation with ~50 ns/day to meet the 55 years required
- 8-GPU HPC Node runs this simulation with ~500 ns/day to get ~5.5 years
- Anton Supercomputer runs this simulation with ~100 µs/day to get ~10 days
HPC Architectures
- Distributed-memory machines
- Shared-memory machines
- GPUs
- FPGAs
- Hybrid machines
HiPer-FC
- High Performance Computing Lab of the Faculty of Computers and Information – Cairo University
- Offers Department of Computer Science
- Two clusters: 12 Nodes Single Core and 14 nodes Multicore
- Total 26 Nodes Running 50 Cores
- Access to CyTera
- Access to BlueGene
- One of the tests run is Matrix Multiplication
AIC – HPC (Imhotep)
- Includes MGMT Node PE R640 Bright Cluster and Mellanox EDR Switches 2 x S4148F 1 x S3048
- Has Login Node PE R640
- Utilizes Intel Studio Programming Env Libraries and Schedular
- Includes GPU Nodes 12 x C4140 and FAT Nodes 20 x PE R640
- Provides Parallel FS BeeGFS 600T and Isilon 300TB
What we need for the course
- Any modern computer (Multicore of course)
- Linux
- GNU Compiler (C or Fortran)
- MPI
- OpenMP
- Cuda/OpenCL
Studying That Suits You
Use AI to generate personalized quizzes and flashcards to suit your learning preferences.
