wayground logo

Free Printable Worksheets

Font size

S
M
L
XL
Worksheets

High Performance Computing

Total questions: 120

Worksheet time: 30hrs 0mins

Name
Class
Date
1.

  1. What are the four reasons for performance degradation according to the acronym SLOW?

a)

  1. Slow Access, Latency, Overconsumption and Waiting

b)

  1. Starvation, Latency, Overhead and Waiting

c)

  1. Slow Access, limited accessibility, Overconsumption and Waiting

d)

  1. Security Threats, Latency, Overconsumption and Waiting

2.

  1. What is true about Sustained performance? (Multiple selection)

a)

  1. Maximum rate at which operations can be accomplished theoretically by the hardware resources of a supercomputer

b)

  1. Actual or real performance achieved by a supercomputer system in the performance of an application program

c)

  1. Sustained performance is considered a better indicator of the true value of a supercomputer than its specified peak performance

d)

  1. Peak performance is considered a better indicator of the true value of a supercomputer than its specified sustained performance

e)

  1. While sustained performance cannot exceed peak performance, it can be much less and often is

3.

  1. What is true about strong scalability

a)

  1. Both the number of processors and the problem size are increased

b)

  1. Number of processors is increased while the problem size remains constant

c)

  1. Results in a reduced workload per processor

d)

Results in a constant workload per processor

e)

  1. Mostly used for long-running CPU-bound

4.

  1. What type of problem representatives are used for machine learning, fraud detection and security services?

a)

  1. Linear Algebra

b)

  1. Stochastic systems

c)

  1. Graph problems

d)

  1. Solution of partial-differential equations

e)

  1. Large systems with pairwise force interactions

5.

  1. What is the element that incorporates all the functional elements required for computation, and is highly replicated to achieve large scale in a supercomputer?

a)

  1. Core

b)

  1. Chip

c)

  1. Blade

d)

  1. Node

e)

  1. Rack

6.

  1. What type of problem representatives are used for Navier–Stokes equations, Black–Scholes equation and Weather Prediction?

a)

  1. Liner Algebra

b)

  1. Stochastic systems

c)

  1. Graph problems

d)

  1. Solution of partial-differential equations

e)

  1. Large systems with pairwise force interactions

7.

  1. What is true about Commodity Cluster?

a)

  1. is a form of HPC assembled from commercially manufactured subsystems

b)

  1. cluster “node” is a computer that can be directly employed individually as a PC

c)

  1. Provides economy of scale to increase performance to cost dramatically compared to custom-designed MPPs of the same scale

d)

  1. Examples are Touchstone Paragon (1994), the Thinking Machines Corporation CM-5 (1992), and the IBM SP-2

8.

What is true about FLOPS?

a)

  1. Stands for Floating-Point OperationS

b)

  1. is an addition or multiplication of two real (or floating-point) numbers

c)

  1. Stands for Floating-point operations per second

d)

  1. We uses the greek prefixes kilo, mega, giga, tera, and peta to represent 1000, 1 million, 1 billion, 1 trillion, and 1 quadrillion

e)

is an addition or multiplication of two integer numbers represented

9.

  1. Historically supercomputers have been applied to science and engineering,and the methodology has been described as the “third pillar of science”alongside and complementing what other pillars?

a)

  1. Experimentation (empiricism) and Mathematics (theory)

b)

  1. Simulation and Mathematics (theory)

c)

  1. Empiricism and Simulation

d)

Experimentation (empiricism) and Simulation

10.

  1. What is true about “highly parallel Linpack” (HPL)?

a)

  1. Widely used supercomputer benchmark

b)

  1. Solves a set of linear equations in dense matrix form

c)

  1. measures the rate at which the system can transpose a large array

d)

  1. gives a means of comparative evaluation between two independent systems by measuring their respective times to perform the same calculation

11.

  1. Which of the following best describes Moore's law?

a)

  1. A. The prediction that supercomputer performance would double every 5 years.

b)

  1. The hypothesis that device transistor density will double every decade.

c)

  1. The prediction by Intel's Gordon Moore that device transistor density would increase by a factor of two every 2 years.

d)

  1. The law that all supercomputers should use Intel processors.

e)

  1. The guideline for assessing the efficiency of processors based on their energy consumption

12.

  1. When assessing the performance of supercomputers using the Top 500 list, which benchmark is primarily used?

a)

  1. Moore's benchmark 

b)

  1. HPC linear algebra 

c)

System stack performance rate

d)

SIMD array evaluation

e)

  1. HPL or "Linpack" benchmark for dense linear algebra

13.

What is the primary difference between sustained and peak performance in HPC?

a)

  1. Sustained performance refers to the maximum rate theoretically possible by the hardware, while peak performance refers to the real-world application performance.

b)

  1. Sustained and peak performance are the same.

c)

  1. Sustained performance refers to the real-world application performance, while peak performance refers to the maximum rate theoretically possible by the hardware.

d)

  1. Peak performance is only relevant for supercomputers, while sustained performance is for personal computers.

e)

  1. Sustained performance is only achieved after a supercomputer has been operational for 5 years.

14.

  1. Which of the following is NOT a source of performance degradation in HPC systems?

a)

  1. Starvation 

b)

  1. Overclocking 

c)

Latency

d)

Overhead

e)

  1. Contention

15.

  1. With the potential end of Moore's law in sight, which epoch of supercomputing evolution reflects a possible direction for HPC architecture?

a)

  1. Von Neumann architecture in vacuum tubes 

b)

  1. Calculator mechanical technology

c)

  1. SIMD arrays 

d)

  1. Multicore petaflops 

e)

  1. Vector processing

16.

  1. Imagine you work for a company specializing in weather forecasting. The company's predictions are crucial for several industries, like aviation and agriculture. To ensure accurate predictions, you rely on a supercomputer to process vast amounts of data. Which performance metrics would you prioritize for real-time weather prediction?

a)

  1. Peak performance: the maximum rate that can be achieved theoretically by the hardware.

b)

  1. Sustained performance: the actual or real-world performance of the application.

c)

  1. Moore's Law performance rate. 

d)

Performance as rated by the Top 500 list.

e)

  1. SIMD array evaluation speed

17.

  1. A biomedical research institute is analyzing complex protein folding patterns. The process involves dense linear algebra calculations. To compare two supercomputers for potential acquisition, which benchmark would be the most relevant?

a)

  1. Von Neumann benchmark 

b)

  1. SIMD performance metric

c)

  1. HPL or "Linpack" benchmark for dense linear algebra: a measure of a system's floating-point computing power.

d)

  1. Moore's performance rate 

e)

  1. System stack throughput test

18.

XYZ Corp., a leading AI research facility, just deployed a supercomputer. During initial testing, the system frequently suffers from delays and lags. As a troubleshooter, which source of performance degradation would you first investigate?

a)

  1. Starvation: a situation where a task is perpetually deprived of the resources or conditions it needs to proceed.

b)

  1. CPU brand compatibility 

c)

  1. Moore's law violation

d)

  1. The age of the machine 

e)

  1. Vector processing alignment

19.

  1. BioTech Inc. is working on simulating the human brain. This project requires significant memory and processing power. Knowing that the company is going to require enhanced computational abilities soon, what would be a recommendation in light of Moore's law?

a)

Wait for two years before buying any new supercomputer.

b)

Acquire multiple low-powered computers now.

c)

  1. Invest in a supercomputer with multicore petaflops architecture: a system with the ability to perform a large number of floating-point calculations simultaneously.

d)

  1. Focus on SIMD array systems. 

e)

  1. Just upgrade the current software without changing the hardware

20.

  1. The Oceanography Department of a university wants to model ocean currents to predict the spread of pollutants. They are looking at the Top 500 list to identify a suitable supercomputer. Why is this list relevant for their decision?

a)

  1. The list is updated biannually and ranks supercomputers based on a system's floating-point computing power using the HPL or “Linpack” benchmark.

b)

  1. The list showcases the price of the top supercomputers. 

c)

It lists supercomputers based on their size and weight.

d)

The list ranks computers according to Moore's Law.

e)

It shows the power consumption of each supercomputer.

21.

  1. What are the key properties that determine performance of an HPC architecture and used in P=e×S×a(R)×μ(E)?

a)

  1. Access, Latency, Overconsumption and Waiting

b)

  1. Speed of components, parallelism and efficiency

c)

Security Threats, Latency, Overconsumption and Waiting

d)

  1. Access, Latency, Overconsumption and efficiency

22.

  1. What are the main types of Flynn's taxonomy of parallel architectures?

a)

SISD—single instruction stream, single data stream

b)

  1. SIMD—single instruction stream, multiple data stream

c)

  1. MIMD—multiple instruction stream, multiple data stream

d)

MISD—multiple instruction stream, single data stream

e)

  1. SPMD - single program, multiple data stream

23.

Please select the sentences that are true

a)

HPC architecture is concerned with only the lowest-level technologies and circuit design.

b)

  1. The cost of the software on an HPC system is much less than the cost of the hardware platform.

c)

  1. The greater the performance that is required, the harder it is to optimize the user program.

d)

  1. Code reuse is critical to managing application development complexity and difficulty.

24.

  1. Select the correct words:

HPC architecture exploits its enabling technologies to _______ time to solution, ________throughput of operation, and serve the class of computations associated with ________, usually ______ -intensive,applications.


a)

  1. Maximize, Minimize, small , numeric

b)

Maximize, Minimize, large , data

c)

Minimize, Maximize, large, numeric

d)

  1. Maximize, Maximize, large , data

25.

  1. Order the memory technologies by speed from faster to slower:

a)

  1. Magnetic storage, NVRAM, DRAM, SRAM

b)

  1. SRAM, DRAM, NVRAM, Magnetic Storage

c)

  1. DRAM, SRAM, NVRAM, Magnetic storage

d)

  1. Magnetic storage, NVRAM, DRAM, SRAM

26.

What is true about von Neumann architecture?

a)

  1. Define in 1940 the modern architecture for computers

b)

Uses an ALU, registers, processor and main memory

c)

  1. The ALU (arithmetic logic unit) stores memory

d)

  1. A load operation reads a word from memory and places its value into a register, to be used by the ALU.

e)

  1. Operation of the processor is managed by the controller 

27.

  1. What is true about Commodity Cluster?

a)

  1. is a form of HPC assembled from commercially manufactured subsystems

b)

  1. cluster “node” is a computer that can be directly employed individually as a PC

c)

  1. Provides economy of scale to increase performance to cost dramatically compared to custom designed MPPs of the same scale

d)

  1. Examples are Touchstone Paragon (1994), the Thinking Machines Corporation CM-5 (1992), and the IBM SP-2

28.

  1. Suppose 50% of a program can be executed in parallel. What is the maximum speed up for a computer with 5 cores?

a)

  1. 1.82 time

b)

  1. 1.66 times

c)

2.21 times

d)

  1. 0.89 times

29.

  1. What is true about MPPs clusters?

a)

  1. MPPs are single systems comprising many integrated computer processors.

b)

  1. Easily scales to extremes of computing system size and performance

c)

  1. MPPs always have shared-memory by all nodes

d)

  1. They use Message passing by a system area network (SAN)

e)

  1. Typically, MPPs exhibit lower efficiencies with respect to number of cores than Commodity Clusters

30.

  1. What is true about Reliability?

a)

  1. The bigger the system, the most faults will have.

b)

  1. “Hard” faults is when a part of the hardware breaks permanently

c)

  1. Soft” fault a when the Software brakes permanently

d)

  1. “Soft” fault is when a part intermittently fails but otherwise operates correctly.

e)

  1. We can use checkpoint/restart to prevent software errors

31.

  1.  In HPC systems, why is the use of high-speed interconnection networks, like InfiniBand, crucial?

a)

  1. For more storage capacity. 

b)

To ensure low power consumption

c)

  1. To allow RGB lighting synchronization.

d)

  1. To ensure efficient communication between thousands of processors. 

e)

  1. To cool down the system.

32.

Which of Flynn's taxonomy classifications is described as multiple instructions operating on a single data stream and is rare in practice?

a)

  1. SIMD 

b)

SISD

c)

  1. MISD 

d)

  1. MIMD 

33.

  1. In the context of the von Neumann architecture, what is the limitation where the speed of operations is constrained by the rate of data transfer between the CPU and memory?

a)

  1. CPU Throttling

b)

  1. Cache Miss

c)

  1. Memory Leak

d)

  1. von Neumann Bottleneck

e)

  1. Pipeline Stalling

34.

  1. If a modern CPU aims to enhance instruction throughput by dividing instruction processing into distinct stages, with each stage being managed by a different segment of the CPU, what is this technique called?

a)

Vector Processing

b)

Loop Unrolling

c)

Hyperthreading

d)

Branch Prediction

e)

Pipelining

35.

In the context of High-Performance Computing (HPC), what advantage does incorporating accelerators like GPUs or  FPGAs alongside traditional CPUs provide?

a)

  1. Allows better RGB lighting effects. 

b)

Enhances the boot-up speed of the system.

c)

Provides better aesthetics to the system build.

d)

Speeds up specific types of computations.

e)

  1. Reduces the power consumption of the entire system.

36.

  1. Jane is a game developer working on optimizing a graphics engine. She wants to perform the same shading operation on multiple vertices simultaneously. Based on Flynn's Taxonomy, which type of parallelism should Jane's graphics engine utilize?

a)

SISD

b)

  1. SIMD (A paradigm where a single instruction operates on multiple data points concurrently)

c)

MISD

d)

MIMD

37.

  1. Let's say you're using Google Chrome, a popular web browser. When you open a new web page, the browser processes multiple tasks like fetching data, rendering images, and playing videos. To boost efficiency, modern CPUs process several instructions related to these tasks concurrently but in different stages. What is this technique called?

a)

Web Lining

b)

Video Streaming

c)

Data Mining

d)

Pipelining

38.

Apple's M1 chip is known for its impressive performance in both high-end tasks like video editing and everyday tasks like browsing. The chip contains a combination of high-performance and energy-efficient cores. This diverse mixture of cores on a single chip is an example of what computing structure?

a)

Monolithic Computing

b)

Heterogeneous Computing

c)

  1. Homogeneous Computing

d)

Singular Computing

39.

  1. Dr. Thompson is researching the protein structures of a newly discovered virus. He needs to run simulations involving millions of atoms and requires a system that can handle multiple sets of instructions on multiple data points. According to Flynn's Taxonomy, which type of parallelism should Dr. Thompson's HPC system employ for his research?

a)

SISD

b)

SIMD

c)

MISD

d)

MIMD (Multiple instructions operating on multiple data points)

40.

The StarGaze Observatory is attempting to map the entire Milky Way. Their computational needs involve processing vast amounts of data from different telescopes concurrently. They're considering using a Beowulf cluster. Why might a Beowulf cluster be suitable for StarGaze's computational demands?

a)

It focuses solely on SIMD operations.

b)

It is a single powerful machine with a large CPU.

c)

  1. It's a cluster of networked computers acting as one system (often using commodity hardware).

d)

It exclusively uses quantum computing.

e)

It operates on MISD principles.

41.

What is true for commodity clusters? (multiple answers)

a)

Use Standard components used for PCs to take advantage of economies of scale

b)

  1. System software is usually proprietary

c)

System software is usually open-source,nonproprietary, and Linux based

d)

They are a type of Massively Parallel Processor (MPP)

e)

  1. Each machine is largely independent of the others in terms of memory, disk, etc

42.

  1. What are the main components of a commodity cluster?

a)

  1. Node, System Area Network (SAN), Host and Secondary Storage

b)

  1. Node, OS, SAN, Storage, Local area network

c)

Host, Storage Arena Network, Local Area Network, Storage

d)

Host, OS, Storage Arena Network, Local Area Network

43.

What are the main components of a commodity cluster?

a)

  1. Node, System Area Network (SAN), Host and Secondary Storage

b)

Node, OS, SAN, Storage, Local area network

c)

  1. Host, Storage Arena Network, Local Area Network, Storage

d)

Host, OS, Storage Arena Network, Local Area Network

44.

Please select the sentences that are true about Operating Systems:

a)

The Kernel manages hardware resources and arbitrates access to them from other software layers

b)

The System libraries expose a common set of programming interfaces permitting application writers to communicate with the kernel and underlying physical devices

c)

An OS has additional system services performed by the background processes

d)

OS includes various administrator and user utilities that comprise programs invoked by users of the computer to accomplish specific minor tasks

45.

  1. What is true about Storage for Commodity Clusters? (Only one must be chosen.)

a)

  1. A node always has hard disk drives or solid -state devices (SSDs) into each node

b)

Sometimes you can find diskless nodes (less maintenance) or storage replaced by NVRAM

c)

A storage subsystem with its own controllers and possibly its own network may be included as a separate unit within the cluster

d)

  1. Some nodes access external storage system via LAN and share with other systems

e)

It is not possible to have a combination of the three storage options (local drive, storage subsystem, external storage).

46.

What are Unix/Linux based OS?

a)

Windows, Open BSD, Linux, Minix , Solaris

b)

BSD, MacOS, Linux, Minix , HP-UX, Solaris

c)

BSD, MacOS, Linux, OS/400, Solaris

d)

  1. BSD, Novell Netware , Linux, OS/400, VMS

47.

What is true about Parallel programming modalities?

a)

The main three modalities are Throughput computing, Message passing and Shared- memory multiple-thread applications

b)

An example of message passing is OpenMP

c)

An example of Shared-memory multiple- thread applications is OpenMP

d)

An example of message passing is MPI

e)

An example of Shared-memory multiple- thread applications is MPI

48.

  1. What is true about SLURM?

a)

  1. It is a debugger used in Linux parallel systems for message passing in multiple nodes

b)

  1. It is a Simple Linux Utility for Resource Management used in HPC systems

c)

  1. It is the most HPC free open source tool for resource management for Linux

d)

It coordinates accesses to multiple execution units, memory allocation, network selection, and persistent storage allocation

e)

  1. It automates the tasks related to distribution of user workloads across the compute nodes and monitoring of progress of the execution

49.

What is the typical order for the package configuration and building?

a)

  1. rpm to unpack files, configure for configuration, makefiles for create the file needed for the build process and mpirun to run the configuration files

b)

rar to unpack files, configure for configuration, makefiles for create the file needed for the build process and mpirun to run the configuration files

c)

  1. Tar to unpack files, configure for configuration and make for create the makefiles needed for the build process

d)

  1. Tar to unpack files, gnuplot for configuration and make for create the makefiles needed for the build process

50.

What is true about the Linux Filesystem?

a)

“/bin” contains critical system executables that may be used during system boot

b)

  1. “/usr/bin” contains nonessential executables, typically system-wide application binaries.

c)

“/lib” includes libraries for the essential executables in “/bin” and “/sbin”

d)

“/etc” stores host-specific configuration files.

e)

“/opt” contains temporarily mounted file systems.

51.

  1. Which of the following best describes a commodity cluster in HPC?

a)

A group of integrated computer systems created for exclusive use.

b)

  1. A set of supercomputers that work together to perform calculations.

c)

  1. A specialized computing system manufactured for a niche audience.

d)

  1. A group of integrated computer systems made from easily available hardware.

e)

  1. A single powerful computer with specialized software.

52.

The Beowulf Cluster Project is best known for:

a)

Being the first supercomputer.

b)

Its role in quantum computing.

c)

Popularizing the concept of joining off-the-shelf PCs for parallel work.

d)

Its advanced proprietary software.

e)

  1. Being a commercial project for creating profit.

53.

  1. Which of the following components is NOT typically found in a node of a commodity cluster?

a)

Disk drives for nonvolatile storage.

b)

  1. An onboard network connecting components.

c)

A standalone GPU dedicated only for graphics rendering.

d)

One or more processor cores.

e)

  1. Main memory banks.

54.

Parallel library application programming interfaces (APIs) in HPC

a)

  1. Allow for simultaneous computation on multiple data sets.

b)

Are exclusive to the C programming language.

c)

Reduce the computing power of individual nodes.

d)

Are mostly used for designing game engines.

e)

Have bindings only to parallel programming languages.

55.

  1. When using a commodity cluster, what is typically the FIRST step in utilizing its HPC capabilities?

a)

Visualizing the data

b)

Running applications

c)

Debugging the data.

d)

  1. Logging on to the system.

56.

  1. A startup company wants to build a commodity cluster for its data processing needs. They have a limited budget but need efficient parallel computing. They decide to buy off-the-shelf PCs. Which project would provide them with a blueprint for their endeavor?

a)

Quantum SuperChain Project

b)

AlphaNet System Design

c)

Beowulf Cluster Project - A design for creating high-performance clusters using commercial PCs.

d)

Graphene Integration Plan

57.

BioTech Corp. processes large genomics datasets. Their computing node just failed because of an overheated GPU used for graphics rendering, not computation. When replacing this node in their commodity cluster, which component should they reconsider?

a)

  1. Disk drives for nonvolatile storage. 

b)

An onboard network connecting components.

c)

  1. A standalone GPU dedicated only for graphics rendering. 

d)

  1. One or more processor cores.

58.

MeteorTech is a company focused on weather prediction using HPC commodity clusters. They want their programs to process multiple sets of meteorological data simultaneously. Which type of programming interface would best suit their needs?

a)

Serial Processing APIs

b)

Graphic Design Interfaces

c)

  1. Parallel library application programming interfaces - Allows for simultaneous computation on multiple data sets.

d)

  1. Video Encoding APIs 

59.

  1. DataVis Corp. specializes in creating visual representations of large datasets. They use an HPC commodity cluster for processing and need an environment for turning the processed data into visual models. Which tool, known for turning complex data into visual graphics, would be apt for them?

a)

DebugMaster - A tool focused on debugging code.

b)

Paraview - An open-source data analysis and visualization application

c)

  1. CompileRight - A compiler tool for various programming languages. 

d)

  1. SpaceGuard - A tool for optimizing storage.

60.

GameForge, a gaming company, uses HPC commodity clusters to test new games. They recently faced bottlenecks in their testing phase. They want to monitor their applications to see where the applications spend most of their runtime. Which HPC tool, designed to measure the performance of applications, would they use?

a)

DebuggerPrime - A tool to step through and break the code.

b)

PerfSuite - A collection of tools and libraries for performance analysis.

c)

  1. VisMaster - A tool for creating graphical interfaces. 

d)

ConnectAll - A tool for establishing multiple network connections.

61.

  1. What is true for HPC benchmarks? (multiple answers)

a)

  1. Is a way to measure the performance of a supercomputer theoretically

b)

  1. Is a way to measure the performance of a supercomputer empirically

c)

  1. Provides some standardized type of workload that may vary in size or input dataset

d)

  1. Provides some standardized type of workload that is fixed in size and input dataset

62.

What are the two main types of HPC benchmarks?

a)

  1. Synthetic where workloads are designed and created to impose a load on a specific component in the system

b)

  1. TOP 500 where workloads are based on the HPL HPCG benchmark for memory

c)

  1. Application where the workload is derived from a real world application

d)

AITOP500 where workloads are based on the HPL AI benchmark for tensorflow applications

63.

What are the characteristics of a good HPC benchmark?

a)

  1. relevant and meaningful to the target application domain

b)

  1. applicable to a broad spectrum of hardware architectures

c)

  1. adopted by both users and vendors

d)

  1. enable comparative evaluation.

64.

What is true about HPL benchmark?

a)

The Linpack benchmark solves a dense, regular system of linear equations

b)

The Linpack benchmark provides an estimate of a system's effective floating-point performance.

c)

  1. The HPL benchmark is used for ranking supercomputers in the Top 500 list.

d)

  1. HPL is part of the HPC Challenge benchmark suite that contains seven widely used HPC benchmarks

e)

  1. Early versions of the HPL include the floating-point-intensive Whetstone benchmark and the integer-oriented Dhrystone benchmark.

65.

What is true about HPCG benchmark?

a)

  1. Is meant to complement the HPL benchmark

b)

It provides memory and data-access patterns in application workloads that are not well represented by HPL

c)

  1. It centers on a sparse system of linear equations arising from the discretization of a 3D Laplacian partial differential equation

d)

  1. Implements the breadth-first search algorithm and reports TEPS as a key metric.

e)

  1. HPCG performance is similar in value to HPL performance on even the fastest supercomputers.

66.

What is true about Graph500?

a)

  1. Is intended to represent data-intensive workloads.

b)

implements the breadth-first search algorithm and reports TEPS as a key metric

c)

  1. Graph500 benchmark performance has gone flat even while HPL benchmark performance continues to grow exponentially.

d)

  1. is a series of small self-contained programs that encapsulate the performance attributes of a large computational fluid dynamics application

67.

  1. What is true about Mini-applications benchmarks?

a)

  1. Complement HPC benchmarking efforts and better capture real application behavior

b)

Permits exploring the parameter and interaction space of memory, network, accelerators, and processor element, especially in terms of emerging hardware

c)

  1. They are Smaller versions of real applications

d)

They permit testing emerging programming models outside conventional parallel APIs (MPI and OpenMP)

68.

What are problems of NPB benchmark?

a)

  1. large integer sort for testing both integer computation speed and network performance,

b)

  1. embarrassingly parallel random number generation for integral evaluation

c)

  1. a block tridiagonal solver with a 5 × 5 block size AND a pentadiagonal solver AND an LU solver for coupled parabolic/elliptic partial differential equations. 

d)

  1. a conjugate gradient approximation to compute the smallest eigenvalue of a sparse symmetric matrix

e)

  1. a multigrid solver for computing a 3D potential

  2. Fourier transform

69.

  1. What are the key roles of a HPC benchmark?

a)

Help decide size/type of HPC to procure AND Estimate performance user applications

b)

  1. Identify & quantify performance upper bounds and limitations of algorithms

c)

Compare performance between technologies

d)

Assessment efficiency of applications AND Provide an historical record exploring HPC trends

e)

  1. Quantify the percentage of theoretical peak perf.

70.

  1. What are examples of Mantevo Suite benchmark?

a)

  1. MiniAMR— exploring adaptive mesh refinement and dynamic execution with refinement and coarsening of meshes driven by objects passing through the mesh

b)

MiniGhost—exploring halo exchange in context of finite differencing app on uniform 3D mesh

c)

MiniFE—finite element codes

d)

Cloverleaf—solving compressible Euler equations

e)

  1. TeaLeaf—solving linear heat conduction equation.

71.

  1. What is the primary purpose of benchmarking in HPC?

a)

To debug software programs.

b)

To compare different programming languages.

c)

  1. To measure and compare the performance of supercomputers empirically.

d)

o enhance the graphical user interface of software.

e)

To upgrade the operating system

72.

  1. Which of the following is NOT a key property of a good HPC benchmark?

a)

  1. Relevant to the target application domain.

b)

Adopted by a majority of vendors

c)

Enables comparative evaluation.

d)

  1. Optimized for one specific hardware only.

e)

Applicable to various hardware architectures

73.

  1. The NAS Parallel Benchmarks are specifically designed to encapsulate the performance attributes of which domain?

a)

  1. Database management

b)

Machine learning algorithms

c)

Computational fluid dynamics applications

d)

  1. Image processing

74.

  1. Why are mini-applications increasingly being used in HPC benchmarking?

a)

  1. They provide entertainment for the users.

b)

  1. They simulate the entire software development process.

c)

  1. They capture specific performance attributes of real-world applications in a more manageable size.

d)

  1. They are mandatory for all HPC systems.

75.

Why is it essential for an HPC benchmark to be relevant to the target application domain?

a)

  1. To ensure the benchmarking process is time-consuming.

b)

o guarantee that the benchmark results directly influence hardware sales.

c)

  1. To ensure that the benchmark results are reflective of the performance users can expect in their applications.

d)

  1. To ensure easier coding and implementation of benchmarks.

76.

  1. An aerospace engineering company is looking to simulate airflow over a new aircraft design using computational fluid dynamics (CFD). They wish to select the best HPC system for their needs. Which benchmark would be most pertinent for the company to consider for understanding the performance of HPC systems for their specific application?

a)

HPC Challenge Benchmark Suite

b)

  1. Graph500

c)

NAS Parallel Benchmarks (described as: benchmarks encompassing workloads commonly seen in computational fluid dynamics applications)

d)

  1. High Performance Conjugate Gradients

77.

  1. A big data analytics firm is evaluating HPC systems for their graph-based data mining tasks, which involve traversing massive datasets. Given the firm's requirements, which benchmark would be most appropriate for evaluating the suitability of different HPC systems?

a)

  1. Graph500 (described as: a benchmark focused on graph traversal workloads)

b)

Highly Parallel Computing Linpack

c)

  1. HPC Challenge Benchmark Suite

d)

  1. HPC AI500

78.

A national weather forecasting agency is looking to enhance the accuracy of their predictions and is therefore considering HPC systems to run more granular simulations. Which of the following benchmarks would be the least relevant for the agency's use case?

a)

  1. NAS Parallel Benchmarks

b)

  1. HPC AI500 (described as: a benchmark focusing on AI and machine learning workloads in HPC)

c)

  1. HPC Challenge Benchmark Suite

d)

  1. Highly Parallel Computing Linpack

79.

  1. A gaming company is developing a cutting-edge physics engine for their next game and is evaluating HPC systems to accelerate the R&D process. Given that their primary concern is effective floating-point performance, which benchmark might be most informative for their evaluation?

a)

  1. Highly Parallel Computing Linpack (described as: a benchmark that estimates a system's floating-point performance by solving a dense system of linear equations)

b)

  1. Graph500

c)

HPC AI500

d)

  1. NAS Parallel Benchmarks

80.

  1. A research lab is aiming to study genetic data and its patterns of inheritance across generations. They require both traditional HPC workloads and machine learning algorithms. Which benchmark would best help the lab assess an HPC system's capabilities for their specific needs?

a)

  1. NAS Parallel Benchmarks

b)

Graph500

c)

  1. High Performance Conjugate Gradients

d)

  1. HPC AI500 (described as: a benchmark that evaluates performance in artificial intelligence tasks on HPC systems)

81.

  1. What service is responsible for monitoring the state of system resources and deciding when and on which subset of resources each job is to run?

a)

  1. Server

b)

  1. Communication service

c)

Scheduler

d)

MoM

82.

A srun command is submitted on a machine equipped with dual quad-core processors (each core supporting single thread of execution), two nodes will be allocated for the job. Assuming the first socket of node 0 includes cores numbered 0–3 and the second cores 4–7, task 0 will run on cores 0 and 1, task 1 on cores 4 and 5, task 2 on cores 2 and 3, and task 3 on cores 6 and 7. The remaining tasks will be instantiated on node 1, with task 4 using cores 0 and 1 and task 5 cores 4 and 5.

a)

  1. srun–n6–c2–m’block:cyclic’ my_app

b)

srun–n6–c2–mplane=2:fcyclic,NoPack my_app

c)

  1. srun–n6–c2a–m’block:block’ my_app

d)

  1. srun–wnode0[4-6],node08–N6 my_app

83.

  1. What this command will do? (multiple true)

qsub –q debug –l nodes=4,walltime=30:00 job.sh

a)

  1. Submit the job to the specific queue (“debug” in this case)

b)

will start four instances of the job.sh program (one per host)

c)

the maximum amount of time a job is permitted to run is 30 min

d)

will launch a Slurm job in debug partition

84.

What are the resource management tools principal functions? (multiple true)

a)

  1. resource allocation

b)

workload scheduling

c)

support for distributed workload execution and monitoring

d)

Replaces the OS in the nodes

85.

What is true about job queues in resource managers?

a)

  1. Is possible to find interactive queue solely for interactive jobs.

b)

  1. defines the order in which jobs are selected

c)

Most systems typically use multiple job queues

d)

  1. FIFO is a job queue strategy

86.

  1. What common parameters that affect job scheduling include the following? (multiple true)

a)

Requested execution time estimated by the user for the job.

b)

  1. Availability of execution and auxiliary resources

  2. Priority

c)

  1. Maximum number of jobs that a user is permitted to execute simultaneously.

d)

Event occurrence, when job start is postponed until specific predefined event occurs

e)

  1. Software license availability if a job is requesting the launch of proprietary code

87.

What is true about Preemption? (multiple true)

a)

  1. It is exactly the same as gang scheduling

b)

  1. jobs are then executed in an alternating fashion so that only one of them obtains the exclusive access to the resources at a time.

c)

  1. Stopping lower-priority jobs to permit the execution of higher-priority jobs

d)

  1. is implemented using a variant of gang scheduling

88.

  1. What are generic resources (GRES) in Slurm?

a)

  1. Slurm term to refer to the nodes themselves

b)

Resources that cannot be allocated in a partition

c)

  1. large array of resources that are generic and do have restrictions

d)

other hardware devices associated with nodes, most commonly accelerators.

89.

What is Slurm primarily used for in HPC environments?

a)

Data storage

b)

Network security

c)

Graphics rendering

d)

  1. Virtualization

e)

  1. Task scheduling

90.

If a user submits a job using the 'srun' command in Slurm without specifying any nodes, what will Slurm do?

a)

The job will be rejected.

b)

The job will be executed on the master node.

c)

  1. Slurm will choose appropriate nodes based on available resources.

d)

The job will be executed on all nodes in the cluster.

e)

  1. Slurm will prompt the user to specify nodes.

91.

  1. Which command in Slurm provides information about current job status, node availability, and other cluster states?

a)

srun

b)

sinfo

c)

spart

d)

sstate

92.

  1. Imagine a user needs to run a job that requires 4 nodes and 32 tasks with 8 tasks on each node. Which of the following Slurm commands correctly specifies this requirement?

a)

  1. srun -N4 -n32 --ntasks-per-node=8 job_script.sh 

b)

  1. srun -N4 -n8 --ntasks-per-node=32 job_script.sh

c)

srun -N32 -n4 --ntasks-per-node=8 job_script.sh

d)

  1. N8 -n4 --ntasks-per-node=32 job_script.sh

93.

  1. Which of the following statements about Slurm's job priority mechanism is true?

a)

Jobs are executed based on the order they are received.

b)

  1. Priority is determined only by the age of the job.

c)

  1. Priority is based on a combination of factors including job size, age, and partition configurations.

d)

Larger jobs always have higher priority.

94.

  1. Alice is working on a climate modeling project and has three large simulation tasks to run. Each simulation requires 10 nodes. She wants to ensure that if one simulation fails, it doesn't impact the others. Which of the following Slurm commands should she use to submit these simulations? Note: In Slurm, the sbatch command is used to submit a job script for later execution.

a)

  1. srun -N10 simulation1.sh; srun -N10 simulation2.sh; srun -N10 simulation3.sh 

b)

  1. sbatch -N10 simulation1.sh && sbatch -N10 simulation2.sh && sbatch -N10 simulation3.sh

c)

srun -N30 simulation1.sh simulation2.sh simulation3.sh

d)

sbatch --array=1-3 -N10 simulation.sh

95.

  1. Bob is a bioinformatics researcher running a genome assembly task. He realized that his job might run for several days. He wants to prevent the job from being terminated if the cluster undergoes a short maintenance. Which Slurm option should Bob use to make sure his job can be resumed after a node failure or maintenance? 

Note: In Slurm, the --checkpoint option allows users to create periodic checkpoints of their jobs, ensuring that they can be restarted from the last checkpoint if interrupted.


a)

  1. –hold 

b)

  1.  –defer 

c)

 –checkpoint

d)

  1.  –persist 

96.

  1. Caroline, a data scientist, has been given access to an HPC cluster for her deep learning experiments. For her initial tests, she wants to run her TensorFlow training script on a GPU node. Which of the following Slurm commands would be appropriate for her to request a GPU node? Note: In Slurm, the --gres=gpu option is used to request generic resources, specifically GPUs.

a)

  1.  srun --gres=gpu:1 tensorflow_script.sh 

b)

srun --gpus=1 tensorflow_script.sh

c)

srun --request-gpu tensorflow_script.sh

d)

  1. sbatch --gpu tensorflow_script.sh 

97.

  1. David is a physicist who needs to run a simulation in parallel on 16 cores. However, he knows that his simulation can sometimes be memory-intensive. Which Slurm command should he use to request 16 cores and 2GB of memory per core? Note: The --mem-per-cpu option in Slurm is used to specify the amount of memory required per CPU.

a)

  1. srun -n16 --mem=32GB simulation.sh 

b)

srun -n16 --mem-per-core=2GB simulation.sh

c)

  1. srun -n16 --mem-per-cpu=2GB simulation.sh

d)

  1. srun --cpus=16 --memory=2GB simulation.sh 

98.

  1. Emily is managing an HPC cluster and notices that a user has mistakenly submitted a job that's using excessive resources, affecting other users. Which Slurm command should she use to remove that specific job from the queue? Note: In Slurm, the scancel command is used to terminate running or pending jobs.

a)

  1. sdelete JOBID 

b)

sterm JOBID

c)

  1. sremove JOBID 

d)

  1. scancel JOBID 

99.

What are the primary functions of the OS? (multiple true)

a)

exploit the hardware resources of one or more processors

b)

provide a set of services to system users

c)

  1. manage secondary memory and I/O devices

d)

interface between the hardware and application programs

100.

  1. What is the main function of the command interpreter an OS?

a)

  1. to provide the interface between the API and application program

b)

  1. to handle the files in the operating system

c)

to get and execute the next user-specified command

d)

  1. none of the mentioned

101.

To access the services of the operating system, the interface is provided by the

a)

  1. Library

b)

  1. System calls

c)

  1. Assembly instructions

d)

  1. API

102.

CPU scheduling is the basis of

a)

  1. multiprogramming operating systems

b)

  1. larger memory sized systems

c)

multiprocessor systems

d)

none of the mentioned

103.

  1. Which one of the following is NOT true?

a)

  1. kernel remains in the memory during the entire computer session

b)

kernel is made of various modules which can not be loaded in running operating system

c)

  1. kernel is the first part of the operating system to load into memory during booting

d)

kernel is the program that constitutes the central core of the operating system

104.

  1. In operating system, each process has its own _______

a)

open files

b)

  1. pending alarms, signals, and signal handlers

c)

  1. address space and global variables

d)

  1. all of the mentioned

105.

In a timeshare operating system, when the time slot assigned to a process is completed, the process switches from the current state to?

a)

  1. Suspended state

b)

  1. Terminated state

c)

  1. Ready state

d)

Blocked state

106.

  1. Which of the following do NOT belong to queues for processes?

a)

  1. job Queue

b)

  1. PCB queue

c)

  1. Device Queue

d)

  1. Ready Queue

107.

  1. What will happen when a process terminates?

a)

It is removed from all queues

b)

It is removed from all, but the job queue

c)

  1. Its process control block is de-allocated

d)

  1. Its process control block is never de-allocated

108.

  1. Which of the following best describes the primary role of an operating system?

a)

  1. Facilitating communication between user and hardware. 

b)

  1. Only managing the computer's memory.

c)

Only executing application programs.

d)

  1. Managing hardware and providing services for computer programs.

109.

  1. In OS structures, what does the 'system call' typically do?

a)

Calls the system administrator for help.

b)

  1. Provides a way for the OS to interrupt system processes.

c)

  1. Provides a way for the application to request services from the OS.

d)

Requests external devices to stop functioning.

e)

  1. Optimizes the memory for application use

110.

  1. Consider a multi-threaded process. Which of the following would be shared among all threads of that process?

a)

Registers.

b)

Stack.

c)

Heap.

d)

  1. Program Counter. 

111.

In the context of memory management, what is the primary purpose of a Translation Lookaside Buffer (TLB)?

a)

Store frequently used page table entries.

b)

Act as a secondary storage.

c)

  1. Directly manage the cache memory. 

d)

  1. Translate virtual memory to cache memory.

112.

  1. Which of the following best describes the concept of 'paging' in memory management?

a)

Dividing physical memory into fixed-sized blocks

b)

  1. Dividing secondary storage into fixed-sized blocks.

c)

Dividing virtual memory into fixed-sized blocks.

d)

  1. Storing and retrieving data from cache.

113.

  1. BioTech Corp. uses an HPC system to analyze DNA sequences. Given the massive datasets, which OS feature related to memory would ensure the quickest access to frequently used data?

a)

  1. Virtual Memory

b)

  1. Translation Lookaside Buffer (TLB) 

c)

Swap Space Management

d)

Disk Partitioning

114.

  1. Cray Inc. recently launched a supercomputer dedicated to climate modeling. Given the nature of climate simulations, which HPC OS service would be PIVOTAL in ensuring the simultaneous execution of multiple tasks within a single simulation?

a)

Filesystem organization settings

b)

User authentication services

c)

  1. Thread management 

d)

Power optimization

115.

  1. AlphaTech, a leading tech firm, is building an HPC system for genome sequencing. The challenge lies in handling massive datasets that often exceed the physical memory. Which memory management technique would be ESSENTIAL for the HPC OS to ensure continuous processing without running into memory limits?

a)

  1. Cache optimization 

b)

Disk partitioning

c)

Virtual memory implementation

d)

Register allocation

116.

  1. BetaLabs has an HPC setup for quantum mechanics simulations. They've noticed that despite having multiple processors, some tasks are not equally distributed, leading to inefficiencies. To address this, which OS process management service should they optimize?

a)

  1. Process spawning 

b)

Process scheduling

c)

  1. Process authentication 

d)

  1. Process partitioning

117.

  1. A research institute uses an HPC system for simulating neural networks. They've designed these simulations to run tasks in parallel, but there's often a need for tasks to communicate intermediate results. Which HPC OS service focused on threads would BEST facilitate this inter-task communication?

a)

Thread spawning

b)

Thread isolation

c)

Thread-level parallelism

d)

Thread synchronization

118.

What are examples of I/O channels that communicate with external devices (outside SMP)? (multiple true)

a)

user interfaces

b)

  1. data storage

c)

system area networks

d)

  1. Memory banks

119.


  1. Suppose that a calculation has a 4% serial portion,

                       a) what is the limit of speedup on 16 processors?

                                   b) What is the maximum speedup?


a)

  1. a)16/(1 + (16 + 1)*.04) = 9.5 ; b) 1/0.04 = 25

b)

  1. b)16/(1 + (16 – 4)*.06) = 9.3 ; b) 1/0.04 = 25

c)

c)16/(1 + (16 – 4)*.06) = 9.3 ; b) 1/0.06 = 16.6

d)

  1. d)16/(1 + (16 – 1)*.04) = 10 ; b) 1/0.04 = 25

120.

  1. The reason for the implementation of the cache memory is

a)

  1. To increase the internal memory of the system

b)

The difference in speeds of operation of the processor and memory

c)

  1. To reduce the memory access and cycle time

d)

All of the mentioned

Similar Resources on Wayground

Fundamentals of EnE

115 questions

Fundamentals of EnE

quiz

Worksheet

•

University

Maternal Newborn Nursing Exam 2

115 questions

Maternal Newborn Nursing Exam 2

quiz

Worksheet

•

12th Grade - University

farma katzung 3

124 questions

farma katzung 3

quiz

Worksheet

•

University

Evaluative Exam 2_Reviewer from Harr | Reviewer by Badong.SJO.

121 questions

Evaluative Exam 2_Reviewer from Harr | Reviewer by Badong.SJO.

quiz

Worksheet

•

University

Amino Acids

116 questions

Amino Acids

quiz

Worksheet

•

University

Anatomy and Physiology Fall Final Exam

120 questions

Anatomy and Physiology Fall Final Exam

quiz

Worksheet

•

9th Grade - University

UHS Y2 Practical Histology MCQs

115 questions

UHS Y2 Practical Histology MCQs

quiz

Worksheet

•

University