Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Introduction to Parallel and Distributed Systems
Stephane Genaud
November 6, 2014
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 1 / 13
Outline
1 Introduction on Use Cases
2 Broad fields of parallelism
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 2 / 13
Table of Contents
1 Introduction on Use Cases
2 Broad fields of parallelism
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 3 / 13
Use Case 1 - Fluid Simulation Visualization
Fluid SimulationIn a fluid dynamics lab, visualization of fluid circulation.Need: from the numerical simulation data, render one particularinstant of the simulation (107 points, 8 GB data in RAM)
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 4 / 13
Use Case 1 - Car Model
In the car industry, a R&D department works on CAD models.Need: compute the heat transfer from the engine to the rest ofthe vehicle on a 108 cell mesh (100 GB of data in mesh).
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 5 / 13
Use Case 2 - Scene Rendering
In the film industry, produce an animated cartoon based ondigital images.Need: compute the scene rendering for 105 images
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 6 / 13
Use Case 3 - RC5-72
Breaking a Ciphering KeyFind the cryptographic keyused to cipher a message.Need: scan all possible keysusing brute force and findwhich one matches.See for example the RSASecret-Key Challenge
Archi. Distributed.net
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 7 / 13
Use-Case 4: Social Network
Compute properties on the graphNeed: Recommend a user new connections, i.e people he/shemay know by transitivity. Do it for 106 users.
Graph
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 8 / 13
Table of Contents
1 Introduction on Use Cases
2 Broad fields of parallelism
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 9 / 13
Multi Threaded
key technologies: OpenMP - CUDA - OpenCL
architecture : single host
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 10 / 13
Message-passing
key technologies: MPI
architecture : multi-hosts connected through fast network(clusters)
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 11 / 13
Bag-of-tasks
key technologies: BOINC - Globus derivatives - MAUI/Torque
architecture : distributed hosts connected through LANs orWANs. Grids, Clouds, Internet Computing
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 12 / 13
MapReduce
key technologies: MapReduce - Hadoop
architecture : multi-hosts connected through fast network(clusters)
Stephane Genaud Introduction to Parallel and Distributed Systems November 6, 2014 13 / 13