Description |
The availability of high-speed networks and increasingly powerful commodity microprocessors are making the usage of clusters, or networks, of computers an appealing vehicle for cost effective parallel computing. Clusters, built using commodity-of-the-shelf COTS) hardware components as well as free, or commonly used, software, are playing a major role in redefining the concept of high performance computing.\nThe School aims to provide the skills needed to benefit from this generation of HPC solutions, giving a basic knowledge of programming, administering and tuning, as well purchasing or building Linux-based clusters.\n\nAs a provisional program there will be lectures on : Interconnections solutions(characteristics and performances) Parallel programming techniques Optimization and profiling techniques for clusters Operating system issues: parallel file systems and IO\n System Administration of a Linux cluster These lectures will be complemented by hands-on lab sessions where participants, grouped in small teams, will build their own cluster from scratch; during the first week of the school these clusters will be assembled, configured and tested. In the second part of the course, some representative parallel codes in the areas of computational condensed matter, engineering and weather forecast will be distributed to the participants. These codes will be presented in a series of tutorials illustrating their usage and the parallelisation strategies they adopt. During Lab sessions, the participiants will be asked to install, analyse and profile some of these codes on the previously built clusters |
JOINT ICTP-INFM SCHOOL IN "HIGH PERFORMANCE COMPUTING ON LINUX
CLUSTERS" | (smr 1395)
Go to day
-
-
09:00 - 11:00
REGISTRATION and administrative formalities
Location: Adriatico Guest House (Lower Level 1) -
09:00
REGISTRATION and administrative formalities
2h0'
-
09:00
REGISTRATION and administrative formalities
2h0'
-
11:00 - 12:00
Introduction
A brief presentation of school and an overview of HPC and the impact of linux cluster approach
-
11:00
Introduction
1h0'
Speaker: S.Cozzini (INFM udr. Sissa, Trieste, Italy) Material: lecture notes
related links
-
11:00
Introduction
1h0'
-
12:00 - 13:00
The Linux O.S. : an overview
An overview of the Linux .O.S. will be presented. Pros and Cons of this OS with respect to other operating systems will be adressed and discusses
-
12:00
The Linux O.S. : an overview
1h0'
Speaker: C. Fonda (ICTP) Material: lecture notes
-
12:00
The Linux O.S. : an overview
1h0'
-
13:00 - 15:00
Lunch
Location: Main Building Cafeteria - 13:00 Lunch 2h0'
-
15:00 - 15:15
Lab Session: presentation
Introduction: how to use the lab sessions
Location: Adriatico Guest House Informatics Lab. -
15:00
Lab Session: presentation
15'
Speaker: S. Cozzini Material: lecture notes
-
15:00
Lab Session: presentation
15'
-
15:15 - 16:15
Again on O.S. Linux System
II part:
Location: Adriatico Guest House Informatics Lab. -
15:15
Again on O.S. Linux System
1h0'
Speaker: C. Fonda Material: lecture notes
-
15:15
Again on O.S. Linux System
1h0'
-
16:15 - 18:15
Practical: Install your own linux box
Every student is supposed to install a Linux Box.
Location: Adriatico Guest House Informatics Lab. -
16:15
Practical: Install your own linux box
2h0'
Speaker: C. Fonda Material: lecture notes
-
16:15
Practical: Install your own linux box
2h0'
-
09:00 - 11:00
REGISTRATION and administrative formalities
-
-
09:00 - 10:50
Cluster Hardware: PC and networks
The basic hardware to build a linux cluster
Location: -
09:00
Cluster Hardware: PC and networks
1h50'
Speaker: R. Innocente (SISSA) Material: lecture notes
-
09:00
Cluster Hardware: PC and networks
1h50'
-
10:50 - 11:10
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 10:50 Coffee Break 20'
-
11:10 - 13:00
Introduction to the Lab Network
An introduction to linux networking and to the network organization of our lab.
Location: -
11:10
Introduction to the Lab Network
1h50'
Speaker: C. Fonda Material: lecture notes
-
11:10
Introduction to the Lab Network
1h50'
-
13:00 - 14:30
Lunch
Location: Main Building Cafeteria - 13:00 Lunch 1h30'
-
14:30 - 17:30
Tutorial: Oscar for linux cluster
OSCAR is a fully integrated easy to install bundle of software designed to make it easy to build and use a cluster for high performance computing. Everything you need to build, maintain, and use a modest sized Linux cluster is included in OSCAR. We invite students to read and study the oscar installation guide ( see more information link)
Location: -
14:30
Tutorial: Oscar for linux cluster
3h0'
Speaker: J. Enos (NCSA) Material: lecture notes
related links
-
14:30
Tutorial: Oscar for linux cluster
3h0'
-
17:30 - 18:00
Forming of Groups
Working Groups are formed. Short presentation among members of the group.
Location: Adriatico Guest House Informatics Lab. -
17:30
Forming of Groups
30'
-
17:30
Forming of Groups
30'
-
18:30 - 20:00
Reception
Location: Adriatico Guest House (Lower Level 1) - 18:30 Reception 1h30'
-
09:00 - 10:50
Cluster Hardware: PC and networks
-
-
09:00 - 11:10
Practical: Install Oscar
Location: Adriatico Guest House Informatics Lab. -
09:00
Practical: Install Oscar
2h10'
Speaker: J. Enos
-
09:00
Practical: Install Oscar
2h10'
-
11:10 - 11:30
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 11:10 Coffee Break 20'
-
11:30 - 12:30
Students' Presentation
How we build our first linux cluster...
Location: Adriatico Guest House Informatics Lab. -
11:30
Students' Presentation
1h0'
Material: lecture notes
-
11:30
Students' Presentation
1h0'
-
09:00 - 11:10
Practical: Install Oscar
-
-
09:00 - 10:50
Parallel Programming in MPI (I)
Location: -
09:00
Parallel Programming in MPI (I)
1h50'
Speaker: C. Cavazzoni (Cineca) Material: lecture notes
-
09:00
Parallel Programming in MPI (I)
1h50'
-
10:50 - 11:10
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 10:50 Coffee Break 20'
-
11:10 - 13:00
Practical: MPI examples
Run your first MPI programs
Location: Adriatico Guest House Informatics Lab. -
11:10
Practical: MPI examples
1h50'
Speaker: C. Cavazzoni Material: related links
-
11:10
Practical: MPI examples
1h50'
-
13:00 - 14:30
Lunch
Location: - 13:00 Lunch 1h30'
-
14:30 - 16:20
Parallel Programming in MPI (II)
Location: -
14:30
Parallel Programming in MPI (II)
1h50'
Speaker: C. Cavazzoni
-
14:30
Parallel Programming in MPI (II)
1h50'
-
16:20 - 16:30
Break
Location: - 16:20 Break 10'
-
16:30 - 18:00
Practical: More MPI examples
Location: Adriatico Guest House Informatics Lab. -
16:30
Practical: More MPI examples
1h30'
Speaker: C. Cavazzoni
-
16:30
Practical: More MPI examples
1h30'
-
09:00 - 10:50
Parallel Programming in MPI (I)
-
-
09:00 - 10:50
Linux Cluster Management (I)
Location: -
09:00
Linux Cluster Management (I)
1h50'
Speaker: S. Martinelli (Cineca) Material: lecture notes
-
09:00
Linux Cluster Management (I)
1h50'
-
10:50 - 11:10
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 10:50 Coffee Break 20'
-
11:10 - 13:00
Tools for HPC
Location: -
11:10
Tools for HPC
1h50'
Speaker: S. Cozzini Material: lecture notes
related links
-
11:10
Tools for HPC
1h50'
-
13:00 - 14:30
Lunch
Location: - 13:00 Lunch 1h30'
-
14:30 - 15:30
MPI programming (III)
Location: -
14:30
MPI programming (III)
1h0'
Speaker: C. Cavazzoni
-
14:30
MPI programming (III)
1h0'
-
15:30 - 17:00
Practical: exercises on MPI programming
Location: Adriatico Guest House Informatics Lab. -
15:30
Practical: exercises on MPI programming
1h30'
Speaker: C. Cavazzoni Material: misc
-
15:30
Practical: exercises on MPI programming
1h30'
-
17:00 - 18:00
Practical: Install free math libraries and the Intel Fortran compiler
Location: Adriatico Guest House Informatics Lab. -
17:00
Practical: Install free math libraries and the Intel Fortran compiler
1h0'
Speaker: S. Cozzini Material: lecture notes
-
17:00
Practical: Install free math libraries and the Intel Fortran compiler
1h0'
-
09:00 - 10:50
Linux Cluster Management (I)
-
-
09:00 - 10:50
Linux Cluster Management (II)
Location: -
09:00
Linux Cluster Management (II)
1h50'
Speaker: S. Martinelli Material: lecture notes
-
09:00
Linux Cluster Management (II)
1h50'
-
10:50 - 11:10
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 10:50 Coffee Break 20'
-
11:10 - 13:00
Advanced Topic: Mosix cluster approach
Mosix Cluster will be presented and discussed. Please note: two attached articles are provided ( follow more information link) and students are invited to download and read them carefully before the lectures
Location: -
11:10
Advanced Topic: Mosix cluster approach
1h50'
Speaker: M. Bar (Mosix's Group) Material: lecture notes
misc
-
11:10
Advanced Topic: Mosix cluster approach
1h50'
-
13:00 - 14:30
Lunch
Location: - 13:00 Lunch 1h30'
-
14:30 - 15:30
A demo on Mosix
Location: -
14:30
A demo on Mosix
1h0'
Speaker: Moshe Bar
-
14:30
A demo on Mosix
1h0'
-
15:30 - 17:00
Practical: configuring PBS and run PBS
Location: Adriatico Guest House Informatics Lab. -
15:30
Practical: configuring PBS and run PBS
1h30'
Speaker: S. Martinelli
-
15:30
Practical: configuring PBS and run PBS
1h30'
-
17:00 - 18:45
Practical: Install Mosix Cluster
Location: Adriatico Guest House Informatics Lab. -
17:00
Practical: Install Mosix Cluster
1h45'
Speaker: S. Cozzini Material: lecture notes
-
17:00
Practical: Install Mosix Cluster
1h45'
-
09:00 - 10:50
Linux Cluster Management (II)
-
-
09:00 - 10:50
Again on Hardware: High -end Processors and High Speed Network
A survey of the high-end processors available and high speed networks
Location: -
09:00
Again on Hardware: High -end Processors and High Speed Network
1h50'
Speaker: R. Innocente Material: lecture notes
-
09:00
Again on Hardware: High -end Processors and High Speed Network
1h50'
-
10:50 - 11:10
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 10:50 Coffee Break 20'
-
11:10 - 13:00
Profiling and Optimization Techinque (I)
Location: -
11:10
Profiling and Optimization Techinque (I)
1h50'
Speaker: L. DeRose (ACTC -IBM)
-
11:10
Profiling and Optimization Techinque (I)
1h50'
-
13:00 - 14:30
Lunch
Location: - 13:00 Lunch 1h30'
-
14:30 - 18:00
Practical: Measure performances of processors/ networks and compare mpich vs LAM
Location: Adriatico Guest House Informatics Lab. -
14:30
Practical: Measure performances of processors/ networks and compare mpich vs LAM
3h30'
Speaker: M. Corbatto (SISSA) Material: lecture notes
-
14:30
Practical: Measure performances of processors/ networks and compare mpich vs LAM
3h30'
-
09:00 - 10:50
Again on Hardware: High -end Processors and High Speed Network
-
-
09:00 - 10:50
Profiling and Optimization Technique (II)
Location: -
09:00
Profiling and Optimization Technique (II)
1h50'
Speaker: L. DeRose
-
09:00
Profiling and Optimization Technique (II)
1h50'
-
10:50 - 11:10
Coffee Break
Location: - 10:50 Coffee Break 20'
-
11:10 - 13:00
Portable MPI Tools at Work - Cracking Performance Problems
Abstract Vampir, the leading MPI performance analysis tools, is now available in a new and improved version. Vampir features a streamlined user-interface, additional displays and source-code display, while keeping all the unique features of previous Vampir releases. This presentation will cover - brief introduction to Pallas, a leading european vendor of software tools for parallel computing. - Vampir, visualization and analysis of MPI programs, focus on 'news' - Vampirtrace, low overhead MPI profiling library, news on Linux - TotalView 5, multi-process debugger
Location: -
11:10
Portable MPI Tools at Work - Cracking Performance Problems
1h50'
Speaker: W. Krotz-Vogel (Pallas) Material: lecture notes
-
11:10
Portable MPI Tools at Work - Cracking Performance Problems
1h50'
-
13:00 - 14:30
Lunch
Location: - 13:00 Lunch 1h30'
-
14:30 - 16:30
Practical: Optimize these codes!
hands-on session on code optimization
Location: Adriatico Guest House Informatics Lab. -
14:30
Practical: Optimize these codes!
2h0'
Speaker: L. DeRose Material: lecture notes
-
14:30
Practical: Optimize these codes!
2h0'
-
16:30 - 18:00
Practical: Vampir at work
Location: Adriatico Guest House Informatics Lab. -
16:30
Practical: Vampir at work
1h30'
Speaker: W. Krotz-Vogel Material: lecture notes
-
16:30
Practical: Vampir at work
1h30'
-
09:00 - 10:50
Profiling and Optimization Technique (II)
-
-
09:00 - 11:00
Lab Session
A free lab session in order to 1. Complete/repeat previous day work 2. Install and run your own code 3. Test different Solution
Location: Adriatico Guest House Informatics Lab. -
09:00
Lab Session
2h0'
-
09:00
Lab Session
2h0'
-
11:00 - 11:20
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 11:00 Coffee Break 20'
-
11:20 - 12:30
Group Presentation
A 10 minutes presentation for each group where results obtained, problem encountered are presented and discussed
Location: Adriatico Guest House Informatics Lab. -
11:20
Group Presentation
1h10'
-
11:20
Group Presentation
1h10'
-
09:00 - 11:00
Lab Session
-
-
09:00 - 10:50
Advanced Topic: Parallel Filesytems GPFS
Location: -
09:00
Advanced Topic: Parallel Filesytems GPFS
1h50'
Speaker: Luiz DeRose
-
09:00
Advanced Topic: Parallel Filesytems GPFS
1h50'
-
10:50 - 11:10
Coffee Break
Location: - 10:50 Coffee Break 20'
-
11:10 - 12:30
Case study A: Engineering application
Location: -
11:10
Case study A: Engineering application
1h20'
Speaker: C. Borges (Universidade Federal do Rio de Janeiro) Material: lecture notes
-
11:10
Case study A: Engineering application
1h20'
-
12:30 - 14:30
Lunch
Location: - 12:30 Lunch 2h0'
-
14:30 - 15:30
case A tutorial
Location: Adriatico Guest House Informatics Lab. -
14:30
case A tutorial
1h0'
Speaker: C. Borges Material: lecture notes
-
14:30
case A tutorial
1h0'
-
15:30 - 17:00
Case A Practical
Location: Adriatico Guest House Informatics Lab. -
15:30
Case A Practical
1h30'
-
15:30
Case A Practical
1h30'
-
09:00 - 10:50
Advanced Topic: Parallel Filesytems GPFS
-
-
09:00 - 10:50
Benchmarking linux clusters
VIDEOCONFERENCE
Location: -
09:00
Benchmarking linux clusters
1h50'
Speaker: M. Guest (Daresbury Lab.) Material: lecture notes
-
09:00
Benchmarking linux clusters
1h50'
-
10:50 - 11:10
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 10:50 Coffee Break 20'
-
11:10 - 12:30
Case study B : Metereological Application
The Met Office climate model is a world class model used for the study of climate change. The model will be briefly described and examples of it's use for the understanding of the climate system will be shown. Experience and results of using the model on various 64 bit and 32 bit platforms and interconnects will be discussed.
Location: -
11:10
Case study B : Metereological Application
1h20'
Speaker: A. Heaps (Centre for Global Atmospheric Modelling Reading (UK)) Material: lecture notes
-
11:10
Case study B : Metereological Application
1h20'
-
12:30 - 14:30
Lunch
Location: Main Building Cafeteria - 12:30 Lunch 2h0'
-
14:30 - 15:30
Case B Tutorial
Location: Adriatico Guest House Informatics Lab. -
14:30
Case B Tutorial
1h0'
Speaker: A. Heaps Material: lecture notes
-
14:30
Case B Tutorial
1h0'
-
15:30 - 17:30
Case B Practical
Location: Adriatico Guest House Informatics Lab. -
15:30
Case B Practical
2h0'
-
15:30
Case B Practical
2h0'
-
09:00 - 10:50
Benchmarking linux clusters
-
-
09:00 - 10:50
Performances of parallel chemistry codes on Linux Cluster
VIDEOCONFERENCE
Location: -
09:00
Performances of parallel chemistry codes on Linux Cluster
1h50'
Speaker: M. Guest Material: lecture notes
-
09:00
Performances of parallel chemistry codes on Linux Cluster
1h50'
-
10:50 - 11:10
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 10:50 Coffee Break 20'
-
11:10 - 12:30
Case Study C: Condensed matter application
Location: -
11:10
Case Study C: Condensed matter application
1h20'
Speaker: S. Cozzini Material: lecture notes
-
11:10
Case Study C: Condensed matter application
1h20'
-
12:30 - 14:30
Lunch
Location: Main Building Cafeteria - 12:30 Lunch 2h0'
-
14:30 - 15:30
Case C tutorial
Location: Adriatico Guest House Informatics Lab. -
14:30
Case C tutorial
1h0'
Speaker: S. Cozzini Material: lecture notes
-
14:30
Case C tutorial
1h0'
-
15:30 - 17:30
Case C practical
Location: Adriatico Guest House Informatics Lab. -
15:30
Case C practical
2h0'
Speaker: S. Cozzini
-
15:30
Case C practical
2h0'
-
09:00 - 10:50
Performances of parallel chemistry codes on Linux Cluster
-
-
09:00 - 10:00
How to build and run a 128 Processor Cluster
Location: -
09:00
How to build and run a 128 Processor Cluster
1h0'
Speaker: S. Martinelli Material: lecture notes
-
09:00
How to build and run a 128 Processor Cluster
1h0'
-
10:00 - 11:00
How to build a cluster for Monte Carlo Simulation
Our experience in building a Beowulf cluster from scratch as a personal parallel computing facility will be presented. The technical solutions which has been implemented in building a cluster dedicated to quantum Monte Carlo simulations will be discussed.
Location: -
10:00
How to build a cluster for Monte Carlo Simulation
1h0'
Speaker: D. Galli (Dip. Fisica Milano) Material: lecture notes
-
10:00
How to build a cluster for Monte Carlo Simulation
1h0'
-
11:00 - 11:30
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 11:00 Coffee Break 30'
-
11:30 - 13:00
Practical: free exercises
Location: Adriatico Guest House Informatics Lab. -
11:30
Practical: free exercises
1h30'
-
11:30
Practical: free exercises
1h30'
-
13:00 - 15:00
Lunch
Location: Main Building Cafeteria - 13:00 Lunch 2h0'
-
14:30 - 15:30
The Enea Experience using Beowulf cluster
Location: -
14:30
The Enea Experience using Beowulf cluster
1h0'
Speaker: M.Celino Material: lecture notes
-
14:30
The Enea Experience using Beowulf cluster
1h0'
-
15:30 - 15:50
Student Presentation: "Diskless Linux clusters"
Diskless Linux clusters". The technical peculiarities of a cluster consisting of diskless Linux nodes are presented. Alternative choices for the installation of such a cluster are compared. A solution based on the GRUB bootloader, Devfs and Linux support for "root over NFS" is discussed in detail.
-
15:30
Student Presentation: "Diskless Linux clusters"
20'
Speaker: Ciro Cattuto (INFM- Perugia) Material: lecture notes
-
15:30
Student Presentation: "Diskless Linux clusters"
20'
-
15:50 - 16:10
Student presentation: "Some Improvements on Monte Carlo Calculations in QFT".
Location: -
15:50
Student presentation: "Some Improvements on Monte Carlo Calculations in QFT".
20'
Speaker: Antun Balaz Material: lecture notes
-
15:50
Student presentation: "Some Improvements on Monte Carlo Calculations in QFT".
20'
-
16:30 - 18:30
Final Group Presentations
Each Group will present results/comments on the three different test cases and all the activities of the school. 10 minutes each group. Slides and/or electronic presentation strongly encouraged
Location: -
16:30
Final Group Presentations
2h0'
-
16:30
Final Group Presentations
2h0'
-
19:00 - 21:00
Final Round Table
An informal round table with a buffet
Location: -
19:00
Final Round Table
2h0'
-
19:00
Final Round Table
2h0'
-
09:00 - 10:00
How to build and run a 128 Processor Cluster
-
-
09:00 - 09:20
Students' presentations:Parallel gravity simulation
Location: -
09:00
Students' presentations:Parallel gravity simulation
20'
Speaker: Bagla Jasjeet Singh Material: lecture notes
-
09:00
Students' presentations:Parallel gravity simulation
20'
-
09:20 - 09:40
Student Presentation: "Parallel Genetic Algorithm: A typical MPI Application"
MPI is used to develop a parallel genetic algorithm program to solve the TSP problem, a typical optimization problem. The decisions made in designing the algorithm and developing and optimizing the code are highlighted, so as to illustrate the process of building a parallel program.
Location: -
09:20
Student Presentation: "Parallel Genetic Algorithm: A typical MPI Application"
20'
Speaker: Alkarouri Muhammad Abdulmuneim Material: lecture notes
-
09:20
Student Presentation: "Parallel Genetic Algorithm: A typical MPI Application"
20'
-
09:40 - 10:00
Student Presentation: to be announced
Location: -
09:40
Student Presentation: to be announced
20'
Speaker: D. Soroka
-
09:40
Student Presentation: to be announced
20'
-
10:00 - 11:00
Future Trends: Java for HPC
Location: -
10:00
Future Trends: Java for HPC
1h0'
Speaker: M. Ronchetti (Universita' di Trento) Material: lecture notes
web page
-
10:00
Future Trends: Java for HPC
1h0'
-
11:00 - 11:30
Coffee Break
Location: Adriatico Guest House (Lower Level 1) - 11:00 Coffee Break 30'
-
11:30 - 12:30
Future Trends: GRID COMPUTING
Location: -
11:30
Future Trends: GRID COMPUTING
1h0'
Speaker: Francesco GIACOMINI (CNAF-INFN) Material: lecture notes
-
11:30
Future Trends: GRID COMPUTING
1h0'
-
12:30 - 12:45
Conclusions
Location: -
12:30
Conclusions
15'
Speaker: S.Cozzini, A. Nobile (ICTP)
-
12:30
Conclusions
15'
-
09:00 - 09:20
Students' presentations:Parallel gravity simulation