Tutorials

Location: The tutorials are located at the Collegium Conference Center, see [directions]. Some tutorials will have hands-on sessions in the afternoon; these are held in terminal rooms at the Dept of Mathematics at the university, see [directions].

Lunch is not included in the tutorial fee.

Further questions: If you have any questions, please contact Patrick Norman.

Tutorials:
October 17, 2005

T1: Linux clusters
10:00-12:30 Using Linux Clusters at NSC
14:00-16:30 Beowulfs—An introduction to design and installation

T2: Grid computing
Cancelled

T3: Matlab for parallel computing
10:00-12:30 tutorial session
14:00-17:00 hands-on session

T4: VASP, a periodic electronic structure program
10:00-12:30 tutorial session
14:00-17:00 hands-on session

T1: Linux Clusters

Purpose and Topics:
Part I: Using Linux Clusters at NSC
The purpose with this tutorial is to give the participant a crash course in how to use linux clusters. Even though the tutorial will focus on the NSC Cluster Environment (NCE) it will also include a fair amount of generic information.

• Logging in and security
• Software environment overview
• Compilers
• Available MPI implementations
• The Maui scheduler
• Understanding system load
• Compiling MPI applications
• Running interactively
• Running in batch

Part II: Beowulfs—An introduction to design and installation
This tutorial is an introduction for anyone who are going to accuire and install a beowulf for high performance computing. It will cover both the time before the purcase and the initial installation after. Installation and execution of application software is no covered in this tutorial. Practial issues will, of natural reasons, be slightly coloured by our experience of Beowulfs at National Supercomputer Centre.

The tutorial consist of two parts. In the first part we focus on the process of creating the hardware specification of a Beowulf system for high performance computing. We will have the cost of solution in mind when we address the questions.

• Why Beowulf—a short motivation
• Typical configurations
• Performance requirements
• A glimpse of current processor architectures
• Interconnects (Ethernet, SCI, Myrinet, Infiniband, others)
• Benchmarking
• Floorspace, Power and Cooling
• Reliablity, Accessibility, Serviceability

• Open source vs. commercial solutions
• Installation procedure
• The Linux operating system and distributions
• Job management, queuing systems, scheduling
• Compilers, MPI, libraries
• Managing nodes and users
• Monitoring
• Security issues

Relevant web-links:
National Supercomputer Centre
Beowulf.org

Prerequisites:
In the morning session, a basic knowledge of Unix is required. In the afternoon session, a basic knowledge of PC hardware architecture is assumed and a knowledge of Linux on administration level is required.

T2: Grid computing

T3: Matlab for parallel computing

Purpose and Topics:
The purpose of this tutorial is to present an easy-way-in to the use of supercomputing facilities for Matlab users. With minimal efforts required from the user, Matlab codes can be executed on the most powerful cluster machines.

Star-P by Interactive Supercomputing is a high-performance parallel computing software solution specifically designed for cluster CPU or HPC server computing applications. Star-P takes the key features of MATLAB functions and shifts the model to the parallel universe of supercomputing. Star-P enables popular simulation, modeling and analysis to occur more quickly enabling faster algorithm development and better performance, saving significant testing time.

Relevant web-links:
Interactive Supercomputing
Star-P

Prerequisites:
A basic knowledge of Matlab (or similar sofware) is recommended.

T4: VASP, a periodic electronic structure program

Purpose and Topics:
There is an increased use of the VASP program at the Swedish HPD centra. NSC is therefore glad to be able to present an opportunity for VASP users to meet a program expert from the Vienna group, since we believe this will help researchers to get the most out of the program.

The morning session will be given in the form of a lecture, whereas the afternoon session will be held in a terminal room and provide a hands-on experience with VASP. There will be suitable exercises for the laboration, but we encourage participants to prepare their own topics for discussion with Dr Vogtenhuber based on their research experiences. A list of topics for the morning lecture reads as:

• short remarks on DFT and program algorithms
• discussion on the included potentials
• capabilities
• limitations
• file handling (input/output)
• understand the structure of input/output
• speeding things up, tricks and techniques
• understand the structure of the program
• running on a Linux cluster

Presentation material:

• Morning presentation (postscript): [here]
• Afternoon hands-on (postscript): [here]
• VASP examples (gzipped archive): [here]

Relevant web-links:
VASP homepage
VASP at NSC

Prerequisites:
A basic knowledge of VASP (or similar software) is recommended.