Distributed and High-Performance Computing
(Advanced Operating Systems D)

Computer Science Honours course 7933

  Dr Paul Coddington
  Department of Computer Science
  University of Adelaide
  Room 1052
  http://dhpc.adelaide.edu.au/
  paulc@cs.adelaide.edu.au

  July - October 2000

Course Home Page

http://dhpc.adelaide.edu.au/education/CS7933/

Course Objectives

This course is aimed at bridging the gap between undergraduate computer science and state-of-the-art research work in distributed and high-performance computing (DHPC). The course provides background material in distributed computing systems, parallel computing systems, and other high-performance computing systems. Some historical background is given on architectures for high performance computing, but the emphasis is on what the software developer needs to know to exploit high-performance hardware for maximising efficiency and performance. The course has a strongly applied outlook with numerous case studies and examples from industry and computational science.

The course will discuss the following questions and provide a close look at past and present research in distributed and high-performance computing.

What is Distributed Computing?
What is High-Performance Computing?
What is the relationship between the two?
Why are they important and to whom?
A review of some important systems.
An introduction into active research in these areas, looking at research efforts worldwide as well as those in our own DHPC project.

Course Contents

Definition of the following areas:

High-Performance Computing: Faster, cheaper, better,...
Parallel Computing: Extra performance or price/performance usually in a tightly coupled environment
Distributed Computing: Added functionality and reliability over a local or wide-area network
Grid Computing or Metacomputing: High-performance computing over wide-area networks and the Internet

We will cover the following broad areas:

Applications and Motivation: An overview of key applications areas of DHPC
Architectures and Applications: A historical look at some important architectural aspects of DHPC systems
Profiling, Measurement and Code Porting: Some examples of how to get code running efficiently on a new DHPC system, how to analyse/model/benchmark/optimize the code
Case Studies: Computational physics; Weather forecasting, etc.
The Research Problems: Security and Authentication; Name Spaces for Control and Storage; Heterogeneous Systems and Integration; Interoperability.
Where to Find More Information: Some web sites and key research groups around the world.

and in particular focus on the following DHPC technologies:

Data Parallelism: (HPF, HPC, ...)
Message Passing: (PVM, MPI, ...)
Shared Memory Systems: (CC-NUMA, OpenMP, Java threads, ...)
Distributed Environments: (DCE, RPC, CORBA, Java RMI, ...)
Metacomputing or Grid Computing: (Globus, Legion, DISCWorld, SETI@Home, ...)

The course is intended to cover as up-to-date material as possible, and will include lectures on recent areas of interest such as:

Cluster computing, and in particular presentation of software and ideas for clustered PCs (Beowulf Project), and new networking technologies such as Myrinet and VIA.
Some Java distributed computing technology such as Java Jini and the DISCWorld Java based middleware being developed by the DHPC group.

The lecture notes and an outline of the topics covered by each lecture for last year's course are available from the course home page. Some of last year's material will be changed to incorporate recent advances in DHPC technologies.

Pre-requisites

You must be able to program at least one of C, Java, or Fortran, and it will be helpful to be able to understand code presented in C and Java.

Administration

There will be 11 Lectures on the following dates:

-  1. Friday, July 28
-  2. Friday, August 4
-  3. Friday, August 11
-  4. Friday, August 18
-  5. Friday, August 25
-  6. Friday, September 1
-  7. Friday, September 8
-  8. Friday, September 15
-  -
-  -
-  9. Friday, October 6
- 10. Friday, October 13
- 11. Friday, October 20

All lectures are from 4.15pm - 6pm in the CS Dept Lecture Theatre on Level 2 of the Plaza Building.

Assessment

The course will be assessed as follows:

Practical programming exercise - choice of several: 40%
Research a topic - choice of several ( 2500 word Essay ): 40%
In addition there will be one other (shorter) Essay to be written earlier on in the course to allow time to provide feedback: 20%

Distributed and High-Performance Computing course,
Paul Coddington (paulc@cs.adelaide.edu.au),
July - October 2000.

Distributed and High-Performance Computing (Advanced Operating Systems D)