WORLDCOMP'08 Tutorial: Prof. Jesus Carretero
|Parallel Input/Output Techniques in High-Performance Computing Systems.
Prof. Jesus Carretero
Universidad Carlos III de Madrid, Spain
Date: July 16, 2008
Time: 6:00 - 9:30 PM
Location: Platinum Room
- Parallel Input/output architectures.
- MPI-IO mechanisms to exploit the I/O infractructures.
- Parallel file systems.
- Application optimizations.
I/O systems are the weak part of high-performance computer architectures, where much more research has been devoted to computing and network nodes. However, for the success of such systems it is very important to implement scalable parallel I/O systems to solve this I/O gap, and to use adequate programming methods to get advantage of the infrastructures.
This tutorial will address the topic of parallel I/O for high-performance computers. It will be divided in two parts. First, the tutorial will address parallel Input/output architectures, MPI-IO mechanisms to exploit the I/O infrastructures, parallel file systems, and application optimizations. Second, some examples of parallel I/O optimizations implemented at the computing research center of Universidad Carlos III de Madrid will be shown and compared with other parallel file systems as IBM's General Parallel File System (GPFS), and Parallel Virtual File System (PVFS) from Argonne National Laboratory, and the available interfaces to such parallel file systems such as MPI-IO. Some of these results are been developed in cooperation with Argonne National Labs, European Space Agency, and other research centers where input/output is a major topic.
The main goal of this tutorial is to understand the problem of input/ouput usage in parallel and distributed computing systems, and the solutions that can be provided by using parallel input/output systems. This tutorial will address:
This tutorial is intended for faculty, engineers, scientists, and graduate students interested on input/output topics and high-performance computing systems. They will learn how to use parallel I/O mechanisms to enhance programs performance in homogeneous and heterogeneous clusters of computers and grids.
Biography of Presenter
Jesús Carretero is Full Professor of Computer Architecture at Universidad Carlos III de Madrid (Spain), where he is responsible for that knowledge area since 2000. He is also Director of the Master in Administration and Management of Computer Systems that he founded in 2004. He also serves and has served as a Technology Advisor in several companies.
Dr. Carretero holds a PhD in Computer Science from the Universidad Politécnica de Madrid (Spain), where he was researching on parallel Input/Output systems since 1989 until 1997. In 1997 and 1998 he was a visiting researcher at Northwestern University of Chicago (Ill, USA). His research activity is currently centered on high-performance computing systems, distributed systems and real-time systems. He has participated and leaded several national and international research projects in these areas, founded by Madrid Regional Government, Spanish Education Ministry and the European Union. He also acts as a reviewer and editor of several journals, and is a Senior Member of IEEE.
Prof. Jesus Carretero
Computer Science Department
Universidad Carlos III de Madrid (Spain)