IPDPS 2004 TUTORIAL

IPDPS 2004 is presenting three plenary tutorials, two on Monday and one on Friday. Registration for tutorials is in addition to registration for the symposium. They are not included in the technical program registration and require separate registration for each tutorial. See the IPDPS 2004 registration form. The tutorials are held on the same days as workshops and there will be morning and afternoon refreshment breaks for all participants.

Below are brief descriptions of each tutorial. We will be posting additional information and links, so please return to this page in the coming weeks. Meantime, for more information you may contact the individual tutorial organizer or the Tutorials Chair, Patricia Teller.

QUICK CLICK [ Tutorial 1 ] [ Tutorial 2 ] [ Tutorial 3 ]

Tutorial 1 – Monday Morning, April 26, 2004

High Performance Computing & Networking Using Infiniband Technology – Hardware and Software Solutions

Presenter: Hsing-bung Chen, Los Alamos National Lab

Level: %Introductory-20 • %Intermediate-60 • %Advanced-20

Description: Infiniband Architecture has been emerging as the next-generation interconnect for I/O and inter-process communication. A major target application of this new industry standard is to use a single switch fabric to design clusters and servers with high performance and scalabalility. Infiniband Architecture provides several mechanisms and services – multiple transport services (reliable connection, unreliable connection, reliable datagram, unreliable datagram, raw datagram), OS bypassed socket direct protocol, RDMA, atomic operations, multicast support, and virtual channels. The main features of IB that make it attractive for high-performance computing and networking are high bandwidth, low latency, moderate cost, and a competitive marketplace. Adapting Infiniband technology brings up a new challenge in design and development of modern hardware and software to take advantage of applying these interesting features.

This tutorial presents information on state-of-the-art hardware and software solutions using Infiniband technology in high performance computing areas such as MPI implementations, highly scalable parallel and distributed file systems, clusters, distributed systems, and SANs. This tutorial also explores how to efficiently program Infiniband transport services and how to best match Infiniband transport mechanisms with application characteristics. On-going Infiniband Architecture research will be addressed and discussed. Participants will benefit by:

Click for more information  |  Contact organizer

QUICK CLICK [ Tutorial 1 ] [ Tutorial 2 ] [ Tutorial 3 ]

Tutorial 2 has been cancelled
Tutorial 2 – Monday Afternoon, April 26, 2004

An Introduction to Distributed Data Mining

Presenter: Hillol Kargupta, University of Maryland

Level: %Introductory-20 • %Intermediate-60 • %Advanced-20

Description: This tutorial will present an overview of the field of distributed data mining (DDM). It will discuss several emerging DDM-application-areas, such as mining over wireless networks, large-scale distributed scientific data mining, and privacy-sensitive multi-party data mining. It will review a wide range of DDM algorithms for these application areas and points out some of the future directions of the field. The tutorial will also discuss some of the existing DDM systems and present a case study of one such system in details. The tutorial will also point the attendees toward publicly available DDM resources and literature.

The attendees will be exposed to the following aspects of this field:

Click for more information  |  Contact organizer

QUICK CLICK [ Tutorial 1 ] [ Tutorial 2 ] [ Tutorial 3 ]

Tutorial 3 – Friday All Day, April 30, 2004

Parallel Computing on Heterogeneous Networks

Presenters: Alexey L. Lastovetsky, University College Dublin and Alexey Ja Kalinov, Russian Academy of Sciences

Description: (Note: Tutorial presenters are the developers of mpC, and the tutorial will be based on the book “Parallel Computing on Heterogeneous Networks” by Alexey Lastovetsky published in June 2003 by John Wiley & Sons.)

In this tutorial, we present common networks of computers as a next milestone in the parallel architecture evolution that follows homogeneous distributed-memory multiprocessors. We formulate and analyze all the main programming challenges posed by this inherently heterogeneous parallel architecture. We will focus on how to develop efficient and portable parallel applications for heterogeneous networks of computers. We will use mpC to present basic approaches to heterogeneous parallel programming. A real-life mpC programming environment, mpC Workshop, will be used to demonstrate development, debugging and execution of the most interesting mpC applications.

From this tutorial the attendee should learn the following:

• A general logic view on the evolution of parallel architectures and related parallel programming models and tools
• Challenges of parallel programming for NoCs
• Basic approaches to design of heterogeneous parallel algorithms
• Performance models of heterogeneous parallel algorithms
• Approaches to parallel programming for heterogeneous NoCs using mpC
• How to implement a heterogeneous parallel algorithm in mpC
• How to debug and execute an mpC application.

Click for more information  |  Contact organizer