Morning Tutorials (8:30am - 12:00pm)

Tutorial 1 - Designing Next Generation Clusters with Infiniband: Opportunities and Challenges, D. Panda (Ohio State University).

Tutorial 2 - Using MPI-2: Advanced Features of the Message Passing Interface, W. Gropp, E. Lusk, R. Ross and R. Thakur (Argonne National Lab).

Afternoon Tutorials (1:30pm - 5:00pm)

Tutorial 3 - The Gridbus Toolkit for Grid and Utility Computing, R. Buyya (University of Melbourne).

Tutorial 4 - Building and Managing Clusters with NPACI Rocks, G. Bruno, M. Katz, P. Papadopoulos, and F. Sacerdoti, (NPACI Rocks group at San Diego Supercomputer Center), L. Liew and N. Ninaba (Singapore Computer Systems).

Short Descriptions

"Designing Next Generation Clusters with Infiniband: Opportunities and Challenges"

The emerging InfiniBand Architecture (IBA) standard is generating a lot of excitement toward building next generation high performance computing systems in a radical different manner. This is leading to the following common questions among many scientists, engineers, managers, developers, and users associated with Cluster Computing:

1) What is InfiniBand Architecture?

2) How is it different from other on-going developments and standardization effort such as Virtual Interface Architecture (VIA), PCI-X, Gigabit Ethernet, Rapid I/O, Hyper-transport, 3GIO/PCI-Express, TCP off-load engines, etc.?,

3) How does it perform compared to other proprietary cluster interconnects (Myrinet and Quadrics)?

4) What unique features and benefits does IBA bring to designing next generation cluster computing systems?

5) How to exploit novel features of InfiniBand to build clusters for high performance computing as well as cluster-based servers and datacenters?

This tutorial is designed to provide answers to the above questions. We will start with the background behind the origin of the IBA standard. Then we will make the attendees familiar with the novel features of IBA (uniform treatment of interprocessor communication and I/O; provision for multiple transport services and mechanisms to support QoS and protection in the network; hardware support for remote DMA, atomic, and multicast operations; support for virtual lanes and service levels; and support for low latency communication with Virtual Interface). We will compare and contrast the IBA standard with other on-going developments/standards. We will show how the IBA standard facilitates the next generation computing systems to be designed not only to deliver high performance but also RAS (Reliability, Availability, and Serviceability). Open research challenges in designing communication and I/O subsystems of next generation clusters with IBA will be outlined. Challenges in designing clusters for standard high-performance computing (with different programming models such as MPI, DSM, Get/Put) as well as cluster-based servers/file systems and data centers will be outlined. Performance numbers obtained on clusters with latest InfiniBand products and their comparisons with other proprietary interconnects (Myrinet and Quadrics) will be presented. The tutorial will conclude with an overview of on-going IBA related research projects, IBA products, and the market time frame for the IBA products.

Speaker Bio:

Dhabaleswar K. Panda is a Professor of Computer Science at the Ohio State University. He obtained his Ph.D. in computer engineering from the University of Southern California. His research interests include parallel computer architecture, high performance computing, user-level communication protocols, interprocessor communication and synchronization, network-based computing, and Quality of Service. He has published over 110 papers in major journals and international conferences related to these research areas. Dr. Panda and his research group members have been doing extensive research on InfiniBand. His research group is currently collaborating with Sandia National Laboratory, IBM T.J. Watson, and leading InfiniBand companies (Mellanox and InfiniSwitch) on designing various subsystems of next generation High Performance Computing systems with InfiniBand. The MVAPICH (MPI over VAPI for IBA) package developed by his research group (http://nowlab.cis.ohio-state.edu/projects/mpi-iba/) is being used by many organizations world-wide to extract the potential of IBA-based clusters for HPC applications.

Dr. Panda has served on Program Committees and Organizing Committees of several parallel processing and high performance computing conferences and on editorial boards for several parallel processing journals. He was General Co-Chair for the 2001 International Conference on Parallel Processing; Program Co-Chair of the 1999 International Conference on Parallel Processing, 1997 and 1998 Workshops on Communication and Architectural Support for Network-Based Parallel Computing (CANPC); Program Co-Chair of the Int'l Workshop on Communication Architecture for Clusters (CAC '01 and CAC '02); an Associate Editor of the IEEE Transactions on Parallel and Distributed Computing; Co-Guest-Editor for two special issue volumes of Journal of Parallel and Distributed Computing on "Workstation Clusters and Network-based Computing"' an IEEE Distinguished Visitor Speaker and an IEEE Chapters Tutorials Program Speaker. Currently, he is serving as a Program Co-Chair of International Workshop on Communication Architecture for Clusters (CAC '03). Dr. Panda is a recipient of the NSF Faculty Early CAREER Development Award, the Lumley Research Award (1997 and 2001) at the Ohio State University, and an Ameritech Faculty Fellow Award. Dr. Panda is listed as a distinguished scientist in "Who'sWho in America" and in "American Men & Women of Science".

"Using MPI-2: Advanced Features of the Message Passing Interface"

This tutorial is about how to use MPI-2, the collection of advanced features that were added to MPI (Message-Passing Interface) by the second MPI Forum. These features include parallel I/O, onesided communication, dynamic-process management, language interoperability, and some miscellaneous features. Implementations of MPI-2 (or significant subsets thereof) are now available both from vendors and from open-source projects. For example, the one-sided communication functions of MPI-2 are being used successfully in applications running on the Earth Simulator. In other words, MPI-2 can now really be used in practice.

This tutorial explains how to use MPI-2, particularly, how to use it in a way that results in high performance. We present each feature of MPI-2 in the form of a series of examples (in C, Fortran, and C++), starting with simple programs and moving on to more complex ones. We also discuss how to combine MPI with OpenMP. We assume that attendees are familiar with the basic message-passing concepts of MPI-1.

The tutorial will feature a hands-on session in which attendees will be able to run MPI-2 programs on their own laptops with the latest version of MPICH2, which we will distribute on CDs.

Speakers Bio:

William Gropp is a senior computer scientist and associate division director in the Mathematics and Computer Science Division at Argonne National Laboratory. His research interests are in adaptive methods for PDEs, software for scientific computing, and parallel computing. He was a member of the MPI Forum from the beginning and was one of the chapter authors in the MPI-2 standardization process. He is one of the designers of MPICH and is a co-author of the books Using MPI, Using MPI-2, and MPI - The Complete Reference: Volume 2, the MPI-2 Extensions.

Ewing Lusk is a senior computer scientist in the Mathematics and Computer Science Division at Argonne National Laboratory. His research interests are in portable parallel-programming libraries, performance visualization, and automated theorem proving. He was a member of the MPI Forum from the beginning and played a leading role in the MPI-2 standardization process. He is one of the designers of MPICH and is a co-author of the books Using MPI, Using MPI-2, and MPI - The Complete Reference: Volume 2, the MPI-2 Extensions.

Rob Ross is an assistant computer scientist in the Mathematics and Computer Science Division at Argonne National Laboratory. He received a Ph.D. in Computer Engineering from Clemson University in 2000. His research interests are in the area of high-performance computing, particularly cluster computing and high-performance I/O. He is the primary author of PVFS, a high-performance parallel file system for clusters, and is currently involved in the implementation of the next-generation MPICH.

Rajeev Thakur is a computer scientist in the Mathematics and Computer Science Division at Argonne National Laboratory. He received a Ph.D. in Computer Engineering from Syracuse University in 1995. His research interests are in the area of high-performance computing in general and high-performance networking and I/O in particular. He was a member of the MPI Forum and participated actively in the definition of the I/O part of the MPI-2 standard. He is the author of ROMIO, a high-performance, portable implementation of MPI I/O. He is also a co-author of the book Using MPI-2 together with Bill Gropp and Rusty Lusk.

"The Gridbus Toolkit for Grid and Utility Computing"

Computational Grids enable the sharing, selection, and aggregation of geographically distributed resources (such as computers, data bases, scientific instruments) for solving large-scale problems in science, engineering, and commerce. However, application development, resource management, scheduling, and supporting end-to-end quality-of-services (QoS) in these environments is a complex undertaking. This is due to the geographic distribution of resources that are owned by different organizations having different usage policies and cost models, and varying loads and availability patterns. To address these challenges, we have developed distributed computational economy framework for resource allocation and regulation of supply-and-demand for resources. We applied this framework in the design and development of scheduling systems that manage distributed resources in a single administrative domain (cluster computing) and also in multiple administrative domains (grid computing).

The Gridbus Project is engaged in the design and development of cluster and grid middleware technologies for service-oriented computing. They include visual Grid application development tools for rapid creation of distributed applications, competitive economy-based Grid scheduler, cooperative economy-based cluster scheduler, Web-services based Grid market directory (GMD), Grid accounting services, and a widely used GridSim toolkit. These tools have been used in Grid-enabling applications such as molecular docking and neuroscience and deploying them for distributed proceedings on Global Grids.

This tutorial covers four topics. First, we briefly review emerging trends in network-based high performance computing and identify application development and resource management challenges. Then, we introduce our framework on Grid Architecture for Computational Economies (GRACE) that leverages existing technologies such as Globus and provides new services that are essential for constructing industrial-strength Grids. We discuss Gridbus technologies and their use in Grid enabling application, the use of our economic grid infrastructure in scheduling parametric computations containing hundreds of jobs for execution on the World Wide Grid (WWG) testbed. Particular emphasis will be placed on Grid economy, how to design and develop Grid technologies and applications capable of dynamically leasing services of distributed resources at runtime depending on their availability, capability, performance, cost, and users' quality of service requirements. Finally, we present the usage of tools in composition and distributed execution of data-intensive applications (e.g., molecular docking, brain activity analysis, and high-energy physics) on the Grid to demonstrate capabilities of Gridbus system.

Speaker Bio:

Rajkumar Buyya is the founder and program leader of the Grid Computing and Distributed Systems (GRIDS) Laboratory in the Dept. of Computer Science and Software Engineering at the University of Melbourne, Australia. He is one of the creators of system software for PARAM Supercomputers developed by the Centre for Development of Advanced Computing (C-DAC), India. He has authored three books Microprocessor x86 Programming, BPB Press, New Delhi, 1995, Mastering C++, Tata McGraw Hill Press, New Delhi, 1997, and Design of PARAS Microkernel. The books on emerging topics that he edited include, High Performance Cluster Computing published by Prentice Hall, USA, 1999; and High Performance Mass Storage and Parallel I/O, IEEE and Wiley Press, USA, 2001. He has published over 70 research articles in international conferences and journals. For further information, please visit: http://www.buyya.com/

"Building and Managing Clusters with NPACI Rocks"

NPACI Rocks is an open source clustering distribution for heterogeneous x86 and IA64 HPC Linux Clusters. The first version of Rocks was released in November of 2000, and has averaged 3-4 releases each year since. NPACI Rocks is unique in that it is a complete “cluster on a CD” software distribution, everything from the base Red Hat OS, to de facto standard job schedulers and monitoring systems are included. Rocks was designed with the goal of making clusters easy and accessible to domain application scientists. To achieve this goal, Rocks automates the common and most time consuming tasks of cluster administration and deployment. NPACI Rocks software and documentation is available at www.rocksclusters.org.

This tutorial assumes minimal experience with cluster administration and use. Experience with common clustering toolkits is useful but not required. We will cover the basic design and philosophy of Rocks and the procedures to customize Rocks for unique sites and non-HPC clustering roles. A hands-on lab will guide participants through building their own cluster. Toward the end of the lab, the individual clusters will be bound together as a Grid where each participant will launch Grid-wide jobs.

Speakers Bio:

Greg Bruno is a Programmer Analyst IV for the Cluster Development Group at the San Diego Supercomputer Center (SDSC). Mr. Bruno received his MS from UCSD in Computer Science and he is currently enrolled in the PhD program. For 10 years, Mr. Bruno worked with Teradata Systems developing cluster management software for the systems that supported the world’s largest databases. Recently, he has spent the last 3 years helping to architect, design and implement the Rocks Cluster Distribution, a freely available software stack that enables domain-specific scientists to build and manage their own clusters.

Mason Katz is Group Leader for Cluster Development at the San Diego Supercomputer Center (SDSC). Mr. Katz received his BS in Systems Engineering from the University of Arizona. He worked for 5 years as an embedded software engineer on networks of lightning detection sensors. Recently, he has spent the last 6 years working at both the University of Arizona, and UCSD/SDSC on projects ranging from network security protocols, operating systems (x-kernel, Scout), and commodity clustering (HPVM, NPACI Rocks).

Philip M. Papadopoulos is the Program Director for Grid and Cluster Computing at the San Diego Supercomputer Center (SDSC). Dr. Papadopoulos received his BA in Applied Mathematics from UCSD, MS in Mechanical Engineering from UC Berkeley, and PhD from UC Santa Barbara in Electrical and Computer Engineering. He worked for over 5 years at Oak Ridge National Laboratory as part of the PVM development team before moving the Computer Science Department at UCSD. In 1999, Dr. Papadopoulos joined SDSC as a group leader for cluster development and started the NPACI Rocks Clustering project. He has authored more than 25 peer-reviewed papers on cluster and distributed computing, has served on the organizing committees for several years on the SC conference series and has given numerous invited talks on cluster and distributed computing. Dr. Papadopoulos will be the technical program chair of the IEEE Clusters 2004 conference. Dr. Papadopoulos is deeply involved in several grid and cluster computing research projects including OptIPuter, GEON, The National Biomedical Computational Resource (NBCR), and the Biomedical Informatics Research Network (BIRN).

Federico Secerdoti is currently a member of the Cluster Development group, San Diego Supercomputer Center (SDSC), at the University of California. Mr. Sacerdoti received his BS in Computer Engineering fromWashington University in St. Louis, and MS in Compute Science from University of California, San Diego. Mr. Sacerdoti has been involved with the Rocks cluster distribution effort at SDSC for over three years, and has contributed to many HPC projects including the Ganglia monitoring system and the KeLP parallel message-passing libraries. His thesis work was in the area of dynamic cache optimizations for parallel applications.

Laurence Liew graduated from the National University of Singapore (NUS) with First Class Honors in Mechanical Engineering. He also holds a Masters in Knowledge Engineering from the Institute of Systems Science in NUS. Since 1998, Laurence has been actively involved with Linux and Beowulf supercomputing. He was involved with the design and deployment of Beowulf clusters in Singapore’s foremost R & D and academic institutions, and also the deployment of enterprise Linux solutions in major government and commercial organizations. Laurence joined Singapore Computer Systems in 2001 and presently heads the SCS Linux Competency Centre (LCC) in Singapore, Malaysia and Thailand. Under Laurence, the LCC have partnered world-class leaders in Linux and this includes Red Hat as a Red Hat Certified Training and Education Centre, Oracle, Sendmail, Computer Associates, Myricom, VMware and others.

Najib Ninaba graduated from Singapore Polytechnic with a diploma in Computer Information Systems. Najib has been a Linux user, programmer and administrator since 1996. He was introduced to Linux Beowulf supercomputing in 2001 and have since been hooked on NPACI Rocks. Najib joined Singapore Computer Systems in 2001 and currently leads the NPACI Rocks development in Singapore. A co-developer of NPACI Rocks, Najib integrated the Parallel Virtual Filesystem (PVFS) and Sun Grid Engine (SGE) into Rocks in 2002 and currently maintains the PVFS, SGE, Myrinet and other packages for NPACI Rocks.