Research in Indiana researcher participation in SC2002 technical program

The Proteus Multiprotocol Message Library

Session: Message Passing
Date: Wednesday, November 20
Time: 10:30am - 12:00pm

AUTHORS: Kenneth Chiu (Indiana University)
Madhusudhan Govindaraju (Indiana University)
Dennis Gannon (Indiana University)

ABSTRACT: Grid systems span manifold organizations and application domains. Because this diverse environment inevitably engenders multiple protocols, interoperability mechanisms are crucial to seamless, pervasive access. This paper presents the design, rationale, and implementation of the Proteus multiprotocol library for integrating multiple message protocols, such as SOAP and JMS, within one system. Proteus decouples application code from protocol code at run-time, allowing clients to incorporate separately developed protocols without recompiling or halting. Through generic serialization, which separates the transfer syntax from the message type, protocols can also be added independently of serialization routines. We also show performance-enhancing mechanisms for Grid services that examine metadata, but pass actual data through opaquely (such as adapters). The interface provided to protocol implementors is general enough to support protocols as disparate as our current implementations: SOAP, JMS, and binary. Proteus is written in C++; a Java port is planned.

Interoperable Web Services for Computational Portals

Session: Web Services
Date: Wednesday, November 20
1:30pm - 3:00pm

AUTHORS: Marlon Pierce (Indiana University)
Geoffrey Fox (Indiana University)
Choonhan Youn (Syracuse University)
Stephen Mock (San Diego Supercomputer Center)
Kurt Mueller (San Diego Supercomputer Center)
Ozgur Balsoy (Florida State University)

ABSTRACT: Computational web portals are designed to simplify access to diverse sets of high performance computing resources, typically through an interface to computational Grid tools. An important shortcoming of these portals is their lack of interoperable and reusable services. This paper presents an overview of research efforts undertaken by our group to build interoperating portal services around a Web Services model. We present a comprehensive view of an interoperable portal architecture, beginning with core portal services that can be used to build Application Web Services, which in turn may be aggregated and managed through portlet containers.

Introduction to Parallel Programming with OpenMP

Session: M11
Date: Monday, November 18
Time: 8:30am - 12pm (half day)

Tim Mattson, Intel
Rudolf Eigenmann, Purdue University

The OpenMP Application Programming Interface (API) defines compiler directives and library routines that make it relatively easy to create programs for shared memory computers. It first appeared in 1997 and has become the de facto standard for programming shared memory computers.

In this tutorial, we will provide a comprehensive overview of OpenMP. The OpenMP specifications are simple enough that in half a day we will be able to cover the entire API. We will start with basic concepts in programming shared memory computers and then move onto the basic constructs in OpenMP. This will include thread creation, assigning work to threads, managing data within the program and finally synchronization of threads.

The focus of the tutorial is using OpenMP to write real applications. Therefore, we will demonstrate the OpenMP constructs using excerpts from real applications. We will use these application to explain how to use the basic OpenMP constructs, but also how to use them effectively to maximize performance and robustness.

Computational biology and high performance computing

Date: Tuesday, November 19
Time: 1:30-3:00pm
Room: 307/308

Moderator: Craig Stewart, Indiana University
Panelists: Chris Johnson(Utah); John Reynders (Celera); David Bader (New Mexico); Debra Goldfarb (IDC); Rick Stevens (Argonne National Laboratories and the University of Chicago)

The biological/biomedical/bioinformatic/genomic/proteomic sciences now require the use of massive high performance computing (HPC) resources. In fact, biological sciences may more than double the size of the HPC market according to some estimates. This panel will be drawn primarily from the user community - people whose background is in the biological or chemical sciences - who have been successful in implementing HPC solutions to current biological and biomedical problems.

The questions to be addressed by this panel include the following:

• What has been the nature of the problems and the HPC solutions when HPC technologies have been critical to the solution of important problems?
• What are the current best practices in use of HPC in biological/biomedical sciences?
• Where has use of HPC approaches not yet been helpful in solving biological problems, and why not?
• What are the big future problems in the biological sciences that will require use of HPC systems?
• What does the biological research community need from the HPC community in order to foster better and more rapid implementation of HPC solutions to biological problems?
• What hardware architectures are most effective now, and what will the most effective hardware architectures be in 5 years?
• What are the software frameworks and applications that most effectively facilitate solution of important biological and biomedical problems?

Building a Massive Data Storage Infrastructure for the Masses

AUTHORS: Anurag Shankar (Indiana University)
Gustav Meglicki (Indiana University)
Jeff Russ (Indiana University)
Haichuan Yang (Indiana University)
Chris Garrison (Indiana University)

ABSTRACT: At Indiana University, we have built a massive data storage system (with a 500 Terabyte capacity) using the High Performance Storage System (HPSS) software. Using the secure, Distributing Computing Environment Distributed File System (DCE DFS) interface to HPSS and gatewaying software, we have extended the reach of a high-end HSM system to the masses. A user at IU is now able to access Terabytes of data with easy-to-use, secure access methods from their Unix, Linux, Windows, or Mac desktops. We use gateway software such as secure Apache server, Samba, Netatalk, and secure FTP to make this possible. We have also developed a distributed storage infrastructure which operates over a fast, wide-area network (WAN) between our Bloomington and Indianapolis campuses, located around 50 miles apart. Finally, we are implementing remote dual copies of data across the WAN for disaster recovery.

The Grid Services Flow Language

AUTHORS: Patrick Wagstrom (Illinois Institute of Technology and Argonne National Laboratory)
Sriram Krishnan (Indiana University and Argonne National Laboratory)
Gregor Von Laszewski (Argonne National Laboratory)

ABSTRACT: GSFL provides an XML-based language that allows the specification of the workflow description for Grid services. The salient features of the language are the list of service providers and activities that are part of the workflow, the composition model which describes the interaction between the individual services allowing their recursive composition, and the lifecycle model which describes the lifecycle for the various activities and services that are part of the workflow. The language provided is rich enough to model data and control flow between services and support recursive construction of workflow into larger services. GSFL is built on top of the OGSI Technology Preview. Rather than coding the XML bindings by hand, tools are used to automatically generate data bindings, allowing us to always keep current with the evolving standards. The information from a GSFL workflow can then be used to automatically generate the WSDL for a workflow instance. A generic GSFL Coordinator is spawned for each workflow instance by a GSFL Provider, which intercepts all calls to the instance and maps it to a generic coordinator function. Thus it allows the creation of "virtual ports" that behave as standard WSDL ports, to encapsulate dynamically generated methods of the workflow.

Microsoft .NET

Date: Wednesday, November 20
Time: 5:30 - 7:30pm
Room: 314/315

Andrew Lumsdaine, Indiana University

Microsoft has recently announced .NET as a platform for supporting a wide variety of services and applications. Although at this early stage, it is not clear that .NET will be a suitable platform for high-performance computing, the certain ubiquity of .NET in the near future indicates that the technology should at least be investigated. The purpose of this BoF is to gauge community interest in .NET as a platform for high-performance computing and to initiate the process of creating a on-going discussion and development of .NET as it relates in particular to high-performance computing.