|
Ian
Foster
Computation
Institute Argonne National Laboratory & University
of Chicago
Service-Oriented
Science: Scaling eScience Impact
Computational approaches to problem solving have proven
their worth in many fields of science, allowing the collection and analysis
of unprecedented quantities of data, and the exploration via simulation of
previously obscure phenomena. We now face the challenge of scaling
the
impact of these approaches from the specialist to entire communities. I speak here about work that seeks to
address this goal by rethinking science's information technology foundations
in terms of service-oriented architecture.
In principle, service-oriented approaches can have a transformative
effect on scientific communities, allowing tools formerly accessible only to
the specialist to be made available to all, and permitting previously manual
data-processing and analysis tasks to be automated. However, while the
potential of such "service-oriented science" has been demonstrated,
its routine application across many disciplines raises challenging technical
problems. One important requirement is to achieve a separation of concerns
between discipline-specific content and domain-independent infrastructure;
another is to streamline the formation and evolution of the "virtual
organizations" that create and access content. I describe the
architectural principles, software, and deployments that I am and my colleagues
have produced as we tackle the first of these problems, and point to future
technical challenges and scientific opportunities.
|
|
Fran
Berman
Director, San Diego
Supercomputer Center
Professor and High Performance Computing Endowed Chair, UC San Diego
One hundred years
of data
The 20th century brought about an "information
revolution" which has forever altered the way we work, communicate, and
live. In the 21st century, it is hard to imagine working
without an increasingly broad array of enabling technologies and the data
they provide. Much of this data will form the foundation for new
discovery and advances for the next 100 years and beyond. In this talk
we focus on the development and deployment of Data Cyberinfrastructure, and
the challenges of developing a grid-based framework for data management and
preservation over the foreseeable future.
|
|
Andrew Grimshaw
Enterprise Data Integration with Grid
Technology
Both
compute grids and data grids have been used in academic computing. In the
extended enterprise (beyond a single site) the most compelling need is for
data grids to provide real-time, and near real-time, access to data
throughout the enterprise for both analytic purposes and to satisfy
operational needs. I begin my talk with a discussion of why compute
grids in the extended enterprise are not a priority for companies. I will
then examine the common data landscape in enterprises, and show how data grid
techniques can be used to provide data integration. I will contrast data
grids with alternatives such as data warehouses.
|
|
William
E. Johnston, ESnet
Lawrence Berkeley
National Laboratory
The Evolution of
Networks Over the Next 10 Years to Support Grid-Based, Large-Scale
International Science
The U. S. Department of Energy’s Office of Science
(SC) funds about 40% of the physical science research in the US. Much of this
research is related to the large facilities operated by SC – the high energy
accelerators like the Spallation Neutron Source and the Relativistic Heavy
Ion Collider, specialized instrument centers like the National Center for
Electron Microscopy, the Environmental Molecular Sciences Laboratory (EMSL),
and the Joint Genome Institute, several major supercomputer centers, etc. Additionally
SC is a major contributor to two of the largest-ever science experiments –
CERN’s Large Hadron Collider and the International Fusion Reactor (ITER) both
of which are located in Europe. All of these facilities have massive data
handling problems.
Each of these facilities is unique and involves
large and widely distributed collaborations. Each of these facilities has its
own requirements for providing collaborators access to the facility and for
managing and analyzing the data generated at the facility, frequently using
Grid-based data analysis and simulation approaches that involve computing and
storage systems that are not located at the science facility.
The network capabilities needed to support the data
movement and the Grid-based analysis systems are characterized in three ways:
by observing current trends in network usage, by characterizing the data
generating instruments of the facilities and the locations of the components
of the analysis and storage systems, and by case studies of how the
associated science community expects to be doing their science over the next
ten years.
In this talk I will discuss the network capacity and
functionality requirements that arise from the data movement and Grid
communication of large scientific instrument centers, the network
architecture and services needed to satisfy the requirements, and how these
will be provided to the science community.
|
|
Jack Dongarra
Innovative Computing Laboratory
Computer Science Dept. University of Tennessee
Challenges with Multicore and Specialized
Accelerators
|
|
Chris Jesshope
Managing Concurrency in Computer
Systems - A Journey Across Scale
Currently
only high performance computers need to manage concurrency.In the near
future, with the projected advances in chip density, the characteristics of
grid applications, i.e. massive concurrency and highly-latent asynchronous
communication, will need to be managed in on-chip environments. this means
all computer systems from embedded ones, through desktops right up the scale
to grid systems will need to manage concurrency explicitly. This represents
an enormous challenge for hardware manufacturers and scientists alike. It
also represents a major opportunity to design future computer systems that
make the problems of programming concurrency in a scale-independent manner
automatic, without any intervention from the programmer.
Ideally
the programmer should be able to program using a well understood and
completely deterministic model, such as the one that has been used (HPC
excepted) since the dawn of the computer age... namely the sequential one.
However, from this model all of the problematic issues, such as the full
extent of an application's concurrency, the distribution and scheduling
of that concurrency and the communication, synchronisation and latency that
this induces, must all be extracted by our compilers. Moreover, these
compilers must capture that information from the level of instruction sets at
the lowest end of our journey in scale right up to the high-level protocols
and layers of software that provide the framework for our grid systems. Some
would say that this is just a pipe dream, which is true if you look to
existing processor technology and software infrastructure. However, there
have been a number recent, disruptive advances computer architecture that
have begun to make this scenario look distinctly possible in the not too
distant future.
This talk
will review the issues involved in managing concurrency across this scale,
highlight the requirements and illustrate some solutions found in recent
concurrency models being introduced into microprocessor architecture. The
issues are of course the same as need to be extracted from our applications,
namely: concurrency capture, concurrency distribution that manages locality
and communication, dynamic scheduling that manages latency of communication
in an asynchronous environment and synchronisation to bind it all together.
These issues will be explored by reviewing a range of recent architectural
approaches in supporting concurrency from purely sequential code.
|
|
Dr. Frank Baetke - Global HPC
Technology Program Manager
Old and New Trends in High Performance
Computing
HP’s HPC
product portfolio is based on standards at the processor, node and
interconnect level. The strategic approach comprises a triangle with
computation, data management and visualization representing the sides of the
triangle. The success of such a standard-based strategy is reflected in major
wins across all market segments.
In
general trends in High Performance Computing can be separated into two basic
categories: one covers the directions towards smaller, faster and hopefully
cooler components, the other includes more innovative directions which might
lead to real paradigm shifts. A few of those will be addressed in this talk.
Since last
year certain changes became obvious specifically in the components which make
up the computational part. Those changes provide vast opportunities in terms
of system efficiency but will also lead to new challenges. The presentation
will focus on those developments within the context of UCP (Unified Cluster
Portfolio) and will also address the data management part known as SFS
(Scalable File Share) which is based on the Lustre cluster file-system. New
trends will also be covered for the visualization part which is based on an
architecture called SVA (Scalable Visualization Array).
For
all the segments mentioned above characteristic sites will be shown as
examples and a few hints concerning roadmaps and future trends will be added.
|
|
Juan Jose’ Porta
From gaming to serious supercomputing fun
The Cell
Broadband Engine (CBE, a.k.a. Cell microprocessor) has been jointly developed
by IBM, Sony and Toshiba. The Cell Broadband Architecture is intended to be
scalable through microprocessor integration of parallel vector processing,
where a general-purpose Power processor core is interconnected with eight
special-purpose streaming SIMD cores ("Synergistic Processing
Elements" or SPE). The first major commercial application of Cell is in
Sony's upcoming PlayStation 3 game console. Cell was developed to be a
general purpose processor and also a multimedia processor.
Exploiting
the CBE as a standard Power Architecture processor with tightly attached
(on-chip) SIMD acceleration capabilities opens up a non-disruptive migration
path to significantly improve the price/performance competitiveness (Teraflop
per €, Watts and/or floorspace) of Power blades and clusters. The
accelerator-based programming model has the additional attractiveness to
leverage the Linux on Power and VMX ecosystems while opening up opportunities
to push selected applications into the supercomputing realm.
Shifting
computational work from the main PPE processors to SPEs (by running special
subroutines that take over portions of the application) is a straighforward
extension of current VMX tuning techniques, enabling users to solve complex
problems in less time without having to deal with the complexity of
reconfigurable computing techniques like FPGAs. HPC segments that can
potentially benefit from the CBE capabilities include astrophysics,
bioinformatics, computational chemistry, medical imaging, seismic processing,
climate modelling, computational fluiddynamics, financial engineering,
quantum chromodynamics and other compute-intensive areas.
This
presentation will discuss the concepts of the new Cell Broadband Engine and
its impact on software development and performance. Operational enviroment
and programming models concepts, as well as first performance results for
different benchmarks and applications will be presented.
|
|
Toshiyuki Furui
NEC HPC Strategy and Products
NEC, as
one of the leading HPC solution and platform providers, has contributed to the
advances by offering broad range of HPC customers innovative, sustainable and
dependable solutions based on the platform products available from within the
company and, when deemed appropriate, from the market.
Mandatory
for the good product development is good computer architecture design and
good technology development from device up to system level.
This
talk will start from discussing NEC's strategy on High performance computing,
together with its wide range of HPC product line.
|
|
Joshua Harr
Does Multi-paradigm Computing Have a Future?
The
recent advent of multi-core processors and the lackluster improvement in
processor clock speeds gives reason to question future increases in cost
effective, single-thread application performance. Many HPC vendors are looking to some kind
of heterogeneous, or multi-paradigm, computing to obtain future
speedups. Most of these architectures
involve a general purpose CPU used to host the operating system and other
core system functions and some kind of a co-processor such as an FPGA or
floating-point accelerator to improve application performance. This talk will present the state of the art
in multi-paradigm computing and present some results for some CFD, seismic,
and life sciences applications. We’ll
also discuss recent developments in this area and the key advances that must
be achieved for this computing paradigm to succeed.
|
|
Bernard Schott
Proceedings on Grid Computing technology by
Platform Computing serving multiple HPC and Enterprise workload types based
on unified resource grid architecture.
Unified resource Grid architecture - Platform
Enterprise Grid Orchestrator - is the enabling element and differentiator to
other well-known Grid implementations today that are specialized to serve
selective workload types only, e.g. only sequential jobs or only parallel
batch workload, not supporting other workload types.
As specialization may be justified by the aim to
optimize the Grid for a certain class of applications, it is creating new silo-type
architectures, reducing the value of Grid technology to organizations and
companies.
Unified resource Grid architecture - Platform
Enterprise Grid Orchestrator - is delivering scalability and performance
for classical HPC computing concurrently in the same grid together with
SOA-real-time applications and typical business- or Enterprise-workloads.
This includes policy based automated orchestration of virtual machines and
application server architectures.
Operational stability – incident management - and
operations and management support technologies complement the unified
resource Grid solution scenario.
Further, an outlook on current and future
projects presents upcoming advancements in European Grid research.
|
|
Markus Fischer
Myricom, Inc., Arcadia,
CA, USA
Myri-10G and the Future of HPC Clusters
Myri-10G,
Myricom's 4th generation of high-speed networking products, is a convergence
at 10 Gbit/s data rates of Ethernet, the ubiquitous networking standard, and
Myrinet, by far the most successful "specialty network" for HPC
applications. In only a year, Myricom has made the transition from a
specialty-network company to a volume supplier of both 10-Gigabit Ethernet
and 10-Gigabit Myrinet products. In addition to the superb performance and
microbenchmarks for the 10-Gigabit Myrinet kernel-bypass mode of operation
for HPC applications, Myri-10G's seamless interoperability with 10-Gigabit
Ethernet is a great asset for grids, and an ideal transition path to
high-performance Ethernet clusters.
|
|
Richard Croucher
Scalable Building Blocks for Deploying Large
Grids
This
presentation will describe Sun's design approach for deploying Grids. Showing how very large Grids using
either Gig Ethernet or InfiniBand can be used. Why we are seeing a shift to fatter, and often diskless, compute
nodes. The new Sun servers which are ideally suited for Grids.He will also
decribe some of the large
configurations which Sun has deployed recently including the 38 TF TSUBAME
cluster recently commissioned at Tokyo
Institute of Technology.
|
|
Miron Livny
Managed Data
Placement in a Distributed Environment
|
|
Carl Kesselman
Virtual Communities and Science in the Large
Increasingly,
significant activities in science, business and society at large take place
within the context of distributed, computationally enabled collaborations.
Large-scale science collaborations such as those found in astrophysics,
astronomy, geophysics, and particle physics are typical of this new type of
collaboration. Driven by requirements for a broad range of skill sets,
participants and resources, the concept of community becomes central
organizing principal for these emerging computationally empowered
explorations. However, unlike traditional communities, which tend to have
well defined members and boundaries, today's scientific communities are
dynamic, distributed, and span institutional boundaries. This has lead to the
description of these structures as virtual organizations. Virtual
organizations more then just the people, but encompasses the services,
resources and capabilities that are shared to achieve the goals of the shared
endeavor. This leads to the inevitable question of how these distributed are
communities formed, how are they maintained, how to they create new services
and capabilities for their members, how do they get work done. Technologies
such as service oriented architectures and Grids provide underlying
foundation, but now need have mechanisms for identifying, creating and
operating distributed virtual communities. In this talk, I will explore the
question of how to create and empower virtual communities and how we can
support community formation within the context of our information technology
infrastructure.
|
|
Daniel A. Reed
Chancellor’s Eminent
Professor
University of North
Carolina at Chapel Hill
Director,
Renaissance Computing Institute
Scaling:
Realizing the Potential of Grids and High-end Computing
Large-scale
Grids containing thousands of sites are being contemplated, developed and
deployed. Similarly, node counts for terascale systems have grown to tens of
thousands, with petascale system likely to contain hundreds of thousands of
nodes. In addition, a tsunami of new experimental and computational data
poses equally vexing problems in analysis, transport, visualization and
collaboration. We must rethink traditional assumptions about software
scaling, component integration and hardware reliability. Our thesis is that
the “two worlds” of software – Grids and parallel systems – must meet,
embodying ideas from each, if we are to build usable and useful
infrastructure. This talk describes approaches to scalable performance
measurement, dynamic adaptation and Grid integration.
|
|
Pete Beckman
Urgent Computing: The Future of Emergency
Computation and Shared Grid Resources
High-performance
modeling and simulation are playing a driving role in decision making
and prediction. For time-critical emergency support applications such as
severe weather prediction, flood modeling, and influenza modeling, late
results can be useless.
With HPC
computing and distributed Grid resources becoming more widely available,
the community can begin to build global systems and policy frameworks
for utilizing high-end computational resources in support of emergency
computation.
A
specialized infrastructure is needed to provide computing resources quickly,
automatically, and reliably. SPRUCE is a system to support urgent or
event-driven computing on both traditional supercomputers and distributed
Grids. Scientists are provided with transferable "right-of-way"
tokens with varying urgency levels. During an emergency, a token has to
be activated at the SPRUCE portal, and jobs can then request urgent access.
Local policies dictate the response, which may include providing
"next-to-run" status or immediately preempting other jobs.
Additional components for this architecture include a periodic testing
mechanism of applications in "warm- standby" mode ensuring
readiness and an automated "advisor" that helps find the best
resource to submit based on deadline, queue status, site policy, and
warm-standby history.
|
|
Geoffrey Fox
Grids for Scholarly Research
We
discuss the support for modern research and in particular the emerging
distributed interdisciplinary data deluged scientific methodology as an end
(instrument, conjecture) to end (publication) process. This includes
integration of components at many different timescales (microseconds for MPI,
milliseconds for Web services and perhaps years for paper reviews). We need
to support both Web 2.0 and its mashups and social networks as well as the
traditional concerns of the Grid and HPC communities. We can speculate on
both the implications for research and the growth and sunset of many
organizations.
|
|
Domenico
Talia
Carmela Comito and Domenico
Talia
Data Integration in OGSA Grids for
Distributed Information Sharing
In a
distributed environment structured and semi-structured data stored in
databases can be heterogeneous in their formats, schemas, quality, access
mechanisms, ownership, access policies, and capabilities. We need models and
techniques for managing different data resources in an integrated way.
Therefore, data integration is a key issue for exploiting the
availability of large, heterogeneous, distributed and highly dynamic
information sources.
Data
integration in Grid systems has to deal with highly dynamic data volumes
provided by independent parties that happen to be participating at any given
time. So, traditional approaches to data integration, such as FDBMS and the
use of mediator/wrapper middleware, are not the best choice for supporting
semantic data integration and cooperation in dynamic computing environments
like Grids: a centralized structure for coordinating all the nodes may not be
efficient because it becomes a bottleneck when a large number of nodes are
involved and, most of all, it does not benefit from the dynamic and
distributed nature of data resources. As a consequence, decentralized
approaches, such as P2P models, can be adopted to overcome the limits of
current database systems and to address semantic heterogeneity of data
sources over a Grid. A P2P data integration system uses a decentralized
semantic network of mappings between source schemas to reformulate user
queries over every source in the network. According to this approach, in this
talk we present a framework for integrating heterogeneous, XML data sources
distributed over large-scale Grids. The proposed integration model is based
on XML schema MAPping (XMAP) in
terms of path-to-path mappings using the XPath language. The key issue of the
XMAP framework is the XMAP query
reformulation algorithm allowing to combine and query XML sources through a decentralized
point-to-point mediation process among the different data sources by using P2P
schema mappings. The above cited XML integration formalism is exposed as a
Grid Service within an OGSA-based Grid architecture that we refer to as Grid Data Integration System (GDIS).
GDIS is a service-based architecture for providing data integration in Grids
using a decentralized approach. The underlying model of such architecture is
discussed and we showed how it fits the XMAP formalism/algorithm.
|
|
Sergei Gorlatch
Sergei Gorlatch and Jens Müller
University of
Muenster, Germany
Scaling the Grid for Real-Time Online
Computer Games
We report
on our ongoing research in developing scalable, Grid-based solutions in the
novel area of real-time, interactive Internet applications, including online
computer games. While offering new opportunities for Grids, in particular, a
very broad user community and industrial interest, online games pose several
challenges in terms of Grid architecture and middleware, due to their high
dynamicity, interactivity and real-time requirements.
Conventional
architectural solutions -- client-server and p2p -- cannot provide the
required level of scalability for the very popular class of massively
multiplayer online games (MMOG), with hundreds and thousands of users playing
in a single world. We present an alternative, scalable proxy-server
architecture and describe its replication concept and consistency model. We
develop an analytical Game Scalability Model (GSM) for evaluating the
scalability of our proxy approach. The model provides a detailed forecast of
the required computation time and communication bandwidth for a multiplayer
application, thus allowing to compare different architectural and middleware
solutions.
We
illustrate the approach by the novel real-time strategy game Rokkatan
developed at the University of Muenster, which allows several hundreds of
users to participate in a fast-paced game session. The game is based on the
proxy-server architecture with an eventual consistency mechanism and uses the
GSM model to obtain forecasts for the main game characteristics with an error
of less than 15%.
|
|
Lucio Grandinetti
An Optimizated Algorithm for Local Grid
Scheduling
|
|
Ian Foster
(Joint
work with Ewa Deelman, Carl Kesselman, Mike Wilde, Yong Zhao, and others)
The GriPhyN Virtual Data System
The
Grid Physics Network project, GriPhyN, sponsored by the NSF, brought together
four large-scale experiments in physics and astrophysics (ATLAS, CMS, LIGO,
and SDSS) in partnership with computer scientists, to explore approaches to
harnessing the Grid for data intensive scientific collaborations. A
particular fccus of GriPhyN has been the development of data analysis and
management tools that facilitate discovery, collaboration, workflow
composition, and validation, and enable large-scale Grid resources to be
applied to data-intensive computations. This work has led to the development
of the GriPhyN Virtual Data System (VDS), a set of tools that enable the
specification and execution of complex data-intensive workflows, and the
subsequent management and discovery of both the resulting data products and
the workflows that created them. These tools include a virtual data language
for expressing in an abstract, machine-independent manner the graph of data
transformations that yield an output dataset from specified input data;
various planners, in particular the Pegasus system, for optimizing workflow
execution; and a virtual data catalog, for tracking the results of job
execution. In this talk, I provide an overview of VDS technology and review
some of its application successes.
|
|
Hai Jin
Image Processing Grid: A Concrete Example in
ChinaGrid
Image
Processing Grid, as one of the five typical grid application in ChinaGrid,
demonstrate its feasibility, efficient and practical through a set of
concrete grid applications. In this talk, I will introduce the overview of
Image Processing Grid, its platform and key techniques associated with this
platform, discuss the current status of three grid applications, that is
reconstruction of 3-D digital human being, remote sensing image processing,
and medial image processing. Besides, I will also discuss the grid
middleware, ChinaGrid Supporting Platform (CGSP), used in ChinaGrid, and how
it applies to Image Processing Grid to support various grid applications.
|
|
Charlie Catlett
Toward The InterGrid
For
several decades there have been distributed software systems, at one point
aimed at "Metacomputers," in the past ten years "Grids"
and more recently "service oriented architectures" based on web
services. Today most production grids
achieve interoperation by coordinating the deployment of a uniform set of
software and configurations - involving the installation of binary
distributions in some cases and in other cases involving coordination of the
build and install processes. The
emergence of standards and common approaches today presents the potential to
achieve interoperation both within and among grid facilities based on a
service description rather than a software and configuration (i.e.
implementation) specification. Such an
approach allows individual sites within a grid, or individual grid
facilities, to select or develop the best implementations, while also
providing a means for achieving interoperation. During 2006 the TeraGrid project in the
United States is making a transition from a software-implementation-based to
a service-based definition for interoperation of resources, involving a core
set of services (authorization, information, accounting, verification and
validation) combined with optional service "kits" (e.g. job
execution, data storage, application development, application hosting,
etc.). Using a similar approach,
TeraGrid is partnering with EGEE, NAREGI, the UK National Grid Service,
DEISA, and more than a dozen other grids in a project named "Grid
Interoperation Now" (GIN). GIN is
attempting to prototype the "InterGrid" based on currently
available methods for authorization, information services, job submission,
and data movement. We will discuss
this approach, including use cases and both near-term selected services and
long-term directions, and provide an update on TeraGrid plans and on the work
of the GIN collaboration.
|
|
Franck Cappello
Status and early results of the largest experimental
platform for Grid research: Grid'5000
Large
scale distributed systems like Grids are difficult to study only from
theoretical models and simulators. Most Grids deployed at large scale are
production platforms that are inappropriate research tools because of their
limited reconfiguration, control and monitoring capabilities. In this talk,
we present the status and early results of Grid'5000, a 5000 CPUs
nation-wide infrastructure for research in Grid computing.
Grid'5000
is designed to provide a scientific tool for computer scientists similar to
the large-scale instruments used by physicists, astronomers and biologists.
We describe the motivations, design, architecture of Grid'5000 and
performance results for the reconfiguration subsystem. We also present early
results obtained with this platform.
|
|
Frank Würthwein
The Open Science Grid (OSG)
The
OSG is a national computing infrastructure for high throughput computing in
the US that is open for all of science, and strives to be interoperable with
the LCG production infrastructure in Europe and Asia. We will describe status
and goals of the OSG from an applications development perspective. We will
address questions regarding the use of computing, storage, IO, as well as
scalability goals for the OSG middleware stack.
|
|
Ewa Deelman
USC Information Sciences Institute
Challenges of Managing Large-Scale Scientific Workflows in Distributed
Environments
Recently,
workflows have emerged as a paradigm for conducting large-scale scientific
analyses. The structure of a workflow specifies what analysis routines need
to be executed, the data flow amongst them, and relevant execution details.
These workflows often need to be executed in distributed environments, where
data sources may be available in different physical locations and the
processing steps may have different execution requirements. Workflows help
manage the coordinated execution of related tasks. They also provide a
systematic way to capture scientific methodology and provide provenance
information for their results. Scientists in many disciplines are approaching
data volumes and resource sharing facilities that would enable a new stage in
scientific discovery. Based on our experiences in generating and managing
workflows with thousands of tasks that execute over 1.8 CPU/years and process
approximately 10 TB of data, this talk will describe open research problems
in workflow management. We will explore the challenges from the perspective
of workflow creation, compilation and execution.
|
|
Dennis Gannon
Indiana University
Science Portals for Grid Applications
This
talk will describe the service-based architecture used by many scientific
communities to provide applications and workflow tools to users. We will
illustrate the ideas with the LEAD project and a service architecture that
has been developed to help atmospheric scientists conduct large-scale
simulation of extreme weather events such as tornadoes and hurricanes. The
user accesses the system by a web portal, which provides a secure environment
in which to host tools running on remote Grid resources such as the Teragrid.
The workflow composer that is accessed via the portal allows the user to
create simulations by composing useful data analysis and simulation
components into larger application templates which can be instantiated with
data that comes from remote instrument sources or data archives.
|
|
Andrew Grimshaw
Enterprise Data Integration with Grid
Technology
Both
compute grids and data grids have been used in academic computing. In the
extended enterprise (beyond a single site) the most compelling need is for
data grids to provide real-time, and near real-time, access to data
throughout the enterprise for both analytic purposes and to satisfy
operational needs. I begin my talk with a discussion of why compute
grids in the extended enterprise are not a priority for companies. I will
then examine the common data landscape in enterprises, and show how data grid
techniques can be used to provide data integration. I will contrast data
grids with alternatives such as data warehouses.
|
|
Luc Moreau
An open provenance
model for scientific workflows
|
|
David Abramson
Grid Middleware for e-Science software
deployment, testing and debugging
A critical
ingredient for success in e-Science is appropriate Grid-enabled software
which, to date, has lagged behind the high-performance computers, data
servers, instruments and networking infrastructure. All software follows a
life cycle, from development through execution, and back again. Grid software
is no exception, although there are sufficient differences in the details of
the various phases in the life cycle to make traditional tools and techniques
inappropriate. For example, traditional software development tools rarely
support the creation of virtual applications in which the components are
distributed across multiple machines. In the Grid, these types of virtual
applications are the norm. Traditional methods of distributing, testing and
debugging software do not scale to the size and heterogeneity of
infrastructure found in the Grid, and thus they are ineffective. In this talk
I will discuss some new middleware components that specifically target the
deployment, test and debug phases of the software life cycle.
|
|
Felix Heine
BabelPeers: Scalable and Semantically Rich
Grid Resource Discovery
|
|
Dieter Kranzlmuller
Authors: Dieter Kranzlmüller & Owen Appleton
The EGEE-II Project: Evolution Towards a
Permanent European Grid Initiative
EGEE-II
operates the world's largest multi-disiplinary Grid infrastructure. It is a
European Commission funded project bringing together more than 90 partners in
32 countries to produce a reliable and scalable computing resource available
to the European and global research community.
Bringing
experience from a wide range of earlier Grid projects and many of the world's
top Grid experts, EGEE-II is a dramatic step toward a sustainable European
Grid infrastructure that can provide a platform for research in the
twenty-first century.This presentation provides an overview of the work of
EGEE-II: its infrastructure, middleware, applications and support structures.
From this not only the achievements but the lessons learned relevant to other
projects and initiatives will be identified. Following this, the
relationships with other Grid projects, both large generalized initiatives
and more focussed efforts, will be examined. This will demonstrate the
increasing cooperation between projects, and the synergies and problems that
have emerged from such cooperation. Finally, from this experience the current
state of future plans will be explained. This will cover the emerging
federated model for sustainable future Grid infrastructures, and the current
proposals for centralized oversight of the emerging European Grid
infrastructure.
|
|
Wolfgang Gentzsch
D-Grid Coordinator
D-Grid
Globalization
demands global infrastructures, e.g. for transportation, communication, and
collaboration. Collaboration will be facilitated by grid infrastructures,
which are currently being built in hundreds of grid projects around the
world. The 5-year German D-Grid initiative aims at developing a common grid
infrastructure for "Services for Scientists", to be tested by
Community Grids (high-energy physics, astronomy, climate, medicine,
engineering, libraries). Recently, a 2nd Call invited new communities, IT
service providers and industry users to help expand the user community, and
to strengthen D-Grid towards a productive and sustainable service platform.
This lecture will present an overview on D-Grid, and its challenges and
benefits for the national and international science and industry community.
|
|
Mirco Mazzucato
National Grid initiatives in Italy
Started at the beginning of 2000,
with the INFN Grid project http://grid.infn.it/ the national production Grid.it
(http://grid-it.cnaf.infn.it/) infrastructure is now a reality. It includes
more than 40 resource sites, a Certification Authority, Registration
Authorities, release and central support of main general grid services as,
Computing and Storage Access, VO management and VO policy enforcing
framework, data management, information system, resource brokering and
scheduling. An operation and support center is operational 24x7 at CNAF
Bologna. The Italian grid is currently used by more than 20 Virtual
Organizations, is fully integrated with the European EGEE and the worldwide
LHC Computing Grid e-Infrastructure. It also provides a powerful grid demonstrator
to attract newcomers, see https://grid-demo.ct.infn.it/ . Application fields that are of maximum
interest for such high performance Grid platform include:
• Earth
Observation
• Geophysics
• Astronomy
• Biology
and Genomics
• Computational
chemistry
This talk will review the current
major middleware developments made by INFN and the progress towards a
national grid organization supporting the Italian Grid infrastructure (IGI)
and the construction of the Consortium for the Open Middleware Enabling Grid
Application (c-OMEGA) which aim at providing a general support for the
diffusion of the grid technology in Industry, Commerce and Government.
|
John O’Callaghan
Australian
Partnership for Advanced Computing
The
APAC National Grid
The
Australian Partnership for Advanced Computing (APAC) is a national partnership
of 8 organisations, one in each State as well as the Australian National
University (ANU) and CSIRO.
APAC and
its partners are developing Australia’s capabilities in advanced computing,
data management and grid infrastructure and are involved in research,
development, education, training and outreach activities.
The APAC
National Grid is being installed to allow researchers easy access to
distributed computation and data management facilities at the APAC National
Facility and the partner facilities and to services that support research
collaboration, nationally and internationally.
The
National Grid is being developed in collaboration with specific research
teams in astronomy, high-energy physics, bioinformatics, geosciences,
chemistry and earth systems science.
The core
grid middleware is the Virtual Data Toolkit (VDT), complemented with toolkits
for virtual organisation management, resource discovery, job scheduling and
job monitoring.
The APAC
Certificate Authority provides an authentication service for users of the
National Grid and has been recently recognised as a production level service
by the Asia-Pacific
Grid Policy Management Authority (APGridPMA).
The
presentation will focus on the design of the National Grid and its services
for distributed computing and data management. It will provide examples of the use of
these services by the applications.
|
|
Satoshi Sekiguchi
NAREGI as National Cyber Science
Infrastructure and beyond
The
Grid Middleware developed at NAREGI makes the various computational resources
(heterogeneous high-performance computers, high-end servers) appear as one
huge virtual computing resource to a user. This grid is based on the open
grid services architecture (OGSA), which is steadily becoming a global
standard. It serves as the core to the Cyber Science Infrastructure and is
the backbone of future advanced science research. In this talk, its current
status and future plan will be presented.
|
|
Guy Tel-Zur
Grid and High-Performance Computing in Israel
This presentation
is meant to give a broad overview of the present and future activities in
Grid and High-Performance Computing in Israel. There will be a distinction
between the Academia and the Industry.
In the
Academia the following topics will be described: The Inter University
Computation Center (IUCC) – which is responsible for common infrastructures
like Networking, the Israel Academic Grid (IAG) which coordinates and
represents all the academic institutes and universities that are involved
with EGEE. Then, several important Grid and HPC projects by individual groups
will be described with special attention to a few Condor implementations.
The
Industry sector is mainly organized under the umbrella of the Israel
Association of Grid Technologies (IGT). This young, non-profit, association
already has 30 members from leading High-Tech firms and some of its
achievements will also be described.
The
presentation will be concluded by describing future plans such as the
proposed IsraGrid initiative.
|
|
Jean Pierre Prost
Presenters: Jean-Pierre Prost, IBM WW Grid
Strategy and Technology Team, Montpellier, France
Rosa M. Badia, Grid Computing and Clusters Manager, Barcelona Supercomputing
Center, Spain
Latin American Grid Initiative Joint Research
Program Overview
The
LA(Latin-American) Grid is an IBM initiative, started in late 2005 and aimed
at dramatically increasing the quantity and quality of Hispanic Technical
Professionals entering the Information Technology fields. At the core of this
initiative is the development of a computer grid across multiple universities
and IBM to serve as a platform and a catalyst for education, advanced
research, collaboration and talent development, in the critical and emerging
fields of grid computing, distributed systems and supercomputing. Today, five
joint research projects are on-going, involving Florida International
University, University of Puerto Rico at Mayaguez, Barcelona Supercomputing
Center, University of Miami, Tecnologico de Monterrey, and IBM Research, in
the areas of grid technology (e.g. metascheduling, autonomic resource
management) and applications that can leverage the technology (e.g. hurricane
mitigation and bio-technologies).
This talk will provide a brief overview of the LA Grid initiative with more
focus on the joint research program. For each project, the goals and
directions will be presented as well as some of the underlying technologies
the projects are building upon. In conclusion, we will outline future plans
of the initiative, both in terms of collaborative research and in terms of
infrastructure buildout and talent development.
|
|
Zhiwei Xu
Research Agenda
for China National Grid and Vega Grid
This talk will present two grid research projects
in China, a multi-group China National Grid project and the Vega Grid project
at Institute of Computing Technology, Chinese Academy of Sciences. It will
give an overview of the advances of the two projects in 2002-2005, and
discuss their research agenda for 2006-2010. The talk will emphasize
essential issues and international cooperation.
|
|
Kihyeon Cho
Cyberinfrastructure
in Korea: Status Report
Overview
of cyberinfrastructure in Korea, which consists of computing resource,
network, middleware, and application, will be presented. In particular,
emphasis will be given on the issues and strategies on the integration of the
components for production level national cyberinfrastructure. Applications
which take full advantage of Korea’s advanced broadband infrastructure will
also be discussed.
|
|
Kyriakos Baxevanidis
European Commission 7th FP Preview
|
|
Sanzio Bassini
DEISA Project -
European Supercluster Alliance
|
|
Thomas Lippert
Blue Gene - A Paradigm Shift in HPC?
Presently,
we witness a rapid transition of cutting edge supercomputing towards highest
scalability, utilizing up to hundreds of thousands of processors for
cost-effective capability computing. As so often experienced in the past, a
great leap forward in performance will result in new emerging phenomena and
insights for many fields of computational science and engineering. A
considerable number of "highly scalable" problems already benefits
from this development and more and more simulation science will enter this
class, with ongoing advances of scalable algorithms and software. Still, many
HPC-problems, maybe the majority, are less scalable by nature, might require
intricate communication capabilities or a large non-distributed memory space
with fast memory access. General purpose supercomputers like clusters of SMPs
or NUMA-clusters are needed for the latter class. It has become a hot debate
within the community if the advent of highly scalable systems like Blue Gene
and the emergence of a more heterogeneous hardware landscape signal the onset
of a paradigm shift in HPC.
Staying
abreast of changes, the German national supercomputing centre "John von
Neumann Institute for Computing" (NIC), the supercomputer centre of the
Helmholtz Association, has recently complemented its 9 Teraflop/s general
purpose SMP-cluster by a 46 Teraflop/s Blue Gene/L supercomputer, currently
being one of the fastest computers in Europe. Both systems share a huge
global parallel file system which is part of the file system of the European
DEISA alliance. With this configuration, the NIC is able both to meet the
requirements of a very broad range of projects and to support a selected
number of high-end simulation projects, aiming at leadership in science and
engineering.
After a
discussion of these developments, I will present a few examples demonstrating
the added value through a dual hardware approach. I will describe NIC's
concept of simulation laboratories to accommodate and support a diversity of
scientific disciplines. I will sketch our mid-term and long-term plans and
the future role of the NIC as part of the European e-Science ecosystem.
|
|
Antonio Arena
Grid Portals: Experiences and Perspectives
|
|
Rizos Sakellariou
University of
Manchester, UK
Parallel Job Scheduling for HPC on the Grid:
The Case for Novel Forms of Scheduling
Current
parallel systems typically provide a limited level of service to their users;
essentially this can be summarized to "run a job whenever it gets to the
head of the job queue". New patterns of usage (arising from Grid
computing) have resulted in the introduction of advance reservation, where
jobs can be made to run at a precise time. However, this is again an
excessive level of service, which has several disadvantages for the resource
owner. In recent work (http://www.gridscheduling.org), we have focused on the
benefits resulting from the usage of Service Level Agreements in job
scheduling for Grid-enabled resources.
This
talk will highlight the problems arising in parallel job scheduling on the
Grid, will discuss the key issues related to SLA based job scheduling and
will present results as well as challenges to be addressed.
|
|
Claudio Arlandini
Present and Future of Project Sepac
Project
SEPAC (South European Partnership for Advanced Computing) is a Grid Project involving
CILEA, ETH Zurich, HP, SPACI and University of Zurich. Its focus is in
testing Grid technologies in a large supercomputing centers environment.
Great
importance is attributed to usability by actual customers, industrial or
academics. In this talk the main achievements so far and the future prospects
will be illustrated.
|
|
Giovanni
Aloisio
G. Aloisio, M.
Cafaro, I. Epicoco, S. Fiore
University of Lecce & SPACI Consortium
An integrated approach to Grid Resource
Management
Grid
computing is a rapidly developing and changing field, involving the shared
and coordinated use of dynamic, multi-institutional resources. Grid resource
management is the process of identifying requirements, matching resources to
applications, allocating those resources, and scheduling and monitoring Grid
resources over time in order to run Grid applications as efficiently as
possible. While Grids are almost commonplace, the use of good Grid resource
management tools is far from ubiquitous because of the many open issues of
the field, including the multiple layers of schedulers, the lack of control
over resources, the fact that resources are shared, and that users and
administrators have conflicting performance goals. We present our experience
in building, administering and using the SPACI and SEPAC grids through the
Grid Resource Broker portal, a web gateway to Globus Toolkit based grid
environments. The GRB Portal main goal is to mediate between the user's
request and the Grid offering, thus hiding the complexity of the underlying
middleware. The GRB acts as a mediator component, with the aim of bridging
the gap between the scientist and the Grid. GRB leverages grid technologies
developed at the University of Lecce, Italy, on top of the Globus Toolkit:
many grid libraries providing high-level grid services, a GSI plug-in for the
gSOAP Toolkit, the iGrid Information Service. The GRB also exploits the GRelC
Data Access libraries to manage through the web portal Grid Databases (access
to relational databases, flat files and so on), submit SQL queries and
retrieve information about stored relational data sources. Moreover we have
been integrating the GRelC Data Storage libraries to manage workspaces and
collections of files, perform (reliable) data transfer, access file using
Posix-like APIs, and so on.
|
|
Marcelo Pasin
An Efficient Interoperable High Performance
Programming Architecture for the Grid
|
|
Christopher Huggins
DAS -The World's Fastest Grid
The
DAS-3 grid will consists of 5 clusters provided by ClusterVision and hosted
by four universities in the Netherlands. The grid will have a total
theoretical peak performance of more than 3.8TFLOP. Each cluster has a
Myri-10G network from Myricom, which connects over 8 10GigE links to the new
Dutch academic SURFnet-6 network. The DAS-3 grid will be a key research tool
in the context of the Dutch Science Foundation (NWO) funded StarPlane project
led by the University of Amsterdam and the Vrije Universiteit. The StarPlane
project aims to empower grid applications to use the most advanced features
of the dynamically reconfigurable multi-colour optical backbone of the Dutch
National Research and Education Network SURFnet-6. Using photonic devices
such as wavelength selective switches and micro electro mechanical switch
devices the topology can be adapted to dynamically optimise the network
capacity to the grid application needs.
|
|
Umit Catalyurek
Application of Grid Technology in Cancer
Research: Middleware and Tools
Information
technology is playing an increasingly important role in cancer research.
Basic and clinical research to understand and cure this complex disease
requires synthesis of information from a range of data sources. Moreover,
cooperative studies involving researchers from multiple institutions have
great potential to significantly improve and add to the understanding of
cancer. There is a need for advanced information technologies to support the
discovery and integration of information, the sharing of data and analysis
tools, the management and analysis of large datasets, and the execution of
multi-institutional, coordinated studies. We will describe and discuss the
application of Grid infrastructures and middleware tools for cancer research.
We will present an overview of our Grid middleware systems (STORM, Mobius,
DataCutter) as well as the NCI-funded cancer Biomedical Informatics Grid
(caBIG) effort and its Grid infrastructure, called caGrid. We will also
present several application projects that employ these middleware tools to
support distributed data management and integration for design of experiments
and comparative studies, and large scale, distributed image analysis.
|
|
Barbara Chapman
Support for Specification and Scheduling of
an Air Quality Forecast Code on the Grid
Large-scale
applications are increasingly composed of a number of different components,
which may interact in complex ways at run time. Such applications may consume
substantial resources and are often loosely coupled; hence they are obvious
candidates for grid computing.
At the
University of Houston, we have been working to develop and deploy a
state-of-the-art Air Quality Forecast application in order to reliably
predict atmospheric pollution in our region.
This application has a complex workflow and non-trivial amounts of
data must be transferred between different components during execution. In our search for gridware to support its
automated scheduling and execution, we did not find a system that permitted
its specification, scheduling and job launch as desired. To address this gap in functionality, we
have created the dataflow language GAMDL to describe a workflow application's
logic.
In this
presentation, we describe our application, its requirements, and the features
of GAMDL as well as the metascheduling architecture being implemented to
complete our support for the production of air quality
forecasts
for our local and state officials.
|
|
Stephen M. Pickles
Collaboration grids: costs, benefits, and
scaling laws
"Group-forming
networks have the property that the benefit of oining the network grows
exponentially with the size of the network.
In
this talk, I make the conjecture that collaboration grids are examples of
group-forming networks. The argument proceeds by ounting the number of
collaborations that a common grid infrastructure allows members to form with
each other. I go on to consider some of the costs of joining a grid from the
different perspectives of resource providers nd consumers, and examine
how these costs scale. I illustrate the provider perspective with
examples drawn from the operation of the UK National Grid ervice, and the
consumer Prspective with examples drawn from the experience of
the UK e-Science project RealityGrid. If my conjecture is true, then
combining two grids into a larger one yields a grid whose alue is
greater than the sum of its parts. In this light, I describe the
challenges faced by recent Grid Interoperation Now (GIN) initiative,
re-examine the concept of Virtual Organisations and reflect on the need
for standards."
|
|
Anatoly Sachenko
Deployment of a GRID System for Space Weather
Prediction using Neural Networks
|
Main Aim