Nessun titolo diapositiva

1
Track 2 Metacomputing
Recent Developments in Metacomputing The Grid


IBM Training Center, La Hulpe, Brussels 29-30
November 2000
Domenico Laforenza CNUCE-Institute of the Italian
National Research Council CNR Research Area, via
V. Alfieri,1 - 56010 Ghezzano, Pisa,
Italy and Department of Computer Science,
University of Pisa Phone [phone] - Fax
39-050-313.8091 domenico.laforenza_at_cnuce.cnr.it
http//brunello.cnuce.cnr.it/domenico/domenico.ht
ml
2
Outline

• Supercomputing - Is it really important ?
• Computing Platforms Evolution Metacomputing
• Metaproblems - Multidisciplinary Applications
• How to Build a Meta Application?
• Future Trends Conclusions

3
Supercomputing - Is it really important ?

• Access to Supercomputing Resources
• The needs of computational researchers far exceed
  the Supercomputing resources available now or in
  the next future.

• Community of users for Supercomputing far exceeds
  the available resources.

• Budget problems for HPC acquisition and
  maintenance, ..

4
How to satisfy these computational needs ?
More than 9,000 Pentium Pro processors and about
580 Gbytes of system memory housed in 86
standard, 19 cabinets, the ASCI TFLOPS system
will occupy approximately 1600 sq. ft. and draw
up to 800 KW of power.
5
How to satisfy these computational needs ?
1997 - ASCI RED - 1.8 Tflop/s - INTEL
Oct. 1998 - ASCI BLUE/Pacific - 3.9 Tflop/s - IBM
Nov. 1998 - ASCI BLUE/Mountain - 3.072 Tflop/s -
SGI
6
(No Transcript)
7
200 Million, 30TeraOPS Project to be at Los
Alamos National Laboratory

• HOUSTON, August 22, 2000 - In a move that
  reinforces Compaq Computer Corporation's (NYSE
  CPQ) emergence as the leading supplier of high
  performance computing systems, the U.S.
  Department of Energy's (DOE) National Nuclear
  Security Administration (NNSA) announced today
  that it has selected Compaq to build the world's
  fastest and most powerful supercomputer, a 30
  TeraOPS system code-named "Q."

8

• The "Direction des applications militaires (DAM)"
  from the French atomic agency CEA will install a
  Compaq/QSW machine with over 2500 processors.
• The new supercomputer, that will be used for
  numerical simulations of nuclear arms.
• The machine will be enlarged in three phases.
• 5 TFlop/s machine (operational by the end of
  2001).
• Phase 2 a 30-50 TFlop/s machine will be
  installed.
• By 2009, in phase three, a system will be
  installed in the 100 TFlop/s range.

9
INFORMATION TECHNOLOGY FOR THE TWENTY-FIRST
CENTURYA BOLD INVESTMENT IN AMERICAS
FUTUREJanuary 24, 1999, Working Draft
Clearly, many researchers and computer users will
not have local access to all the computing power
and storage they need
10
Metacomputing

• Different resources (computing, instruments, .)

• geographically distributed

used as a single powerful parallel machine.
11
Computing Platforms Evolution
Breaking Administrative Barriers
12
From Metacomputing to Grid Computing
13
What kind of problems does Metacomputing solve
well ?Classes of Applications

• Sequential - Dusty Deck Codes

• Data Parallel
• Synchronous - Tightly Coupled
• Data Elements are essentially identical (adapt
  for SIMD / MIMD)
• Loosely Synchronous
• Data Elements are not identical (adapt for MIMD)

• Asynchronous - Functional Parallelism
• Irregular in time and space
• Hard to parallelize to exploit the massive
  parallism

• Embarrassingly Parallel
• NOW, SIMD, MIMD

14
MetaproblemsMultidisciplinary Applications

• Class of problems which is outside the
  scope/capabilities of a single computer
  architecture
• Best run on a Metacomputer (Grid)
• Consists of an asynchronous collection of loosely
  synchronous components
• Each component can be parallel

15
From Monolithic Applications to Meta Applications
Partitioning into Components
Communication
Monolithic Application
Berlin
M1
Pisa
Paris
16
How to build a meta application?Application
integration
Machine X
Machine Y
WRAPPER
WRAPPER
Application A
Application B
New Application (A B)
17
Commodity Technologies

• COTS (Commercial Off-The-Shelf) Software

• Open interfaces that enable large application
  components to be quickly integrated in new
  applications

• Key Technologies

• HTML, XML, VMRL, HTTP,

• MIME, IIOP, CGI,

• Java, JavaScript, JavaBeans, ActiveX, .

• CORBA, DCOM, JINI,.

18
Multidisciplinary Applications

• Example the design and manufacture of a modern
  aircraft, which presents problems in

• Geometry Grid Generation

• Fluid Flow

• Acoustics

• Structural Analysis

• Operational Research

• Visualization

• Database Management

Research problems are becoming more complex and
interdisciplinary in nature
19
Grid (Metacomputing) Applications

• Distributed Supercomputing
• Stellar Dynamics, Ab initio chemistry, ...
• High Throughput
• Chip design, Parametric studies,
• On Demand
• Medical instrumentation, network-enabled solvers,
  
• Data Intensive
• Sky survey, Physics data, Data Mining,
• Collaborative
• Collaborative design, data exploration,
  education, ...

20
Metacomputing in IndustryApplication-Centered
Metacomputing - Funded by European Union

• MICA (Paderborn Center for Parallel Computing,
  Germany et al.)
• Virtual CFD server that runs on several HPC
  Systems throughout Europe
• PHASE (Paderborn Center for Parallel Computing,
  Germany et al.)
• Allows pharmaceutical companies to access HPC
  Systems for rational drug design projects
• EUROPPA (GENIAS BeNeLux i.c. et al.)
• Production of multimedia products (movies,
  advertising, etc.) in cooperative way
• PROMENVIR (Parallel Applications Centre, Univ.
  Southampton et al.)
• To provide a meta-application software package
  for probabilistic analysis through simulation of
  mechanical systems
• TOOLSHED (Parallel Applications Centre, Univ.
  Southampton et al.)
• To create a STEP-based parallel simulation
  environment for commercial mesh-based analysis
  codes

21
Metacomputing in IndustryApplication-Centered
Metacomputing - Funded by European Union

• METODIS (MEtacomputing TOols for DIstributed
  Systems)
• (University of Stuttgart RUS, CRIHAN,
  AEROSPATIALE, DASA and, PALLAS)
• is a project which aim is to develop tools to
  Metacomputing at European level.
• The tools should make the use of distributed
  resources for intensive numerical simulation
  between remote sites easier.

22
The DataGrid Projecthttp//grid.web.cern.ch/grid/

• is a project funded (October 2000) by the EU
• The project has six main partners
• CERN
• CNRS, France
• ESRIN (ESA), Frascati, Rome, Italy
• INFN, Italy
• NIKHEF, The Netherlands
• PPARC, UK
• and fifteen associated partners

23
The DataGrid Project Objective of the project

• To enable next generation scientific exploration
  which requires intensive computation and analysis
  of shared large-scale databases, from hundreds of
  TeraBytes to PetaBytes, across widely distributed
  scientific communities.
• Such sharing is made complicated by
• the distributed nature of the resources to be
  used
• the distributed nature of the communities
• the size of the databases
• and the limited network bandwidth available

24
Molecular Science Software
http//www.emsl.pnl.gov2080/docs/ecce/

• Extensible Computational Chemistry Environment
  (Ecce)
• a comprehensive, integrated, problem-solving
  environment focused on computational chemistry
• consists of a suite of distributed client/server
  applications enables research scientists to
  easily use computational software to perform
• complex molecular modeling and analysis tasks
• by accessing networked, high-performance
  computers from their desktop workstations.

25
Molecular Science Software
http//www.emsl.pnl.gov2080/docs/ecce/

• Developed as part of the construction of the
  Environmental Molecular Sciences Laboratory
  (EMSL)
• Envisioned to be used as an integrated component
  in solving DOEs grand challenge environmental
  restoration problems which efficiently use
  available computational resources
• Designed and developed to be a highly efficient
  and portable MPP computational chemistry package
  with associated problem solving environment
• Provides computational chemistry solutions which
  are scalable with respect to chemical system size
  as well as MPP hardware size
• Extensible framework supporting development and
  use of new computational methods

26
Molecular Science Software
http//www.emsl.pnl.gov2080/docs/ecce/

• Ecce currently supports
• remote submission of calculations to UNIX
  workstations, and supercomputers running
  LoadLevelerª, Maui Scheduler, and NQEª.
• Remote communications are accomplished using
  either remote shell (rsh), secure shell (ssh), or
  Globus.
• Globus was developed by Argonne National
  Laboratory and uses private key encryption
  technology to run computations on remote systems.

27
Molecular Science Software
http//www.emsl.pnl.gov2080/docs/ecce/

• Research and development efforts focus on
  developing
• a three tiered architecture to support the
  enhancement of the user interface to a Java
  environment,
• integration with collaborative technologies such
  as an electronic laboratory notebook,
• and a CORBA communication layer to various
  database servers.
• To provide a
• problem-solving environment that addresses the
  needs of computational scientists
• working in a distributed environment with a
  variety of data sources.

28
Metacomputing (Grid Computing)
IF YES, What will be the dominant grid approach
in the next future ??
29
The Computational Grid is analogous to
Electricity (Power) Grid and the vision is to
offer a (almost) dependable, consistent,
pervasive, and inexpensive access to high-end
resources irrespective their location of physical
existence and the location of access.
30
Trends
It is very difficult to predict the future and
this is particular true in a field such as
Information Technology
I think there is a world market for about five
computers. Thomas J. Watson Sr., IBM Founder,
1943
31
Conclusions

• Research problems are becoming more complex and
  interdisciplinary in nature

• Research will be conducted in virtual
  laboratories

• Scientists and engineers will be able to
  perform their work without regard to physical
  location

• interacting with colleagues

• accessing instrumentation, sharing data and
  computational resources

• accessing information in digital libraries

32
Conclusions
GRIDsThe time is very exciting but .
EU should gain momentum or risk to lag behind..
33
(No Transcript)