Something different ?

1
Something different ?

• Pierre Kuonen
• EIA-FR

2
Why CoreGRID ?

• Most current Grid projects are goal driven
• There is strong pressure to develop a working
  environment from the Grid promoters, end-users
  and institutions that provide funding
• There is a lot of engineering activities but
  little research in computer science
• Lack of visibility of European contributions to
  the development of the Grid Foundations
• Lack of critical mass (research activities are
  widely dispersed and uncoordinated)
• What are the consequences ?
• Lack of long term research to shape the Next
  Generation Grids
• A research community that is not yet well
  identified and sustainable
• Serious possibility of repeating history (GRID
  next buzzword after HPCN, AI)

3
CoreGRID objectives

• To build a European-wide research laboratory
• Create the European Grid Lighthouse and be seen
  as such worldwide
• To achieve integration and sustainability
• To build solid foundations for GRID and
  Peer-to-Peer (P2P) technologies
• Both on a methodological basis and a
  technological basis.
• Support medium and long term research activities
• Achieve and promote scientific and technological
  excellence
• Gather and disseminate European research
• A think-tank for spin-off projects
• EC funded, bilateral projects,

4
CoreGRID membership

• 42 partners
• 18 Countries (1 from S. America)
• 118 Researchers
• 199 PhD Students

VTT
IC RAL-CCLRC U. Belfast U. Cambridge U.
Cardiff U. Manchester U. Newcastle U. Wesminster
KTH SICS
FhG FZ-Julich HLRS U. Munster U. Dortmund U.
Passau ZIB
Delft U. Vrije U.
PSC CYFRONET
U. Chile
CETIC UCL
Masaryk U.
INRIA CNRSERCIM
EPFL EIA-FR
SZTAKI
CLPP-BAS
U. Coimbra
CNR-ISTI INFN U. Calabria U. Lecce U. Pisa
UPC
ICS-FORTH
U. Cyprus
5
Main issues

• Exploiting huge but elusive resources
• resource discovery, reservation/allocation,
  scheduling, fault tolerance, load balancing
• Handling massive, uncertain and unreliable state
  information
• distributed and decentralised management and DB,
  trust and security
• Going beyond the conventional programming model
• multiple and flexible parallelism, at run time
  adaptation

6
Our approach
7
What scientists need

• From program centric vision toservice
  centric vision
• Program-centric
• Execute my program on my specified resources
• Service-centric
• Provide me some services with a given QoS, no
  matter where they are located and executed

8
A set of well identified research challenges

• Knowledge Data Management
• Handling information/data that are
  required/produced by a wide range of diverse
  processing power
• Programming Model
• Making the programming of Grid infrastructures as
  simple and transparent as possible
• Resource Management
• Handling efficiently a large range and quantity
  of heterogeneous resources
• Scheduling
• Scheduling jobs/applications/tasks/computation
  within a Grid environment
• System Architecture
• Designing the next generation Grid middleware
• Problem Solving Environments, tools and GRID
  systems
• Integrating various middleware, tools and
  applications for problem solving

9
How to share resources

• We do not built HPC machines
• We use existing ones
• Who agree to give us cycles
• To whom do we agree to give cycles
• How can we take advantage of your cycles
• Resource discovery and management
• Which applications can take advantage of your
  cycles
• Mapping applications to machines

10
What CoreGRID can possibly do for you??

• Is this approach of any interest for you ??
• Resource management is a major issue in huge
  systems
• Mapping applications to machines is an important
  issue
• How to represent an application together with its
  requested QoS
• To go beyond MPI programming paradigm
• Standard..