Virtual Laboratory for e-Science (VL-e) Henri Bal Department of Computer Science

advertisement
Virtual Laboratory for
e-Science (VL-e)
Henri Bal
Department of Computer Science
Vrije Universiteit Amsterdam
bal@cs.vu.nl
vrije Universiteit
Outline
•
•
•
•
e-Science and virtual laboratories
The VL-e project
VL-e and networking
Case studies:
o Visualization
o Interactive problem solving environments
o Distributed supercomputing
• Computing/networking infrastructure
e-Science
• Web is about exchanging information
• Grid is about sharing resources
o Computers, data bases, instruments, services
• e-Science supports experimental science by
providing a virtual laboratory on top of Grids
Virtual Laboratories
Distributed
computing
Application
Specific
Part
Potential Generic
Visualization & part
Management
collaboration of comm. &
computing
Knowledge
Data &
information
Application
Specific
Part
Potential Generic
part
Virtual
Laboratory
Management
of comm.
& services
Application
oriented
computing
Application
Specific
Part
Potential Generic
part
Management
of comm. &
computing
Grid
Harness multi-domain distributed resources
Virtual Laboratory for e-Science
Interactive
PSE
Adaptive
information
disclosure
High-performance
distributed computing
User Interfaces &
Virtual reality
based visualization
Security & Generic
AAA
Collaborative
information
Management
Virtual lab. &
System integration
Optical Networking
The VL-e project
• 40 M€ (20 M€ BSIK funding)
• 2004 - 2008
vrije Universiteit
• 20 partners
• Academic - Industrial
VL-e and networking
• e-Science applications generate much (distributed) data
o High-resolution imaging
o Bio-informatics queries
o Particle physics:
o Currently: 1 PByte per year
o LHC (2007): 10-30 PByte per year
• Virtual laboratories need high-speed networks for
o Remote visualization
o Interactive problem solving environments
o Distributed supercomputing
VL-e and networking
i PSE
A.I.D.
High-performance
distributed computing
Visualization
Security
CIM
Virtual lab
Optical Networking
Visualization on the Grid
Visualization on the Grid
Visualization on the Grid
Visualization on the Grid
Visualization on the Grid
Interactive Problem Solving Environments
From Medical Image Acquisition to Interactive Virtual
Visualization…
MRI, PET
Patient at MRI
scanner
MR image
MR image
Segmentation
Cave, Wall, PC,
PDA
Virtual Node
navigation
Bypass creation
Simulated
blood flow
Shear stress, velocities
ce (e.g., Valencia)
se (e.g., Leiden)
MD login and Grid
Proxy creation
Monolith, Cluster
LB mesh
generation
Job submission
ce (e.g., Bratislava)
Job
monitoring
ui (VRE)
P.M.A. Sloot, A.G. Hoekstra, R.G. Belleman, A. Tirado-Ramos, E.V. Zudilova, D.P. Shamonin, R.M. Shulakov, A.M. Artoli
, L. Abrahamyan
Simulated
Blood Flow
Distributed supercomputing
(parallel computing on grids)
DAS-2
VU (72 nodes)
UvA (32)
GigaPort
Leiden (32)
Delft (32)
Utrecht (32)
Distributed ASCI Supercomputer 2
Distributed supercomputing
(parallel computing on grids)
HPC on a grid?
• Can grids be used for High-Performance Computing
applications that are not trivially parallel?
• Key: grids usually are hierarchical
o Collections of clusters, supercomputers
o Fast local links, slow wide-area links
• Can optimize algorithms to exploit this hierarchy
o Message combining + latency hiding on wide-area links
o Optimized collective communication operations (broadcast etc.)
o Often gives latency-insensitive, throughput-bound algorithms
Ibis: a Java-centric grid
programming environment
• Written in pure Java, runs on heterogeneous grids
o “Write once, run everywhere ”
• Many applications:
o Electromagnetic simulation (Jem3D)
o Automated protein identification
(VL-e application from AMOLF)
o N-body simulations
o SAT-solver
o Raytracer
Jem3D (see SC’04)
Available from www.cs.vu.nl/ibis
Networking demands
• Low latency is needed for
o Interactive visualization
o Interactive Problem Solving Environments
o Synchronous, latency-sensitive parallel algorithms
• High throughput is needed for
o Data-intensive e-Science applications
o Visualization of large data sets
o Asynchronous, throughput-bound parallel algorithms
• Efficient collective (group) communication for
o Collaborative visualization between multiple sites
o Collective operations in parallel algorithms
Outline
•
•
•
•
e-Science and virtual laboratories
The VL-e project
VL-e and networking
Examples:
o Visualization
o Interactive Problem Solving Environments
o Distributed supercomputing
• Computing/networking infrastructure
VL-e environments
Application
specific
service
Application
Potential
Generic service
&
Virtual
Lab. services
Grid
&
Network
Services
Telescience
Medical
Application
Bio ASP
Virtual Laboratory
Virtual Lab.
rapid prototyping
(interactive simulation)
Grid Middleware
Additional
Grid Services
(OGSA services)
Gigaport
Network Service
(lambda networking)
VL-E Proof of concept
Environment
VL-E Experimental
Environment
DAS-3
• Proposed next generation grid in the Netherlands
• Partners:
o ASCI research school (VU, UvA, TU Delft, Leiden)
o Gigaport-NG/SURFnet: DWDM computer backplane
(dedicated optical group of 8 lambdas)
o VL-e and MultimediaN BSIK projects
• Topology controlled by applications through the
Network Operations Center
DAS-3
NOC
Summary
• VL-e (Virtual Laboratory for e-Science)
studies entire e-Science chain, including
applications, middleware and grids
• High networking demands from applications
and generic methods
• New state-of-the-art Grid infrastructure
planned for 2006 using optical networking
Download