“Use cases” Stephen Pickles e-Frameworks meets e-Science workshop Edinburgh, 27/4/2006

advertisement
“Use cases”
Stephen Pickles
e-Frameworks meets e-Science workshop
Edinburgh, 27/4/2006
Combining the strengths of UMIST and
The Victoria University of Manchester
NGS & Partners Today
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
2
NGS
Points to notice
 NGS provides compute and data resources to UK users and user
communities
– based on PKI (mostly GSI)
 Care about making NGS relevant to wider community
– e.g. provisioning of domain-specific portals using NGS resources
– need to bridge AA universes
• this is the ShibGrid problem
 Care about making wider (e.g. JISC-funded) developments accessible
in a Grid context
– e.g. access to centrally curated data in Grid workflows
– need to bridge AA universes in opposite direction
• this is the GridShib problem
– need “services” to be “Web services” (not just capabilities), and secure
• Web browser context is insufficient
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
3
SHEBANGS
 Shibboleth-Enabled Bridge to Access the NGS
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
4
ConvertGrid
2001
Census
Experian
Postcode
ONS
Convert
Tables
Wards
Query Web
Service
ConvertGrid Web
Service
X509
X509
Postcode
Sectors
Athens
authentication
Client 1
Website
Client 2
Command Line Client
CommonGIS
Applet
Services to automate extraction, geography conversion
and fusion of aggregate, geography-based datasets.
Powered by OGSA-DAI on NGS
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
5
GEMEDA
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
6
GEMS
 The purpose of the Grid Enabling MIMAS Services (GEMS) project is
to provide Grid enabled access to the aggregate statistics from the
2001 Census via OGSA-DAI on the National Grid Service (NGS).
 This will involve connecting the SQLServer databases holding the
2001 Census data directly to the Grid via the NGS and also Grid
enabling the current data access system (Casweb).
 The advantage of this approach is that it will maximise and build upon
the ESRC/JISC investment in the establishment of existing data
infrastructure, avoid having to maintain multiple database systems to
support different forms of access and also provide a more rapid and
flexible method of building Data Grids via the NGS.
 The project will be working closely with two of the National Centre for
e-Social Science (NCeSS) nodes who have a requirement for grid
enabled access to the aggregate statistics from the 2001 Census and
who will be early adopters of the service.
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
7
MIMAS & data provision
Points to note:
 users must register to access many relevant datasets
– e.g. census, satellite data, ...
 wealth of access control information in things like Athens,
Shibboleth
– not readily accessible from Grid context
• this is the GridShib problem
 access enforced by data portals
– must shift to service layer to make grid-accessible
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
8
Nektar, SPICE & Vortronics (2005)



Funding from NSF (US) and EPSRC (UK) for projects linking TeraGrid and NGS
Inspired by TeraGyroid Experiment (2003)
All involve parallel codes and need high-bandwidth cross-site communication.
TeraGyroid at SC Global 2003
TRICEPS
Winner SC’03
HPC Challenge
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
9
Steering in an OGSA framework
Steering
WS
bind
Steering library
publish
Client
find
connect
Steering
Library
client
Simulation
Registry
publish
data
transfer
Steering library
bind
Steering
WS
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
Visualization
Visualization
10
RealityGrid
Points to notice:
 requires simultaneous access to multiple compute and
visualisation resources and sometimes Access Grid
nodes, at times to suit the humans
– and ideally network
– advance reservation + co-allocation = co-reservation
– significant pain point
 provenance
– provenance (for reproducibility) not just a question of capturing
inputs and outputs; also need steering commands
 monitoring of running application/simulation code is
different to monitoring of job
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
11
The Nano-CMOS grid
NeSC
Glasgow
Edinburgh
HPC(x
)
Manchester
York
CSA
R
Southampton
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
12
NanoCMOS Grid
Points to notice:
 will have significant cottage industry in turning design
tools, design flows into services and workflows
 secure invocation on multiple Grids matters
 provenance
 markup/metadata/discovery
 security of data (sensitive IP)
– don’t trust sysadmins
Combining the strengths of UMIST and
The Victoria University of Manchester
28 April 2006
13
Download