UK Grid Support Centre
CLRC + EPCC + MC
Rob Allan
CLRC e-Science Centre
Daresbury Laboratory
http://www.grid-support.ac.uk
http://www.e-science.clrc.ac.uk
r.j.allan@dl.ac.uk
22nd October 2001
UK Grid Support Centre - GSC
Outline
• The issues
– why we need a Grid Support Centre
• The team
– who we are
• The service
– what we will provide
22nd October 2001
UK Grid Support Centre - GSC
HPC and Scientific Opportunity
in the Electronic Era
Report of the Trends and Opportunities Panel (P.J. Durham, March 2001)
•
From Molecules to Matter – Electronic Structure
–
–
–
–
–
The Structure of Proteins
“Simulation of whole systems, and not
Materials Design
just system components”
Heterogeneous Catalysis
Environmental and Atmospheric Science
Rational Drug Design
– New areas which will require exceptional computer resources:
• Bridging Length and Time Scales
• Quantum Computation
• Thermodynamics
•
•
•
•
From Molecules to Cells and beyond – Computational Biology
From Eddies to Aircraft - Fluid Dynamics
From Oceans to the Earth - Environmental Modelling
From the Earth to the Solar System – Solar Plasma Physics
22nd October 2001
UK Grid Support Centre - GSC
Scientific Software Infrastructure
One of the Major Software Challenges
Peak Performance is skyrocketing (more than Moore’s Law)
– In past 10 years, peak performance has increased 100x; in next
5+ years, it will increase 1000x
but ...
– Efficiency has declined from 40-50% on the vector
supercomputers of 1990s to as little as 5-10% on parallel
supercomputers of today and may decrease further on future
machines
Research challenge is software
– Scientific codes to model and simulate physical processes and
systems
– Computing and mathematics software to enable use of advanced
computers for scientific applications
– Continuing challenge as computer architectures undergo
fundamental changes: Algorithms that scale to thousandsmillions processors
22nd October 2001
UK Grid Support Centre - GSC
Improvements in Large-Area
Networks
•
•
•
Network vs. computer performance
– Computer speed doubles every 18 months
– Network speed doubles every 9 months
– Difference = order of magnitude per 5 years
1986 to 2000
– Computers: x 500
– Networks: x 340,000
2001 to 2010
– Computers: x 60
– Networks: x 4000
Moore’s Law vs. storage improvements vs. optical improvements. Graph from Scientific American (Jan2001) by Cleo Vilett, source Vined Khoslan, Kleiner, Caufield and Perkins.
22nd October 2001
UK Grid Support Centre - GSC
e.g. DOE SCIDAC Hardware Infrastructure
A Future Model for the UK ?
•
Flagship Computing Facility
– To provide robust, high-end computing resources for all
research programs e.g. TeraGrid
•
Topical Computing Facilities
– To provide the most effective and efficient computing
resources for a selected set of scientific applications
– To serve as a focal point for a scientific research community
as it adapts to new computing technologies
•
Experimental Computing Facilities
– To assess new computing technologies for scientific
applications
•
All Facilities on “The Grid”
22nd October 2001
UK Grid Support Centre - GSC
Why Topical Facilities ?
Variation in Scientific Application Needs*
Time
Memory
Storage
(TBYTES)
(TBYTES)
Node I/O
Code
Application
Cactus
ARPS
MILC
PPM
PUPI
ASPCG
ENZO
Astrophysics
Weather
Particle Physics
Turbulent Flow
Liquids
Fluid Dynamics
Galaxies
300
25
10,000
500
150
5,000
1,000
1.8
0.25
0.2
0.5
0.1
0.5
0.9
20
16
1
54
0.2
50
10
5
18
3
6
3
3
12
Variation
400x
18x
100x
6x
(TFLOP/S-HRS)
(MBYTES/S)
* From “High-level Application Resource Characterization,” NSF/PACI (National
Computational Science Alliance, May 2000)
22nd October 2001
UK Grid Support Centre - GSC
The 13.6 TF/s TeraGrid:
Computing at 40 Gb/s
26
24
8
4
Site Resources
Site Resources
HPSS
HPSS
External
Networks
Caltech
HPSS
5
Argonne
External
Networks
External
Networks
Site Resources
External
Networks
SDSC
4.1 TF
225 TB
NCSA/PACI
8 TF
240 TB
TeraGrid/DTF: NCSA, SDSC, Caltech, Argonne
22nd October 2001
Site Resources
UniTree
www.teragrid.org
UK Grid Support Centre - GSC
Grid/ Cluster Computing:
Where do distributed Clusters fit?
1 TF/s delivered
Distributed
systems
•
•
•
•
•
•
•
•
SE
TI@
ho
me
Int
ern
et
Co
nd
or
Le
gio
n\G
lob
us
Be
ow
ulf
Be
rk l
ey
NO
W
Su
pe
rcl
us
AS
ter
CI
s
Re
Tfl
d
op
s
Potential 15 TF/s delivered
Gather remote resources (unused)
System S/W manages resources
System S/W adds value
10% - 20% overhead is OK
Resources drive applications
Time to completion is not critical
Time-shared
Commercial: Entropia,
PopularPower, United Devices,
Centrata, ProcessTree, Applied Meta
22nd October 2001
•
•
•
•
•
•
•
•
MPP
systems
Bounded set of resources
Apps grow to consume all
cycles
Application manages resources
System S/W gets in the way
5% overhead is maximum
Apps drive purchase of
equipment
Real-time constraints
Space-shared
UK Grid Support Centre - GSC
Why we need a GSC
•
•
•
•
•
Software quality
– much Grid software is relatively new and immature
Local support
– most Grid software is from US research groups
UK strategy
– focus on a set of software and provide good local support
User skills
– many users will be scientists and not software experts
Coordination
– large user base needs coordinated support and development
22nd October 2001
UK Grid Support Centre - GSC
EPSRC Projects
Name
Partners
DAME
g
CombEchem
g
Reality Grid
g
GEODISE
g
Discovery Net
Theme
York, Oxford, Sheffield, Leeds, Rolls
Royce, Data Systems and Solutions
Distributed Aircraft Maintenance Environment
Southampton, Bristol, Roche
Discovery, Pfizer, IBM, Cambridge
Crystallograpic
Structure-property Mapping: Combinatorial Chemistry
and the Grid
QMW, Edinburgh, Manchester,
Loughborough, CfS, Schllomberger
Cambridge, Edward Jenner, SGI,
AVS, FECIT
Southampton, Oxford, Manchester,
Rolls Royce, BAe Systems, Fluent,
Intel, Microsoft, Epistemics,
Compusys
A tool for investigating condensed matter and
materials
ICST, Inforsense, Deltadot, Rvco
E-Science Testbed for high-throughput Informatics
Manchester, Nottingham, Sheffield,
EBML, IBM, GlaxoSmithKline, Sun,
AstraZeneca, Merck, Epistemice,
Network Inference, geneticXchange
Directly supporting the e-Scientist
Grid enabled Optimisation and Design Search for
Engineering
g
MyGrid
g
22nd October 2001
UK Grid Support Centre - GSC
PPARC Projects
Name
URL+Sponsor
GridPP
Www.GridPP.ac.uk
Theme
UK LHC analysis programme linked to EU DataGrid project
g
AstroGrid
Astrophysical Images and analysis
g
Other Research Councils are currently reviewing proposals
22nd October 2001
UK Grid Support Centre - GSC
Core Projects
Name
Regional Centres
URL
Theme
Www.nesc.ac.uk
Support for e-Science projects and outreach to UK industry
Www.grid-support.ac.uk
Helpdesk and in-depth support for funded e-Science
projects
AccessGrid
www.mcs.anl.gov/FL/
accessgrid
Create & deploy group collaboration systems using
commodity technologies
Grid Starter Kit
Www.grid-support.ac.uk
Tutorial material, Software and documentation, Globus,
Condor, SRB
Consolidated report to be
available soon...
Evaluate Globus and related middleware based on
experience of UKHEC sites
Grid Support
Centre
g
g
g
Evaluation Reports
g
22nd October 2001
UK Grid Support Centre - GSC
UK Grid Support Centre
key challenges
Develop a support mechanism for grid applications and
distributed job execution. UKHEC expertise available.
Help with distributed user administration
• authentication and certificates (UK CA)
• registration and grid-mapfiles
• e-Science project database
Help with installing Grid middleware
• Globus, Condor, SRB
• accounting and resource management
middleware
• new middleware components, e.g. coscheduling and instrumentation
• network monitoring
Help with system administration
• reference system and demonstrators
grid
Who do you call when something goes wrong ?
22nd October 2001
UK Grid Support Centre - GSC
Who is the Support Team ?
•
•
•
UKHEC Sites:
CLRC e-Science Centre (RAL + DL)
– running our own OST-funded e-Science programme
– major involvement in GridPP and other Grid projects
– HPCI Centre and extensive HPC applications development
Edinburgh Parallel Computing Centre/ University of Edinburgh
– National e-Science Centre (Edinburgh + Glasgow)
– HPCI Centre and extensive computing applications experience
Manchester Computing/ University of Manchester
– NW Regional e-Science Centre
– operates CSAR national HPC service
www.ukhec.ac.uk
22nd October 2001
UK Grid Support Centre - GSC
UKHEC and GSC Background
In January 2000 EPSRC funded a 3 year core activity to track and
disseminate information on international activities in computer
architectures, software and programming tools and to promote good
programming practice for the HEC community via workshops,
seminars and mentoring.
GSC commenced operation with distribution of the Grid Starter Kit
and announcement of the helpdesk at the e-Science Core Programme
meeting on 27th July 2001.
Staff at sites:
UKHEC
GSC
CLRC
2
4
Edinburgh
2
1
Manchester
1
1
also HPCI and CCL support teams
A strategic collaboration between the main UK centres offering
nation-wide academic computing support.
22nd October 2001
UK Grid Support Centre - GSC
UKHEC Key Topics
• Hardware: SMP/ DMM (ASCI) architectures + clusters (Beowulf)
• Software Development and QA Tools: faster development, maintenance and
exploitation
• Languages (Java, C++, Fortran90): ease of use and performance
• Optimisation: need highest performance - can tools help ?
• Visualisation and VR: demonstrate VR capabilities for scientists
• Data Management: demonstrate fast storage and access for science apps.
• Grid computing environment: evaluation and use of Globus, coordination of eScience activities between centres
• Standards: portability and longer code lifetime
• OpenMP/ MPI Programming: optimise for heterogeneous architectures
22nd October 2001
UK Grid Support Centre - GSC
Background:
Reports, Meetings and Workshops
Survey
“Survey of Computational Steering, Meta-computing and Network
Information Tools” (R.J. Allan DL-TR-99-002) and updated edition on-line
www.dl.ac.uk/TCSC/HPCI/reports.html
EPSRC GRID awareness meetings, e.g.
RI, 27 March 2000
Polaris House, 1 June 2000
UKHEC Grid Seminar and Workshop - 21-22 June 2000
www.dl.ac.uk/TCSC/UKHEC/GridWorkshop
Technical Report and Report for EPSRC
“Grid-based High Performance Computing”
www.dl.ac.uk/TCSC/UKHEC/metacomputing/metacomputing.pdf
“A Review of UK HEC Grid Infrastructure: State-of-the-art and Next Steps”
www.ukhec.ac.uk/publications/reports/ukhec-grid.pdf
UKHEC-funded early activities meant that the partners were in a good position to
lead in the UK’s e-Science programme. Set up of e-Science Centres and UK Grid
Support Centre has been achieved.
22nd October 2001
UK Grid Support Centre - GSC
What will GSC provide ?
a) Information
b) Technical support
c) Development
d) Technical liaison
22nd October 2001
UK Grid Support Centre - GSC
e-Science and Grid Support Centres:
What we have done so far ?
Experience of Globus and GSI installations: SUN Solaris, IBM AIX, Linux
(RedHat and SUSE), Compaq Tru64, SGI Irix
Globus working with local RMS: PBS, LoadLeveler, NQE
Evaluations of Condor and SRB on Sun and Linux
Grid software distribution portal: linked to DisCo archive, now part of Grid
Support Centre
Grid Starter Kit, help desk and Web sites
Re-launch of HPCProfile as HPCGrid Magazine
Membership of GGF working groups, mailing lists and NPI
UKHEC Globus Course under development
Centres established:
National e-Science Centre www.nesc.ac.uk
CLRC e-Science Centre www.e-science.clrc.ac.uk
NorthWest Regional e-Science Centre
UK Grid Support Centre www.grid-support.ac.uk
7 other Regional Centres
22nd October 2001
UK Grid Support Centre - GSC
Information
•
Support Web site www.grid-support.ac.uk
– downloadable software
– installation guides
– documentation
– introductory material
– evaluation reports (soon)
•
Grid Starter Kit via main site
– access via the Web site (kept up to date) or on CD-ROM
– supports installation of Globus, Condor, SRB
– additional software as required, e.g. GridEngine, LSF
•
National e-Science Centre www.nesc.ac.uk
– links to other Regional Centres
22nd October 2001
UK Grid Support Centre - GSC
Grid Support Centre Web Site
www.grid-support.ac.uk
22nd October 2001
UK Grid Support Centre - GSC
Technical support (1)
•
Help desk support@grid-support.ac.uk
– staffed during normal office hours (9h-17h)
– contactable via Web, email (or phone)
– normally first point of contact with the Support Centre
– located at RAL but provides links to experts at all centres
– uses ARS Remedy help desk management system
•
Expert staff at 3 sites
– provide help with installations
– work closely with Regional Centres
– visit user sites where necessary
– provide in-depth technical expertise for complex problems
•
Intended to work closely with Regional Centres
22nd October 2001
UK Grid Support Centre - GSC
Technical support (2)
•
Certificate Authority ca@grid-support.ac.uk
– issues digital certificates to e-Science programme participants
- will require nominated project, National or Regional Centre
contacts to validate applicants
– policy statement to be defined, agreed and published
•
Reference systems
– running recommended software installations
– will be available for users to access and study
•
Skeleton (Prototype) Grid
•
Training courses
– required for both developers and system administrators
– e.g. this course!!!
22nd October 2001
UK Grid Support Centre - GSC
Software Distribution and Support
: Grid Starter Kit (CD and Web)
Software Download Portal:
22nd October 2001
UK Grid Support Centre - GSC
Skeleton Grid
•
•
•
•
•
Link Regional Centres
Prototype production Grid
Identify issues for UK research and development
Provide facilities for UK demonstrator and pilot projects
Take part in International Testbeds, e.g. SC’2001, EcoGrid…
watch this space!
22nd October 2001
UK Grid Support Centre - GSC
Future work
•
•
•
Evaluate new software
– leading to recommended additions to the supported
programme software base
Custom developments
– software integration, enhancement, and filling gaps
– new capabilities provided through the Support Centre
User registration and accounting
– proposal to base on UoM Unix User Registration System
– ease problem of registering certified users on multiple Grid
resources shared within Virtual Organisations
22nd October 2001
UK Grid Support Centre - GSC
Technical liaison
•
Collaborate with Research Council application programmes
and Core Programme Centres
•
Develop links with US software development teams
•
Collaborate on software developments
•
Exchange technical staff with development teams
•
Participate in Consortium for Open Grid Software
•
Contribute to Global Grid Forum WGs, W3C, IETF, etc.
•
Develop links with industry
22nd October 2001
UK Grid Support Centre - GSC
Working with Industry
Grid Middleware and Services
End Users
IBM - e-Services
SGI - e-Services
Unilever
SUN - GridEngine, DRB, iPlanet, TCP
Jaguar
Platform Computing - Load Sharing
Facility (LSF)
BAe Systems
Entropia - harnessing Windows
platforms
others in future ...
New Productivity Initiative (NPI) - DRM
standardisation
Extreme Networks - QoS and highperformance networking solutions
22nd October 2001
NorthWest Development Agency
Advert:
First GSC workshop in association
with 12th Daresbury Machine
Evaluation Workshop
28-30/November 2001. See
www.cse.clrc.ac.uk/Activity/DisCo
UK Grid Support Centre - GSC
How to contact us
•
Via the Web
– http://www.grid-support.ac.uk/
•
By email
– support@grid-support.ac.uk
•
By phone
– 01235 446822
•
Web or email are preferred
22nd October 2001
UK Grid Support Centre - GSC
Summary
•
Its all happening very quickly !
•
Support service is now operational
– but please bear with us as we get up to speed
– staff recruitment in progress...
•
Constructive comments welcome
•
Priority is to get new users up and running
•
This is a partnership with the UK e-Science community
– we’ll do our best to help you
– but please also try to help us as well !
22nd October 2001
UK Grid Support Centre - GSC
Publications and URLs
Publications
R.J. Allan Survey of Computational Steering, Meta-computing and
Network Information Tools DL-TR-99-002 (Daresbury Laboratory, 1999
and 2000)
UKHEC Grid-based High Performance Computing (2000)
UKHEC A Review of UK HEC Grid Infrastructure: State-of-the-art and
Next Steps (2000)
R.J. Allan et al. Evaluation of Globus and associated Middleware (CLRC,
2001)
R.J. Allan et al. A Globus Developer’s Guide (2001)
R.J. Allan Developing a Web Portal for the Computational Grid (2001)
URLs
www.grid-support.ac.uk
www.e-science.clrc.ac.uk
www.ukhec.ac.uk
www.dl.ac.uk/TCSC/UKHEC/GridWorkshop
www.dl.ac.uk/TCSC/UKHEC/WG
www.dl.ac.uk/TCSC/HPCI/reports.html
22nd October 2001
CD-ROM
Grid Starter Kit
More stuff available,
please call us!
UK Grid Support Centre - GSC