1. The Academic Service Partnership
Peter Dew & Joanna Schmidt
plus
White Rose Grid Teams

2. Overview
The White Rose Grid as an example of an inter-university collaboration providing an early production Grid service
Computing Services roles
Delivering the WRG
Organisational structure
Technical implementation
User management
Lessons learned
Benefits
Concluding remarks

3. About the WRG
The White Rose Grid (WRG) works under the auspices of the White Rose University Consortium (WRUC) – an association of the three research Universities
Leeds, York & Sheffield
Employs complementary skill bases to support both: larger projects than can be delivered by any one University, and a broad research agenda
WRUC featured as a model of collaboration and enterprise in the HEFCE White Paper
Why Grid?
Enhances the competitive position of the three Universities to attract funding
Enables inter-enterprise computing resources optimisation and an increase in the effectiveness of service delivery to researchers
In this slide we want to get across the idea that WR Grid supports a Virtual Organisations with strong business reasons for its exist

4. WRG aims
to strengthen e-Science research (using experience gained from e-Science projects such as DAME, HYDRA, or gViz)
initial focus
decision support (engineering, health, social science)
scientific visualisation
to support and enlarge new scientific communities including bio-technology, aerospace, tissue engineering and healthcare
to assess and grow, in collaboration with YF, regional demand for Grid technology

5. Commitment
Senior staff from the three Universities (White Rose Grid Executive: Chief Exec of White Rose Univ Consortium - M Doxey; P Dew and K Brodlie – Leeds; J Austin – York; P Fleming – Sheffield)
Senior Computing Services staff (C Cartledge – Sheffield, S Chidlow – Leeds) & Computing staff from Comp Science Dept (A Turner – York)
White Rose Grid staff
WRG Project Teams (Computing Services staff & Computer Science staff)
IT Vendors – Esteem involving Sun & Streamline

6. Computing Services roles
To provide a stable, well-managed and responsive HPC service
To promote effective use of HPC facilities across the three Universities through a variety of training including joint seminars and user group meetings
To offer user-support & training in basic HPC techniques (e.g. parallel programming), Globus and e-Science applications
To support an early production Grid service under the leadership of WRG technical staff

7. WRG staff responsibilities
Technical developments (Leeds)
Grid training (Sheffield)
Coordination of joint activities (Leeds)
Liaison with e-Science communities within the WRG, UK e-Science and others (Leeds)
Business outreach (York)

8. Setting up the WRG
Purposely acquired - with over £3M investment - 4 HPC nodes (in total nearly 500 CPUs)
A heterogeneous facility comprising 3 clusters of Sun shared-memory systems and 2 Intel processor-based Beowulf clusters
25%
To offer both:
local HPC services (75% resources)
the Grid infrastructure (25% resources)
Each node specialises in the provision of a distinct service
75%
WRG resource allocation

9. The WRG architecture CFD node General Purpose HPC node
Engineering Application node
Computer Science node

10. WRG software & hardware stack
Software stack composed largely from open source software
Technical implementation
Software stack composed largely from open source software
Basic building blocks:
- HPC Cluster Tools & Sun One Studio
- Sun Grid Engine Enterprise Edition
- Globus Toolkit v2.4
Portals using Apache Tomcat & Grid Portal Development Kit (GPDK)
Further portal developments planned based on the Struts web application framework and the Java CoG kit

11. Delivering our Grid Procedures & resources:
Strong organisational structure
Computing infrastructure - computer systems, storage currently being expanded with a Storage Area Network (SAN), networking infrastructure (YHMAN reprocurement underway -implementation due Nov'04), software (includes traditional HPC tools, Grid software - Globus Toolkit, and Portals)
Mixture of experienced support staff and research staff working in teams

12. WRG project teams WRG USERS Architecture Team Globus, MyProxy, portals
Authentication, Authorisation
& Accounting Team
user management, usage account.
X.509 digital certificates
WRG USERS
Technical Team
stable service
Training Team
HPC techniques, Grid access & applications
Business Outreach Team
working with regional companies & Universities

13. Access to the WRG
Users baffled by Globus (due to its novelty and lack of easy documentation)
Preferred way of access:
using Grid portals developed by the WRG
running Grid-enabled applications e.g. the gViz project has developed Grid-enabled IRIS Explorer modules
Portals need to be developed

14. The DAME XTO portal
Enables aeronautical engineers to identify abnormal behaviour in aircraft engines by performing DSP analyses of vibration data from onboard sensors.

15. User management Includes: user registration
user authorisation for access to resources
user validation & approval of request for a digital certificate
accounting for resource usage
documentation of procedures
These schemes and user administration processes were developed taking into consideration:
the distributed nature of WRG
the cultural differences in registering and managing users at the three sites
the existence of two distinct classes of users: local and WRG as well as including other academic and commercial partners

16. Managing information New registration forms needed to be developed
Documentation for users, e.g.
how to register & obtain a digital certificate
how to access the WRG systems
further local user documentation
Documentation for system administrators such as:
registering users
propagation of Grid distinguished names between systems and mapping out local UNIX user names in the grid-map files
producing usage accountings reports
Development of Web pages

17. User registration & authorisation
Local user registration
University users
Completion of the Application Form for WRG Resources
Industrial partners
Obtaining a digital certificate
Approval of the new WRG project and the user by PI
User requests a digital certificate
User validation by Computing Service
Approval of the new WRG project and the user by the local WRG Executive member
Authorisation
Approval of a UK CA digital certificate
Validation of the new WRG user by Comp Services
requesting update of grid- map files
User registration details
Allocation of local & remote WRG usernames
User registration at remote site 2
Local Computing Service
Request for user registration
User registration at remote site 1
A centralised database of WRG users

18. Digital certification
Globus requires personal X.509v3 digital certificates
WRG systems support certificates from the UK e-Science Grid Certification Authority (CA)
The e-Science CA at Rutherford Appleton Laboratory (RAL) is being run as part of the Grid Support Centre
Registration Authorities (RA) were established at the three Universities
Training of User Administration staff
courses available at RAL
a short introduction to digital certificates issued by the UK e-Science CA available at:
At present the RA interface to the CA system will only work reliably with Netscape 4.79
All certificates issued will expire after one year

19. Lessons learned
New user management procedures must be fully endorsed by Computing Services (e.g. the WRG local identity had to be modified several times to take this into account)
Continuous staff training is required (due to rapidly changing technology – e.g. GLOBUS)
The local system administrators must be involved with their own Globus & other Grid fabric’s installations

20. Benefits to Computing Services
Added dynamic to support issues
Enlarged support team memberships
Broaden knowledge by working in collaboration with other sites
More interesting job specifications for system and user support staff (i.e. include a research approach)

21. Overall lessons learned
Complexity due to:
geographically distributed support teams (lack of full understanding how the three sites work)
large number of support staff involved (new issues may cause confusion who is doing what)
innovative technology (lack of good understanding of new implementations and software dependency/interoperability; lack of good documentation)
human interaction factor (caused by misunderstandings etc)
communication issues within a VO (due to its size)
constantly posed questions of ownership and trust (due to crossing organisational boundaries)
distributed resource management (e.g. software revisions)
software licensing issues ( need licensing for a Grid)
increased exposure to security issues
lack of a central Help Desk

22. Addressing issues within WRG
Many problems resolved through:
effective organisational structure led by the WRG Executive
Computing Services staff involvement and expertise
research element (portal development, Globus installations) led by Computer Sciences staff but with the involvement of Computing Services’ practical approach and using their well-established support infrastructure

23. WRG Evolution e-Science Grid WRG WRG Com- Services C C C WUN Grid
Academic Service
Infrastructure
Companies
“Buy
Services”

24. Concluding remarks The WRG serves as a test-bed Grid environment
Addresses a large variety of problems and issues, including key sociological constraints (human interactions, ownership, trust) reflected in global Grids
Computing services staff expertise is vital to Grid success
Many gaps (see the e-Science Gap Analysis at )
Need to continue to work with Grid users and increase the number of Grid applications and enhance outreach

25. References WRG web site - http://www.wrgrid.org.uk/
DAME XTO portal –
P M Dew, J G Schmidt, M Thompson, P Morris The White Rose Grid: practice and experience – in the proceedings of the All Hands conference
e-Science Gap Analysis

26. Thank you for your attention