Workshop 1: Delivering Grid Services - the role of Central Computing Services
Organisers Clare Gryce (UCL Research Computing) – c.gryce@ucl.ac.uk
Jeremy Yates (UCL Research Computing) – j.a.yates@ucl.ac.uk
Bruce Beckles ( University of Cambridge Computing Service ) - mbb10@cam.ac.uk
Support
The organisers gratefully acknowledge the support of the e-SciNet project ( http://www.nesc.ac.uk/escinet/ ).
eSciNet will fund travel, accomodation and registration costs for funded members of the project.
Significant contributions towards travel, accomodation and registration costs for other attendees may also be available from the e-SciNet project; please contact Clare Gryce ( c.gryce@ucl.ac.uk ) for details.
Background
For the purposes of this workshop we define Grid Services to be any computational activity that includes multiple and/or distributed resources or where middleware of any form is used to mediate or manage job submission or data access.
In order to ensure the usability and long-term sustainability of institutional Grid-based services, it is essential that support for these services is embedded in the central Computing Services Departments of UK HEI's.
To date, much of this support has been provided by Computer Science and other academic departments, making services vulnerable to changes in the local funding environment and institutional ICT policy.
Where such services ARE supported by central departments, this has generally happened in isolation, with the result that solutions are often site-specific. Further, there have been limited opportunities for knowledge transfer in this important area.
Attendance and Registration
Attendance is by registration prior to the event.
This workshop aims to bring together the following communities, all key stakeholders in the vision for a sustainable Grid-services infrastructure:
- Representatives of Central Computing Services Departments from UK Universities (Systems Administration, Network Services, Research Computing etc)
- Representatives of middleware developers such as OMII ( http://www.omii.ac.uk/ )
- Representatives of bodies who fund or provide core infrastructure/services to UK HEI's, such as the JISC, JANET(UK), National Grid Service ( http://www.grid-support.ac.uk/ )
- University-based Computational Scientists wishing to use Grid-related services such as the Collaborative Computational Projects (CCP's).
Please email Clare Gryce ( c.gryce@ucl.ac.uk ) to register your interest in attending this event.
Objectives
- To identify common challenges in the deployment and support of Grid-related services by Central Computing Services Departments of UK HEI's
- To identify and explore common approaches to overcoming the above challenges
- To establish whether there is a need for an ongoing series of meetings in this area, and what the priorities of such a programme might be
- To agree next steps in relation to the above, as appropriate
- The organisers will write a formal report of the workshop for distribution to attendees and key stakeholder groups (above).
Morning Session
- Distributed compute resources for all: users, resource owners and central IT services
Bruce Beckles (e-Science Specialist)
University of Cambridge Computing Service
The role of e-Science Specialist is to serve as a "champion" for e-Science within the University, to develop and manage HTC and other e-Science-related services in the University and to advise the University Computing service on how best it can cater to the requirements of scientists, particularly those working in computational science.
This talk will discuss the challenges of making HTC resources available - in a usable and secure manner - to ordinary scientists in the University of Cambridge . It will focus particularly on the role of the central IT services in managing and facilitating this endeavour. An additional complicating factor is that Cambridge is a federation of autonomous institutions, each of which provide at least some of their own computational resources. The talk will also discuss how the provision of such resources fits into the University's larger IT strategy and outline current thoughts as to the future use and development of these resources in Cambridge .
- GridPP and the Edinburgh Compute and Data Facility or How a general purpose cluster bore the weight of Atlas on its shoulders
Mike Baker and Orlando Richards
University of Edinburgh
This talk will consider some of the challenges encountered in trying to host part of a large grid project on a multi purpose cluster that has to be used by local users.
- Networking requirements for 3 GRID-related projects – some case studies or Unblocking the GRID - Tales of a Lowly Plumber
(Speaker TBC)
Research Computing Network Group, Information Systems, UCL
This talk will consider, as case studies, the networking challenges posed by the requirements of three Grid-related projects (ie. multi-institute collaborative projects across multiple admin domains):
- Culture shock - traditional sysadmins and IT security meet ‘The Grid’
Alan Real (HPC User support) & Jason Lander (NGS Cluster System Manager)
Information Systems Services University of Leeds
IT security people consider 'paranoid' to be a complement. The people who run large, complex systems value stability over having the latest and greatest software; this is not the way The Grid likes to work.
We will describe how we tried to get The Grid running in a world where firewalls exist, existing software cannot be updated and rebooting is the last resort.
Afternoon Programme
The afternoon session will consist of an open, Chaired discussion around the issues raised during the morning's talks. All attendees are encouraged to participate in this discussion session.
Discussion will be structured around a number of key questions, prompted by the morning's talks, including (not not necessarily limited to):
- Job submission – how can users be provided with a convenient, attractive means of seamlessly accessing resources both within their institution, and external to it ( UK and beyond)?
- Securely removing network policy barriers – how can firewalls and other security devices be enabled to accept jobs and data from both within the institution and external to it, without incurring an unreasonable administrative overhead?
- Enabling secure direct addressing of machines – what steps need to be taken to ensure that the operating systems of centrally managed (and other) computational services will support direct addressing of individual nodes?
- Removing network bottlenecks – what policy and funding decisions need to be taken to ensure a UK networking infrastructure to support rapid movement of large data sets (including videostream) between Universities and other UK research institutions and facilities?
- Resource Allocation and Funding Models – what strategies have been adopted by HEI's wishing to support Grid-related IT services in response to Full Economic Costing (fEC) and in the context of more general IT service provision?
|
