JISC

When browsing the web, you type in a URL or click on a hyperlink and the page appears in your browser. Beneath this simple action, however, a complicated chain of events occurs to retrieve and load the elements within the page - images, text, animations etc. - from remote sources spread over the planet.

What is a grid?

A grid shares with the web this ability to access remote, distributed resources while hiding the underlying complexities. Whereas the web retrieves information, however, a grid also retrieves and integrates  services providing computational power, data storage capacity and instruments. These resources are connected into a virtual organisation, which is really a social structure of people who have agreed to share resources in some manner for a period of time. A grid thus enables people to work collaboratively on large scale scientific, engineering and compute and data intensive challenges.

The software that enables these resources to work together with little or no user intervention is called ‘middleware’ because it lies between the network connecting the resources and the user. It is analogous to the protocols a web browser launches when retrieving all the different elements of a web page.

Most grids use middleware based on the Globus toolkit, which is open source software developed by the Globus Alliance. The Storage Resource Broker (SRB), developed by the San Diego Supercomputing Centre, is a commonly used tool for storing and accessing data over a grid. However, grid middleware is still under development and it could be some years before a complete set of standards and protocols is defined.

The UK National Grid Service

The National Grid Service, funded by JISC, the Engineering and Physical Sciences Research Council (EPSRC) and the Council for the Central Laboratory of the Research Councils (CCLRC), was created in October 2003 and entered full production in September 2004. It is led and coordinated through the CCLRC in collaboration with the University of Manchester, the University of Oxford and the White Rose Grid at Leeds. The NGS’s resources are provided by these four founding members plus the national high performance computing facility (HPCx) and a growing number of partner and affiliate sites.

The four core sites currently provide more than 36TB of data storage and 150 dual processor nodes for computation, with Manchester and the CCLRC focusing on data intensive application. This will increase to more than 100TB of storage and 220 dual processor nodes under phase 2 of NGS development which begins in October 2006.

The standardisation of interfaces, based on the Globus toolkit and SRB, allows additional sites to join and the universities of Bristol, Cardiff, Lancaster and Westminster are contributing a variety of additional computing resources. The University of Westminster also provides a portal to allow NGS users to manage their use of the service through a graphical interface.

Standard interfaces also enable the NGS to interoperate with other grids both nationally and internationally. Within the UK, these include GridPP (the particle physics grid), campus grids at a growing number of universities and regional grids such as the White Rose Grid.

Internationally, the NGS provides a national ‘gateway’ to other e-infrastructures, such as the US eraGrid. It also helps to ensure international interoperability by participating in activities related to the European Enabling Grids for E-sciencE (EGEE) project and sitting on international standards bodies such as the Open Grid Forum (formerly the Global Grid Forum).

How to use the NGS

Before applying for an account on the NGS, you need a digital certificate from the UK Certification Authority You can then apply for an NGS account online where you can also find help and advice.

Users who know which resources they want to access, can submit a job directly to their chosen site or to their chosen NGS resource. Others can access the most appropriate, available resources for a task by means of resource brokers or web-based portals.

Portals have the advantage of allowing users to build workflows that, as well as defining procedures for executing a job, also capture metadata about how it was done, thus facilitating later repetition.

The NGS is working with JISC to simplify access by making it compatible and interoperable with the new single sign-on to web-based services that JISC is introducing for the UK academic community. This will mean that a user will have access to the NGS after signing-on once only to his or her desktop computer.

Future plans

In October 2006, the NGS will enter the second phase of its funding and development. As well as an increase in data storage and compute capacity, the NGS will launch a major new initiative with the Open Middleware Infrastructure Institute UK (OMII UK) to ensure that the middleware OMII UK develops meets the needs of NGS users. In addition, a roadmap is under preparation to determine when new services will be introduced, such as visualisation, access to facilities and data repositories. Ultimately, the National Grid Service will provide integrated, coherent access to the full range of the UK’s computation and database research facilities, together with a range of sophisticated services to support novel collaborative and cross- resource activities.

Who’s using the NGS?

The goal of the NGS is to support innovation and collaboration in academic research across a wide range of disciplines. Some examples of NGS use:

• RealityGrid a major UK e-Science project, simulated and visualised complex condensed matter structures, such as oil moving through porous rock. Some of the simulations used supercomputers on two continents - HPCx in the UK, accessed via the NGS, and supercomputers on the US TeraGrid, which were linked in a transatlantic grid. 
• NGS is hosting the database of Computer Crystal Structures for a major Research Council-funded project. It contains the computational predictions of the crystal structures adopted by an organic molecule. Users have easy access to data files via a data portal and the Crystal Navigator web interface. 
• NGS compute nodes are being used to speed up complex simulations of an enzyme involved in replication of the HIV virus, throwing light on the factors that lead to HIV drug resistance. Many simulations, each consisting of 30,000 atoms, can be submitted at once, reducing the time taken to get results.
• A model of individual income, derived from a detailed, but small population sample, has been applied to a much larger sample using NGS compute nodes. Insights into the distribution of poverty within different ethnic minority groups have been revealed.
• The NGS is also developing the grid interfaces to large scale facilities, such as the ISIS Neutron Source and the Diamond Light Source, both located at the Rutherford Appleton Laboratory in Oxfordshire. The aim is to define the middleware that will enable researchers to transfer results directly to compute resources for post-processing, analysis and visualisation.

Example

NGS helps plan radiotherapy treatment ‘Radiotherapy is a critically important tool in the treatment of cancer. To get the most effective treatment possible, we need to simulate (model) the interactions between the radiation and the tumour. A radiotherapy treatment consists of approximately 100 trillion such interactions, and as a consequence, the full simulation requires an excessive amount of computer time.
Until recently, we had to run the simulations on a PC over several days. However, access to the National Grid Service has reduced calculation times by a factor of about 10: answers are now available in a matter of hours.’

Dr Steven Weston, Institute of Medical Physics and Engineering, Yorkshire Centre for Clinical Oncology

Further information

Introduction to grids
http://gridcafe.web.cern.ch/gridcafe
www.grid-support.ac.uk/content/category/7/41/64/
www.gridtoday.com/02/0722/100136.html

Middleware
http://www.omii.ac.uk/
http://www.globus.org/
www.sdsc.edu/srb/index.php/Main_Page

Grid news
www.interactions.org/sgtw

This briefing document was compiled and edited by Dr Andrew Richards, Director of the National Grid Service, and Judy Redfearn from the JISC Communications team.