Whole Earth image (VE record ID: 11622) courtesy of the Visible Earth project,
NASA Goddard Space Flight Center

Motivation

Understanding the factors that influence the Earth's climate are key to understanding phenomena such as global warming. ESM's aim to simulate the long term evolution of the Earth's climate by coupling together individual models of the climate system. The constituent models tend be diverse and often include simulations for the Earth's atmosphere, ocean, sea-ice, marine sediments, land surface, vegetation and soil, hydrology, ice sheets and the biogeochemical cycling within and between components.

Due to the large number of components involved, current ESMs tend to have reduced dimensionality and spatial resolution, or else tend to be too computationally demanding for long-term simulations. Consequently they represent highly idealised states of the Earth's climate that are rarely observed by scientists.

A schematic diagram of the components involved in the GENIE application. Different components "plug together" to produce an application that is Grid enabled through the ICENI middleware. Furthermore, componentisation allows simulation components to be added and/or removed as necessary.
Courtesy of Tim Lenton, Centre for Ecology and Hydrology, Edinburgh

Approach

Grid middleware, such as Globus, binds distributed resources such as computers, networks, on-line instruments, data servers or sensors etc., through a set of common protocols, allowing them to be viewed by the end user as a seamless computing and information environment - a virtual organisation. The virtual organisation allows different organisations to contribute different expertise and resources to a project, creating a more efficient and productive environment of collaboration.

The London e-Science Centre has developed a prototype componentised version of a GENIE simulation environment that utilises the integrated ICENI Grid middleware to provide a component composition, resource scheduling and distributed execution environment. This allows elements of an ESM, currently the atmosphere, ocean and sea-ice, to be placed on different distributed computational resources. The ICENI component based framework ensures flexibility and extensibility, by providing a plug-n-play system whereby models of the individual elements of the climate system can be inserted and/or replaced without rewriting the entire application.

Component composition of GENIE through the ICENI Netbean environment.

The GENIE Portal

As part of our activities we have developed a web portal interface to allow scientists to create experiments consisting of an ensemble of individual GENIE simulations, that can then be submitted to a Condor pool for execution. Results from the individual simulations in an experiment are available for download by the scientist through the portal when the experiment has finished.

Condor is a workload management system that can be configured to use spare CPU cycles on a set of idle workstations -- the Condor pool. The Condor pool at the London e-Science Centre consists of approximately 230 workstations including resources from the Department of Computing at Imperial College London, and the Southampton Regional e-Science Centre. Consequently, Condor allows these resources to be employed as a parallel computing resource.

These efforts have allowed scientists on the GENIE project to transparently harness a large source of computing power, and quickly obtain the results of their experiments.