The University of Edinburgh is a leading participant in the UK's e-Science initiative. Few institutions, if any, are better placed to have an impact on the practices of the e-Science community. Via our personal contacts with e-Scientists in our own institution and via our contacts with the National e-Science Centre and the Edinburgh Parallel Computing Centre we will bring our work to the attention of the widest possible audience. The University of Edinburgh is a partner in the eDIKT initiative (e-Science Data, Information and Knowledge Transformation) at the National e-Science Centre, which provides us with a strong dissemination route to e-Scientists.
Our programme of work on well-founded approaches to Grid resource usage is squarely focussed on the needs of Grid users. The Global Grid Forum Advanced Programming Models Research Group lists flexible composition of Grid resources as one of the primary requirements for improving Grid usage and quality of service [6]. Leading e-Scientists point to proof-carrying-code approaches such as ours as being an important area for future research.
``As the scale of data increases, it becomes progressively more economic to ship computation to the data. The computations that correspond to scientific models linking, selecting or deriving data are much more demanding than those currently permitted as programmed methods in queries. ... Significant work in precisely characterising the behaviour of this class of computation, e.g. with proof-carrying-code techniques, safety, scheduling, monitoring and accounting mechanisms, and new policy and authorisation managers will be needed. This requires substantial research.'' [22]A recent survey from an information technology consultancy adopted by the UK e-Science program also stresses the importance of mobile code approaches thus, ``For certain kinds of research Grids, it may be necessary to allow users to run their own codes in order to pursue their investigations.'' [23]. The report continues by emphasizing the importance of controlling resource consumption: ``However, whether you use explicit account mappings or web services, you still need a scalable way to manage the rights given to individual remote users to access specific Grid resources''. Our techniques provide resource guarantees for mobile code, addressing both of these points.
Static analysis methods as applied in the ESC/Java methodology and the Aspinall/Hofmann resource bound type system improve the quality of program codes. The correctness of scientific codes is a highly significant matter. If scientific computations contain errors then all of the computational expense incurred in executing them and all of the intellectual effort invested in the analysis and interpretation of the results obtained would at best be wasted. In general interpreting the results of a scientific computation with errors could lead to making flawed political, economic or strategic decisions based on erroneous conclusions made from erroneous results. For this reason we believe that scientific codes should be thoroughly analysed by the methods which we have presented above.