Science gateways made easy: the In-VIGO approach

1. Requirements and the In-VIGO approach to grid-computing According to the definition of a science gateway as a " community-specific set of tools, applications, and data collections that are integrated together via a portal or a suite of applications, providing access to Grid-integrated resources " , there are several requirements to be met by its supporting Grid infrastructure [1]. A non-exhaustive discussion of these requirements follows. Tool requirements: Science gateways must support tools that are diverse beyond differences in the programming languages used for their implementation. Some are sequential while others are parallel codes. Some are open-source research-grade programs for an initially restricted set of users while others are commercial codes-for which only binaries might be available-with a large user base. Some tools can be treated as trusted codes but, in general, they must be assumed to be untrusted. Tool interfaces and usage modes also vary greatly. They can be command-line oriented with text-like inputs and outputs. Some tools subsume other software-such as Matlab-to generate graphical outputs from numeric data. Most commercial tools have relatively sophisticated graphical user interfaces (GUIs). GUIs are often integrated with the computational parts of the tools and do themselves require appropriate software libraries. In some tools the data-input, execution and data-output phases are sequential whereas in others they are concurrent, possibly requiring composition with other tools and workflow control. User and developer requirements: Many users are interested in using tools whereas some are only interested in demonstration runs. Tool users may be interested in using the codes for educational and/or research purposes. Both batch and interactive modes must be supported. Tool usage may require extensive resources and execution times or may only use a few computational cycles on a single machine. An important class of users consists of those who develop, modify or compose existing tools in order to maintain them, experiment with new models or adapt existing ones to specific needs. Whereas tool users may only require access to tool interfaces, developers need an execution environment that allows access to code and the use of tools for development, testing and debugging. They also need means for easily turning tools into Grid-services. Shared infrastructure requirements: It should be possible to dynamically add and remove resources across administrative domains, while preserving security and privacy of data, codes and other users' information. Resource owners should never have services denied or compromised by Grid users. These requirements embody …