A Framework for Domain-Specific Science Gateways.

While modern Federated Computing Infrastructures --- Grids, Clouds and other technologies --- continuously increase their computing power, their use for research still stays lower than desired. The authors' diagnosis of this problem is a technology barrier hard to overcome to people who want to focus only on science. The federated infrastructures are difficult to use not only due to the physical distribution of the resources and, thus, need for remote access, but, mainly, due to the fact that everyday patterns of interaction with a computer cannot be directly used for these resources. The way of performing computing operations on them is different than the usual way the scientists do their research using laptops or personal computers. The authors claim that the key to increase the use of modern federated infrastructures for science is making the processing on these infrastructures resemble using a personal computer. The paper collects requirements from different scientific use cases and, from these requirements, derives a processing model that could satisfy all of them, thus, allowing to build a system, in which computations on large infrastructures can be similar to everyday work. This model is implemented by a framework for creating Science Gateways --- InSilicoLab.