There are many ways to accomplish this. One is by expanding the local and distributed computing resources further, a place for many mid-scale and small-scale projects/tasks (c.f., section 2.1) will exist. These services need to be integrated with the supercomputer systems to allow some omnipotent scheduling process (currently users that understand available resources and their applications) to move appropriate tasks to the local and distributed computing class. A compatible applications software environment across the range of computational systems (local, distributed, high performance) is necessary to achieve this goal.
The numerous political and sociological obstacles are compounded by the infancy of true distributed computing environments and the fact that there is little experience in the user community in using coupled but distributed resources and tools. The NERSC facility is again in a unique position to prototype and foster the development of such tools and environments in a production facility. This may require the installation of capacity resources at the NERSC facility and will definitely require the cooperation of the user community.