In the sections above, a model for providing centralized (University operated) clusters within the generally decentralized University cluster environment has been proposed. The design maximally leverages existing administrative resources, improving cluster-specific support to local administrators and using the already existing LAN structure of academic computing as a base that should significantly lower overall administrative costs. It should be capable of providing remotely sited, remotely managed cluster resources to groups physically located anywhere on campus at a cost that is both fair and perceived as being fair by granting agencies and the research groups that own them. Finally, the model itself is one that can fairly easily be reevaluated on an annual basis and adjusted as necessary as actual measurements of cost and utilization are developed and a proper CBA becomes possible.
The model is scalable, and should be able to accomodate anywhere from a hundred to thousands of University managed nodes, with direct recovery of most of the actual costs of installation and operation, recovery of recurring physical infrastructure costs from indirect costs charged to the associated grants, and perhaps a gradual investment in an overall increase in the core infrastructure provided by OIT to the University as a whole (funded out of all the University's revenue streams) in a way that is entirely equitable as clusters become more and more common and more and more important in the University's grant-funded research efforts and educational mission.