Shared Cluster Computing at UCSF #

Purpose #

The shared cluster computing at UCSF was founded by basic scientists to support their research programs. In the process, researchers benefit from maximized economies of scale in building, administering, and using the cluster. The cluster provides computing capacity, network connectivity internally at UCSF and externally to research networks such as the Pacific Research Platform, data storage, job queueing with accounting, and systems administration. As a result, the cluster contributes directly to the basic mission of UCSF, including research, education, and outreach.

Principles #

The cluster is built on a co-op model rather than a fee-for-service model, in contrast to, for example, most technology cores at UCSF and high-performance computing provided by Amazon Web Services. In the co-op model, every contributing research group has immediate access to the cluster computing power and storage capacity proportionate to their contribution; contributions are preferably made in cash, although suitable existing hardware may also be integrated. This funding model suits best the often unpredictable timing and amount of research funds available to individual projects. In addition, our goal is to maximize the number of UCSF users of the shared cluster. Therefore, every member of the UCSF community has access to a fraction of the cluster computing power and storage capacity regardless of their contribution, reflecting the contribution by UCSF and shared instrumentation grants from NIH and NSF. Finally, any unused compute power is available to any user on a first-come, first-served basis. Every effort is made to accommodate and expand the set of applications that can be run efficiently on the cluster, as long as they don’t negatively impact our founding goals. If a contributing member leaves the consortium, they are entitled to take their contribution with them (except for communal storage).

Governance #

The cluster is governed by a Faculty Steering Committee, guided by the cluster’s purpose and organizational principles outlined above. The Committee consists of major contributing basic scientists, a representative of guest users, and a campus Information Technology representative, all of whom have some expertise in cluster computing. The Committee meets as required to review overall utilization of the system, plan future expansions, discuss funding opportunities, etc. The Committee also resolves any disputes concerning use and governance of the cluster. Committee matters are addressed at meetings and by email as needed. To ensure continuity of governance, the Committee appoints new members by a majority vote. The Committee currently consists of Kliment Verba (Co-Chair), Sergio Baranzini, Scott Pegg, Brian Shoichet, Tom Ferrin (Co-Chair), and Mandy Terrill (Campus IT).