Research Data Center Policy
The University of Chicago provides central data center facilities to house e-infrastructure supporting a variety of research activities. The total amount of available research data center capacity (power, cooling, and physical space) is constrained by the significant capital expense required to add data center capacity. Hence, it is prudent to manage how this finite resource is operated and allocated among research activities to optimize its benefit to the University’s research mission.
IT Services is responsible for oversight and management of campus data centers, and the Research Computing Center (RCC) provides hardware, software, and customized support for University research computing. The data centers host shared computing and storage resources available to all faculty that are provided by the University; they also host dedicated, proprietary compute and storage resources for University departments and individual faculty that are purchased by departments or faculty research grants for a fixed period of time. These proprietary resources are governed by Memoranda of Agreements (MOA) with the RCC that align with the policy outlined below.
There are three key principles that guide prudent data center management. The first is maximizing the utilization of the total amount of computing power, storage, and network bandwidth that is available within data center space. The second is ensuring that research data centers are operated to meet physical security requirements across research activities, which helps ensure that research programs are not limited by research data center management practices. Third is an appropriate governance body charged to steward the data centers in accord with prudent management.
The purpose of this University-wide policy is to implement these key principles in order to optimize the overall management and utilization of these vital and costly research resources. The benefits include optimal and equitable allocation of data center resources and regular refreshment of old hardware to increase power usage efficiency.
The Data Center Steering Committee (DCSC) is responsible for proposing policies that optimize data center management and utilization and any subsequent changes to established policies. The Board of Computing Activities and Services (BCAS) and the Research Computing Oversight Committee (RCOC) will advise and revise the policies to be subsequently reviewed and approved by the Provost IT committee. The DCSC will oversee the implementation of the approved policies, review and approve operating practices of the University Data Centers and periodically re-establishes particular standards and acceptable practices for Optimization and Physical Security, and communicate these to management of cognizant University units. Managers of research data centers and other cognizant University units are charged to ensure that their operating practices remain aligned with standards established by the DCSC.
- The management of various University units should adopt practices that align with and help to implement this policy. For example, Procurement Services should guide procurement of research computing hardware accordingly, managers of research data centers should provide periodic notification to remind responsible parties of the dates by which their equipment will be removed, and standard decommissioning procedures should be followed to ensure smooth and predictable transitions. The DCSC will determine such aspects of a holistic optimization program and make management of cognizant units aware of their role in the program.
- All research data centers must be operated to meet a minimum standard of physical security identified by the DCSC.
- The maximum power-efficient lifespan of hosted computational and storage resources in the data center is estimated to be five years.
- Hosted compute and storage resources will be decommissioned five years from the date the hardware is deployed in University data centers or per the specified date in the MOA.
- Managers of data research centers will notify faculty who have proprietary hardware in a data center at least 12 months in advance of the upcoming decommission date.
- Deans and department chairs will be provided with a list of faculty in their units with hardware to be decommissioned at the beginning of each fiscal year.
- Hardware removed from the data center will be wiped and recycled.
Policy Owner: tbarton