Service description

ITS Research Computing manages a variety of servers and compute clusters to help meet the computational needs for Carolina researchers. A wide variety of commercial and open-source software applications is provided for use, as are multiple compilers and parallel operating environments. These systems, located in and administered by Research Computing, include:

1. KillDevil

The KillDevil cluster is a Linux-based computing system available to researchers across the campus. With more than 9,500 computing cores across 774 Dell servers and a large scratch disk space, it provides an environment that can accommodate many types of computational problems.

2. Kure

The Kure cluster is a Linux-based computing system available to researchers across the campus. With more than 1,840 computing cores across 230 blade servers and a large scratch disk space, it provides an environment that can accommodate many types of computational problems.

The KillDevil and Kure research clusters collectively support about 1,200 users.

3. Virtual Computing Lab

  • Windows-based computing through the VCL (Virtual Computing Lab): Provides “anytime, anywhere” access to a wide range of Windows-based research software applications. Windows XP and Windows 7 environments are available.
  • Linux sandbox environments – also based on the VCL: Provides “anytime, anywhere” access to base Linux environments upon which you can build a customized Linux environment for your research needs. Red Hat, CentOS, Ubuntu, Scientific Linux and Fedora base environments are available.

Cluster Charges

NOTE: the cluster charges proposal is formally approved for FY2014-2015. For more information, see timetable and status.

As a condition of procurement of the KillDevil cluster, to motivate considered use, foster a community of stakeholders, and establish appropriate resourcing for proposals, the charging structure for computation on KillDevil is:

  1. Institutional allocation of 200,000 core-hours per year to every lead investigator at no cost to the investigator;
  2. Charge of $0.005 (a half-cent) per core-hour per year for each additional hour beyond the institutional allocation.

Patron status remains available at $6500 per 16-cores, for a 4-year service period; use of patron resources incurs no additional charges for the service period, but overages above patron resources do follow the charging scheme.

For more information, please explore the other tabs in this section.

We anticipate this charging structure to be applied to future clusters, e.g., that which will replace Kure in upcoming years.

The cluster charges has been officially approved by the Office of the Provost:

  1. Waiving charges for FY2013-2014 (that’s all of 7/1/2013 – 6/30/2014);
  2. ITS-Research Computing will not collect charges less than $1000;
  3. Adding an exception process (see exception tab)

Thus, the overall timetable and status was per the table below.

Milestone Completed/Expected
Deliver monthly usage reports to lead-PIs February 2013
Interview affected investigators October / November 2013
Draft/propose modifications (see document here) December 2013 / January 2014
Publish website with status/resources/forms April 2014
Town Hall sessions April/May 2014
Formal approval of scheme June 2014

If the proposed phase-in is approved, billing will occur every six (6) months, with the first bill delivered in January 2015. Thus,

  • 1st bill January 2015 (for period of 7/1/2014 – 12/31/2014)
  • 2nd bill July 2015 (for period of 1/1/2015 – 6/30/2015)
  • Every six (6) months thereafter.

Bills are delivered to the lead-investigator of groups on the clusters (we also allow for designees to receive copies of the messages).

The bill lists the period/term of assessment. It also lists the total hours consumed, represents the 200,000 default allocation, the allocation of any patron cores associated with the lead investigator, and shows the balance of hours. It calculates, based on that balance, and the $0.005 per core-hour, the charges you owe. For a sample bill, follow this link, or follow the links on the Resources tab to the right.

You will need to do one of two things, or a bit of both:

  1. Provide FRS account numbers, the core-hours, and also charges to be applied to each account.
  2. Submit an exception / additional allocation request form for any balance.

Follow this link for a sample bill that has fictitious information supplied; and follow this link for a sample exception / additional allocation request form. And an attached exception narrative is here. If you would like to see a full spreadsheet of utilization for FY2012-2013, that is available as well; finally, the first six (6) months of FY2013-2014 extrapolated in gross fashion to a full year, is here.

Please note: ITS-Research Computing cannot tell you what to charge to what accounts. This is for two reasons. First, we don’t know (although: we can help you discern what codes/modules members in your group are running if you don’t know either). Second, only investigators who are awarded grants/contracts are authorized to determine the appropriate disposition of extramural funds–so even if we did know, we cannot make those judgments.

It is important to have an appropriate exception / additional allocation process. This is for at least two reasons.

  1. To accommodate legitimate uses of the computational clusters that should be sourced from alternate funding sources (e.g., institutional sources related to academic programs that in no way involve extramural projects);
  2. To enable a proper phase-in for projects and programs that are underway.

Rationales for exceptions include, but are not limited to:

  • Education: an investigator may support a project that is part of an educational or degree-seeking program where computation is inapplicable to a research program or project. E.g., this investigator may request an exception of all of his/her student’s computational burden.
  • Methods change: an investigator may have a program or project that was not expected to have any computational demand, yet midway through the effort it became clear that a computational component would further the project and position the investigator/university for future productivity. E.g., this investigator may request an exception to enable/further the existing project and develop future projects in excess of his/her institutional allocation.
  • Unanticipated overages: an investigator may have undertaken a project, and had some wherewithal to complete a significant computational burden, but unanticipated numerical methods or other impacts emerged which entailed a computational overage. E.g., this investigator may request an exception for computational charges in excess of $15,000 in a given year.

The exception process is progressive, from the Assistant Vice Chancellor for Research Computing, to the Research Computing Advisory Committee, to the Research Computing Coordinating Committee.

Follow this link for a fictitious sample exception form and narrative. Or, utilize the Resources tab in the right-hand column.

If you are an investigator, you should do four things:

  1. Include appropriate language in contract/grant proposals. For boilerplate, follow this link, or use the materials in Resources pane.
  2. Review what projects/grants/contracts members of your group are working on. Use your monthly usage reports to do this.
  3. Familiarize yourself with the sample bill. Follow this link, or see the Resources pane.
  4. Familiarize yourself with the exception form. Again, follow this link, or see the Resources pane.

Charges will not be waived in FY2014-2015. After 90 days from billing date, unpaid bills will result first in your accounts having dramatically throttled ability to use cluster resources, and second in being disabled.

For questions, concerns, issues, please contact ITS-Research Computing at research@unc.edu, or Michael Barker at michael_barker@unc.edu, or Mark Reed at markreed@email.unc.edu.

Hours

Services are available 24/7 other than during planned maintenance periods or unanticipated emergencies (such as loss of commercial power to the data center). As the compute clusters comprise large numbers of pieces of commodity hardware, at any point in time, an individual server may not be available.

 

Help and more information

For additional information about computing resources, email research@unc.edu or contact Michael Barker, michael_barker@unc.edu, 919-962-4314.

Research Computing: Getting Started Research Computing Cluster Comparisons