IceCube Institutional Memorandum Of Understanding (MOU)

University of Maryland

Greg Sullivan

Ph.D Scientists (Faculty Scientist/Post Doc Grads): 5 (3 3 1)

Scope of Work

Faculty:

Greg Sullivan (L,+) – Former Spokesperson, Data Systems, ExecCom, ICB, Institution lead, Outreach, NGIC upgrade coordination

Kara Hoffman – filter development, Radio R&D, Outreach

Jordan Goodman – Coordination with Milagro/HAWC, Outreach

Scientists and Post Docs:

Erik Blaufuss –   Former Analysis Coordinator, TFT board Member, PnF, IceTray, SVN repository, Operations Group, ICC

  Analysis topics: GRB, real time alerts

TBD to replace –  Online – near real time GRB analysis

Elim Cheung –  Offline Processing development & testing, Simulation

 Thesis/Analysis topics: Neutrino Oscillations

John Felde   Analysis topics: GRB, Real time alerts, Neutrino Oscillations

Alex Olivas –  Tuesday Call co-convener, Software management, Software Coordinator

 Analysis topics:

Ph.D. Students:

TBD to replace

Ming Song –  Filtering and Detector R&D

 Thesis/Analysis topics: Multi-messenger transient & GRB search

Liz Friedman – Core Software, Datasets for filter testing & common MC datasets

  Thesis/Analysis topics: GRB

UMD General M&O (non-science) IceCube Responsibilities and Contributions:

The Maryland Group’s major responsibilities and contributions towards maintenance and operations of the IceCube experiment include:

·   Primary institutional responsibility for the maintenance of the online PnF filter system.

·   Primary institutional responsibility for the maintenance of the IceTray analysis framework, SVN code repository and software package building.

·   Major responsibility for the maintenance of the IceCube simulation package (IceSim).

·   Software coordinator Alex Olivas

·   The Maryland group maintains a computing cluster of about 750 cpu cores and 48 GPU boards (24 GTX980, 24 GTX1080) with online disk storage of more then 350TB dedicated to IceCube activities. A minimum of 350 cpu cores and all GPUs are reserved for dedicated simulation production under the coordination of the IceCube simulation production manager. Maryland also provides resources to host and maintain a 64 GPU card system for UW.

Institutional (UMD) resource contribution to Computing:

The maintenance and operation of the computing cluster includes:

1.  High quality Computing Space, cooling and power (provided by UMD)

2.  Networking and high speed connectivity to the Internet (provided UMD)

3.  System administration (.5 FTE sys-admin) (partial support by UMD)

4.  Hardware maintenance on a 5-year replacement cycle of $40k/year (partial support by UMD).

5.  First ½ of GPU cluster purchased by UMD ($80k) and maintained with help from UMD

6.  hosting and maintaining an additional 64 GPU cards provided from UW for MC production

1. & 2. Computing Space, cooling and power & Networking and high speed connectivity to the Internet

The University of Maryland provides high quality space, cooling and power. The IceCube group is provided essentially unlimited space in a modern HPC computing facility for research computing on campus. The facility is monitored 24/7 by provided technicians and we have 24/7 secure access. The current Maryland-IceCube system occupies 10 rack spaces with additional space set aside for possible expansion.

Host to UW GPU system of 8 GPU systems for 64 GPU cards.

Maryland is a major hub for the Internet-2 backbone in the northeast US. The University provides a 10 Gb/s fiber connection directly from the interenet-2 backbone into our cluster in the research computing facility. In addition, the university provides a dedicated fiber between the research computing facility and our research group in the physics building.

3., 4. & 5. System administration & Hardware maintenance on a 5-year replacement cycle

The University of Maryland provides $40k per year in funding to be used towards the total. The system administration is approximately .5 FTE and includes administration for the computing cluster as well as about 1 dozen workstations used by the PA group. The hardware maintenance for the compute cluster is $40k per year.

Page 2 of 3