System Description

Host name: hybrid.ncsa.illinois.edu

Hardware

SuperMicro board
- 2x 8-core Intel Xeon CPU E5-2630 v3 @ 2.40GHz
- 64 GB DDR4
- 4 PCI-E 3.0 ports
4x NVIDIA P100 GPUs
- 3584 cores
- 16 GB HBM 2
1x 75GB SSD and 6x 4 TB HDDs, arranged as 2x RAID 0 arrays

Software

need picture

Projects

Start date: 03/2017
End date: TBD
PI: Eliu Antonio Huerta Escudero <elihu@illinois.edu>
- Users: Daniel George: dgeorge5, Miguel Holgado: holgado2, Eliu Huerta: elihu, Daniel Johnson: dsjohns2, Michael Usachenko: usachenk
Objective: Develop deep learning based methodology for gravitational wave detection in support of LAGO project.
- This is a dedicated platform for this specific project because developers need interactive GUI-based access to the system.
Current status
- 03/01/17: 4x P100 donated by NVIDIA were installed
- 03/17: access granted.
Open issues
- need to mount RAID0 arrays and setup user space on them

Start date: 02/2017
End date: 05/2017
PI: Luda
- Users: xnie8, yanzhan2
Objective: Evaluate new congestion control algorithm that should help with long-distance data transfer of large datasets.
- TCP-BBR: Google’s new congestion control algorithm
- This required 4.9 kernel.
Current status
- 02/17: system configured, access granted
- 03/17: waiting for availability of servers abroad
Open issues