Spring2014Meeting

HPC and BD Researchers Meeting 1:30 to 2:30PM 07 May 2014 GITC 3710

Some Candidate Discussion Topics

HPC and BD Wiki

• User input

Computational Hardware

• Tartan HPC Initiative
  • What about new, more capable nodes !!
• Architecture
  • Virtualized headnodes
• Catalog
  • CPUs
  • GPUs
• SMP (Symmetrical Multiprocessor Machine)

Storage

• Storage tiers : Roadmap

Software

• Modules
• Compilers
• Applications
• Commercial

Running Jobs

• Run jobs on compute nodes with AFS token
  • via ksub
  • via SGE
• Using krenew
• HTCondor

2020 Vision Survey

https://wiki.hpc.arcs.njit.edu/2020 | 2020 Vision : Researchers' HPC, Storage and Staff Needs 2015-2020 Results

Workshop Candidates

• Grid Engine Tips and Tricks
• Compiling and Running OpenMP and MPI Program

Documentation

• Additional topics
• User contributions

Use of kong.njit.edu for courses

---

Message from Usman Roshan, CS :

From usman@njit.edu  Wed May  7 16:50:37 2014
To: David Perel <davidp@njit.edu>
cc: ResearchGroupMeeting@njit.edu
Subject: Re: HPC and BD meeting, 1:30 PM 5/7/2014

Hello,

Sorry Zhi and I couldn't attend this meeting today. We had the department
retreat. I have some thoughts and suggestions about our HPC.

1. We need more powerful nodes in Kong both in speed and RAM. I see we
have 16GB RAM nodes in the Kong extension and that is helpful. Nodes with
8 cores would also help.

2. Installing condor was very helpful and we were able to finish days of
work in one day. Disabling the nightly restart on OSL machines would help.

2. I'm satisfied with the gigabyte space allocation but we are now
requiring terrabytes. And so we need to make terrabyte space allocation
available to researchers here. I'm doing a big data project that requires
6TB and am pleased that the university has given the space for it. We need
to ensure that this continues because we will continue such large-scale
projects.

3. We would love to have at least a couple of GPU nodes in the main
cluster. Not only will it allow GPU student training
 but also enable big
data research. Most universities, even ones ranked below NJIT in U.S.
News, have GPU clusters.

Thank you for arranging the meeting and once again sorry I couldn't be
there in person.

Usman

Message from Yun Q. Shi, ECE :

From shi@njit.edu  Wed May  7 14:00:08 2014
Subject: RE: HPC and BD meeting, 1:30 PM 5/7/2014
To: David Perel <davidp@njit.edu>

   RE: HPC and BD meeting, 1:30 PM 5/7/2014

Hello David,

Below is my opinion. I think Big data is important and in fact my group'
work really need powerful computation to handle the big data. Hence, we
strongly appeal that* NJIT should fund the facility** so as to support
NJIT's research involving Big data and high performance computing*.

Look forward to hearing from you regarding any news/issues.

Best,

Yun

Yun Q. Shi

ECE Dept.