You are here: Home News 2014 Newsletter 5 (Nov 2014)

Newsletter 5 (Nov 2014)

Our HPC-Newsletter, published by Freiburg’s IT-Services, aims at informing about HPC activities. It targets administrators, operators, and users of HPC infrastructures.

Dear colleagues,

On November 6th, we will see the review on our bwForCluster ENM at the DFG in Bonn. We are in the final stages of preparing our presentation. Many thanks to our scientific communities for their contributions and support!

At the same time, we see that critical components of the bwGRiD hardware are beginning to fail repeatedly, leading to day-long service outages.

We will therefore speed up the process of migrating to the new HPC service infrastructure. This means

a) Shutdown of the bwGRiD service on December 1st
b) Early access to the new "bwForCluster ENM (pre)" from December 2nd

We hope that you enjoy reading our newsletter. Should you have any further questions, please do not hesitate to contact us.

With best regards,
Your HPC Team, Rechenzentrum, Universität Freiburg

 

Table of Contents 

Important dates

SHUTDOWN: bwGRiD and Lustre in Freiburg

Early access to preliminary bwForCluster ENM, Codename "NEMO"

HPC Competence Center ENM

 

Important dates

  • 06.11.2014: DFG preliminary decision about bwForCluster in Freiburg and Tübingen
  • 01.12.2014: bwGRiD Freiburg goes offline
  • 02.12.2014: Start of the bwForCluster ENM (pre) - Codename NEMO
  • 05.12.2014: Start of the computational/theoretical chemistry cluster (JUSTUS) in Ulm
  • 18.12.2014: Lustre storage of the bwGRiD Freiburg goes offline

SHUTDOWN: bwGRiD and Lustre in Freiburg

Unfortunately, we have to inform you that the bwGRiD cluster in Freiburg will go offline on December 1st (computation service) and December 18th (file service).

The last months have seen numerous failures of the LustreFS that resulted in day long outages of the bwGRiD cluster, despite great efforts to keep the service alive.

The bwGRiD service has been running far longer than originally expected. However, with the current degree of unreliability which cannot easily be dealt with, we cannot afford to keep the service running.

New jobs will only start if they finish by Monday, December 1st (4 weeks from now). After December 1st, no new jobs can be submitted.

Login will be possible until Thursday, December 18th. We will provide instructions on how to transfer the data to another service (e.g. bwFileStorage) on our website: http://www.hpc.uni-freiburg.de/nemo/migration.

Possible alternatives:

  • Early adopters are invited to use our testbed for the new bwForCluster ENM infrastructure, starting December 2nd (see next section)
  • GPU machines will be consolidated into the BFG infrastructure
  • bwUniCluster in Karlsruhe is available as a general computational resource (the share of Freiburg is still not fully utilized, thus queue waiting should be short)
  • If your scientific domain is Theoretical Chemistry, you are also eligible to use the forthcoming bwForCluster JUSTUS in Ulm (terms and conditions apply: entitlement and acceptance by user board)

Early access to preliminary bwForCluster ENM, Codename "NEMO"

We already have started to use parts of the bwGRiD hardware as a testbed for our new cluster infrastructure.

The shutdown of the bwGRiD hardware gives us a window of opportunity to extend the testbed to all salvageable bwGRiD hardware.

If you do not mind being part of a "work in progress" effort, we can offer you early access to this new service, which is running on old hardware.

We need the feedback from early adopters (e.g. software requirements)!

Some key features:
  • bwGRiD entitlements remain valid
  • CentOS 7
  • Virtualization with KVM and OpenStack (ongoing research)
  • Testing MOAB, SLURM and TORQUE, but we will stay compatible to the qsub/msub submit you know from the bwGRiD and bwUniCluster
  • User job monitoring (i.e. You can evaluate to what degree your job is using the allocated resources)
  • If special software is required - please ask

For more information see http://www.hpc.uni-freiburg.de/nemo.

HPC Competence Center ENM

We have started to establish recurring domain specific consultations with our scientific communities. On September 18th, we met with the Elementary Particle Physics community. On October 6th, we met with the Neuroscience community. Our third community, Microsystems Engineering community will follow in November (tba).

In these first consultations, we discussed possible governance models and virtualization options.

Throughout the lifetime of the project, these domain specific consultations will take place in regular intervals. This allows the HPC-Team to continuously synchronize with our scientific communities and turn this into a community-driven effort.


HPC-Team Rechenzentrum, Universität Freiburg
http://www.hpc.uni-freiburg.de

bwHPC-C5
http://www.bwhpc-c5.de

To subscribe to our mailinglist, please send an e-mail to to hpc-news-subscribe@hpc.uni-freiburg.de
If you would like to unsubscribe, please send an e-mail to hpc-news-unsubscribe@hpc.uni-freiburg.de

Filed under: