Slater Computing Laboratory
News and announcements
|
John C. Slater Computing Laboratory
The Slater Lab provideds computing support for QTP.
This includes support for desktops and laptops, the computer
network and high performance computing and visualization.
It also includes support and training for software
engineering and programming, in particuilar parallel
programming.
News and announcements
- Termination of the Slater Lab
October 14, 2011
All business IT services will be handed over to UF
Enterprise Services (email) or CLASnet (Home disk, web
service, and printers), and the HPC Center (research
computing). All QTP servers will be turned off on
Wednesday November 23, 2011.
- Changes in the Slater Lab
August 4, 2011
With the general support for Research Computing by the
Unievrsity, all research computing activities supported
by the John Slater Laboratory will be moved to the UF
HPC Center. This includes all clusters for computation
and all scratch disks /scr/arwen_1 and /scr/arwen_2
(Roitbeerg), /scr/crunch_4 (Cheng), /scr_crunch_5
(Bartlett), /scr/crunch_6 (other QTP groups).
/scr/wukong_2 (merz) was moved over a year ago.
- Changes in Compute Clusters
January 20, 2010
The Merz group has acquired a new cluster, called
kongming.
The new cluster is operated by the UF HPC Center and is
under control of the HPC Center Moab scheduler. Jobs
must be submitted from the HPC submit nodes
submit.hpc.ufl.edu. There are two load balanced
nodes, submit1 and submit2, accessed under the same name.
You can
use HPC nodes test02 and test04 for
interactive work that goes beyond normal job and file
management tasks that can be performed on the submit
nodes.
Thus on the HPC linx64 is replaced by
submit1, submit2, test04, and
test05.
The kongming nodes are dedicated to the Merz group
memebrs and are selected in the job file as other QTP
clusters are selected now, namely from the job script by
specifying the attribute "kongming" with the -l option
#PBS -l nodes=1:ppn=12:kongming
The nodes are identified by their rack and slot number,
such as r14a-s20, which it is the left node in
slot 20 while r14b-s20 is the right-hand node (there are
two independent nodes per rack slot). The nodes are also
known by their "kongming" aliases such as kongming1,
kongming2, etc. but this is mostly irrelevant
since users should not log into individual notes since
they are under the control of the scheduler.
Several changes will occur that are prompted by this
acquisition. Each step will be announced separately.
-
Move of QTP Moab Scheduler to linx64
Next week Tuesday Jan 26, 2010
the QTP Moab scheduler, now running on wukong
will be stopped and will start running on
linx64.
Since all QTP jobs already have to be submitted from
linx64, this will not make a big difference to
users.
The main disruption will be that running
jobs will be killed when the old scheduler is
turned off.
Please let me know if that causes a problem.
-
Move of wukong to Larsen Hall machine room
On the same day as the scheduler move to minimize
disruption of running jobs, the storage
server wukong
will be moved to the Larsen Hall machine room.
This server hosts the file system
/scr/wukong_2 which is accessed by the
kongming, wukong, ra, and
crunch servers.
The wukong cluster will be
moved at the same time as well. This will optimize
network connectivity and power and cooling
management of the server and the cluster.
The /scratch/wukong file system will then
become available on all HPC nodes, including
kongming, wukong, and ra.
It is already available at this time to support
kongming.
It can be accessed from Mac and Linux desktops and
laptops through the HPC center Samba server
smb://samba.hpc.ufl.edu. Please open a
bugzilla
ticket to request access through SAMBA.
The /scr/wukong_2 fle system on the QTP
systems will no longer be available.
-
Maintenance on ra
Finally, the nodes of the ra cluster will be
updated with the latest HPC node software to make it
compatible with the current HPC and kongming
clusters.
The racks of ra will be moved within NPB 1114
to improve air flow and cooling.
Following this work, the ra nodes will
be scheduled by the HPC Center scheduler as well.
-
Wider use of ra capacity
With the addition of the ra cluster to the
HPC Center scheduling pool, QTP users will have
access to a proportionately larger share of the HPC
Center resource pool.
The Merz group members will retain first priority.
Please contact me if you have any questions or concerns
about these changes.
- HyperChem 8 license for NPB
September 20, 2009
The compuational chemistry software HyperChem 8 for
Windows is available to all members of QTP. We have a
license that will work in the Physics Building network
Versions for MacOS and Linux are expected in a
few months.
Details on how to get the CD and how to install
HyperChem on your laptop or desktop can be found here
>> top
|