Remote rendering with ParaView on HPC node with “pvserver”

Posted by on December 2, 2016 No comments

You can launch remote ParaView server on the HPC compute node to render the data stored on HPC (no need to transfer data to your client machine)

  1. Reserve a compute node, say 1 “whole” node for 4 hours
salloc -N1 -n8 --exclusive -t 4:00:00

(you can also do it with a slurm submission script of course)

2. then on the node you have been allocated by slurm (say zeus15), launch ParaView server process:

[aa3025@zeus2 ~]$ salloc -N1 -n8 -t 4:00:00
salloc: Granted job allocation 16726
[aa3025@zeus2 ~]$ qstat
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)
16726 short4 bash aa3025 R 0:12 1 zeus15


[aa3025@zeus2 ~]$ ssh zeus15
[aa3025@zeus15 ~]$ /share/apps/paraview/ParaView-5.2.0-Qt4-OpenGL2-MPI-Linux-64bit/bin/pvserver -display :0.0 --use-offscreen-rendering

Waiting for client...
Connection URL: cs://zeus15:11111
Accepting connection(s): zeus15:11111

Leave this window alone, the server is now accepting connections on port 11111.

3. Now we need to tunnel from your Desktop machine (which must have the same verion of ParaView installed as on the node (ParaView-5.2.0). First we establish the ssh tunnel from port 11111 of the node to port 11111 of your Desktop PC via Zeus’s login node, say zeus2.

So on your Desktop machine set up the tunnel (add user name before zeus2 if necessary, i.e. user@zeus2):

ssh -L 11111:zeus15:11111 zeus2

You will be logged in to zeus2 in this ssh session. keep this terminal running this is your link to the target node zeus15.

4. Next open local ParaView (Linux,Windows) and “connect to server”   “localhost:11111” -> in the 1st console running the server, you will see that connection is made:

[aa3025@zeus15 bin]$ ./pvserver -display :0.0 --use-offscreen-rendering

Waiting for client...

Connection URL: cs://zeus15:11111

Accepting connection(s): zeus15:11111

Client connected.

Now you can open your data files in ParaView directly from your home folder on zeus and process them etc and create your animation:

Once you checked it is all working you can make small 1-line slurm script for starting the ParaView server with sbatch-submitted job.

Alex Pedcenko.

RStudio on Zeus

Posted by on November 27, 2016 No comments

We have R 3.3.2 installed on zeus (login nodes and compute nodes). You can also access R on one login node via RStudio web interace at http://zeus.coventry.ac.uk/R

Alex Pedcenko

Current (experimental) limits of queues

Posted by on November 14, 2016 No comments

The total number of nodes (or CPUs) you can use depend on how long your job has to run (i.e. in which queue/partition it was submitted):

[queues are listed from higher to lower priority, i.e. shorter queues have higher priority in the waiting list!]

for up to 4-hours short jobs:

  • “short4” queue has unlimited number  of CPUs for up to 4 hours  and can use nodes from any of the queues! (Hint: do not specify how many nodes you need, just specify how many CPU-cores you need for your job, i.e. for 1000 CPU-cores for 4 hours “sbatch -p short4 -n 1000 -t 4:00:00  submitscript.slurm”). Default time (if you do not specify walltime) is 1 hr.

for up to 12hours jobs:

  • You can use up to 144 nodes x 8 CPUs  of “all12”  queue for up to 12 hours.

For up to 24hrs jobs:

  • You can use up to 80 nodes x 8 CPUs  in “all” queue (640 CPUs) for up to 24 hours.
  • You can also use 10 nodes x 32 CPUs of Broadwell queue (another 320 CPUs, use less nodes per job, these are “fat” nodes! ) for up to 36 hours
  • You can also use 20 nodes x 8 CPUs of “all48” (another 160 CPUs) queue for up to 48 hours (lower priority than “all“)

For up to 36hrs jobs:

  • You can use up to 10 nodes x 32CPUs in Broadwell queue
  • You can use up to 5 nodes x 32CPUs (160 CPUs and 10 K80 GPUs) in NGPU queue
  • You can use up to 18 nodes x 12 CPUs in (+ 36 K20 GPUs)  in GPU queue
  • You can use 1 SMP node (32 CPUs and 512 GB RAM) in SMP queue

(specialized queues SMP,GPU,NGPU have higher priority in the waiting list, i.e. if you need to use GPUs on these nodes, you have higher “weight”)

For up to 48hrs jobs:

  • You can use 20 nodes x 8 CPUs (160 CPUs in total) in queue “all48” for up to 48 hours

For >48 hrs long jobs:

  • You can use up to 20 nodes (160 CPUs) in queue “long” for unlimited job time

 

 

 

 

“RequiredNodeNotAvailable” status of the Job

Posted by on November 11, 2016 No comments

There is a standing reservation of all nodes in “all” queue and “Broadwell+NGPU+Phi” nodes for this Sunday 13/11/16 from 0:00 to 12:00, which is needed for conducting more performance tests before commissioning of HPC. So If your submitted job spans through this time period you will get this message as a reason for “queueing”.

Alex Pedcenko

Zeus Alive!

Posted by on November 3, 2016 No comments

Just few final tests left after hardware / software upgrades, which took place during couple of weeks at the end of October 2016. The major “full blast heat generation” test will be performed on Monday 7/11/2016 to see if the server room survives. If you want to test some of your codes/jobs before then please ask Alex Pedcenko aa3025@coventry.ac.uk to enable your slurm accounts.

What have we got now:

the HPC specs we will have after the latest upgrade:

Old stuff:

  • 144 nodes with 2xIntel Xeon L5530 @ 2.40GHz CPUs (Nehalem 8 CPU-cores/node), 48 Gb of RAM per node: queues all and long
  • 18 nodes with 2xIntel(R) Xeon(R) CPU E5-2440 0 @ 2.40GHz (Sandy Bridge 12 CPU-cores/node) and 2x NVidia Tesla K20 GPUs, 48 Gb RAM:  queue GPU
  • 1 x 32 CPU SMP Machine Intel(R) Xeon(R) CPU E5-4620 0 @ 2.20GHz (32 CPU-cores), 512 Gb RAM: queue SMP

Compute nodes of HPC are interconnected with High-Speed QDR Infiniband (40 GBps)

New Broadwell CPU based nodes:

  • 44 nodes with 2 x Intel(R) Xeon(R) CPU E5-2683 v4 @ 2.10GHz (32 CPU-cores/node) 128Gb RAM:   queue Broadwell
  • 10 nodes with 2 x Intel(R) Xeon(R) CPU E5-2683 v4 @ 2.10GHz (32 CPU-cores/node) 128Gb RAM + 2x NVidia K80 GPU:  queue NGPU
  • 2 nodes  with 2xIntel(R) Xeon(R) CPU E5-2683 v4 @ 2.10GHz (32 CPU-cores/node) 128Gb RAM + Xeon Phi coprocessor SE10/7120: queue Phi

2 x 15 Tb file servers (/home and /share)

Outside of zeus HPC

  • One “Big Data Analytics” machine Bullion S2 server, 12 Tb storage, Xeon® processor E7 v3 family, 128Gb RAM

Compute nodes of new HPC part  are interconnected with High-Speed FDR Infiniband fabric (54 Gbps)

 

 

Alex Pedcenko

New Intel Broadwell (32 CPU-core) nodes

Posted by on September 15, 2016 No comments
  • 44 nodes with 2 x Intel(R) Xeon(R) CPU E5-2683 v4 @ 2.10GHz (32 CPU-cores/node) 128Gb RAM:   queue Broadwell
  • 10 nodes with 2 x Intel(R) Xeon(R) CPU E5-2683 v4 @ 2.10GHz (32 CPU-cores/node) 128Gb RAM + 2x NVidia K80 GPU:  queue NGPU
  • 2 nodes  with 2xIntel(R) Xeon(R) CPU E5-2683 v4 @ 2.10GHz (32 CPU-cores/node) 128Gb RAM + Xeon Phi coprocessor SE10/7120: queue Phi

at the moment each queue has 36 hrs time limit for a job.

Unplanned shutdown last Friday 20-May-2016

Posted by on May 23, 2016 No comments

Last week, on Friday, zeus HPC did experience some significant NFS (shared filesystem /home) troubles. File system was very slow and unresponsive. Hence, during the weekend there was an unplanned maintenance, system software upgrades and, hence, shutdown of all the queues. Apologies if your running jobs were terminated. All four login nodes zeus[1-4] were reinstalled, as they were also choking with some file system errors. One of the reasons for overfilling of the /tmp space on the login nodes were multiple running VNC sessions on different login nodes. Currently, VNC sessions are limited to only one login node: zeus.

At the moment all queues are up and running again. Let me know if any occasional library is still missing on login nodes.

Website http://zeus.coventry.ac.uk is back on as well. You can access login nodes via their individual login names zeus,zeus2, zeus3, zeus (alias “zeus” isn’t yet operational)

Best Regards,
Alex Pedcenko

Fast & Dirty Comparison of Compilers’

Posted by on February 22, 2016 No comments

A simple C code for parallel multiplication of two 512×512 matrices ran with OpenMPI across 4 nodes (were applicable). Three instances of the same source code were compiled with GNU C (gcc 4.4.7), Intel 11 and PGI 11.8 compilers. Results presented on the graph below.

Both OpenMPI 1.10.2 and “the code” were compiled with the same parallel compilers (the lower the values the better performance).

The winner is:  Intel ICC compiler ! (well PGI is almost the same as Intel, but costs more money)

mmult

 

Alex Pedcenko

Intel Free Software Tools — bring your own license

Posted by on January 12, 2016 No comments

Qualifications

Academic Researchers must provide an academic institution email address to qualify for this program.

Offer:

A free 12 month named-user license for the Intel® Performance Libraries

  • Intel® Math Kernel Library
  • Intel® Data Analytics Acceleration Library
  • Intel® Threading Building Blocks
  • Intel® Integrated Performance Primitives
  • Intel® MPI Library (not available for OS X*)

Get Intel® Performance Libraries for:
Linux* ›
OS X* ›
Windows* ›

Are you interested in more software tools than the libraries listed above?

To apply for a personal user license for Intel Software follow link here

Planned shutdown 19 December 2015

Posted by on December 10, 2015 No comments

HPC zeus will be shut down for the whole day (13 hours starting 6:00AM) on Sat 19-Dec-2015 due to power maintenance works scheduled by Estates. Your reservation of the nodes will not be accepted if it falls within this time slot. HPC will be automatically powered off on 6:15 AM 19-Dec-2015.

Regards,
Alex

css.php