[Gate-users] What is the advantage of the cluster tools and ROOT file size

Philipp Matthies philipp.matthies at gmail.com
Wed Mar 22 15:17:34 CET 2017


Hi Varsha,

we're using HTCondor (formerly condor) as a cluster managing tool as
mentioned in the GATE Users Guide:
http://wiki.opengatecollaboration.org/index.php/Users_Guide_V7.2:How_to_use_Gate_on_a_Cluster

HTCondor is a cluster management software that makes sure to transmit
any required partial data to and back from the individual nodes. It
manages all connections to each node and makes sure they are available
to do work. Once a node finishes its currently assigned task, HTCondor
will assign the next task to any available core.

We installed gjm & gjs
(https://github.com/OpenGATE/Gate/tree/develop/cluster_tools). Both of
these tools are used on the master to separate jobs for the individual
nodes before distributing the workload and merging, respectively.

Each node writes its own output to a separate root file and after the
simulation is done gjm merges all the root files into one file. We
haven't experienced any issues with this setup, yet. Also all the
nodes finish at different times due to the randomness and therefore
write at different times as well.

Best regards,
Philipp

---

Philipp Matthies
Lead Engineer Nuclear Imaging

SurgicEye GmbH
Friedenstr. 18a
D-81671 München
Germany

philipp.matthies at surgiceye.com
www.surgiceye.com


2017-02-24 4:12 GMT+01:00 Varsha Viswanath <vvisw at seas.upenn.edu>:
> Hello GATE users,
>
> I am a new GATE user working on PET simulations and I wanted to know what
> the advantage of the cluster tools are, and whether they are worth
> installing. I am currently running multiple GATE runs on the cluster (1 or 2
> runs per core) but after about 12 - 18 hours, I have the issue where all
> instances of GATE are trying to write at the same time (similar to the issue
> described here:
> http://gate-users.healthgrid.narkive.com/j9rEt4LS/gate-on-cluster). The site
> does include a very useful command, but I wanted to know if the job splitter
> coordinates when multiple instances of GATE are running and when they're
> writing so it doesn't overload the cluster - which is the issue I run into
> now.
>
> In addition, I also notice that writing to my root files after they become
> somewhat large gets slower and slower. Is there a way to define a maximum
> file size in GATE?
>
> Thank you!
> Varsha
> --
> Varsha Viswanath
> HHMI-NIBIB Interfaces Scholar, PhD Student
> Department of Bioengineering
> University of Pennsylvania
>
> _______________________________________________
> Gate-users mailing list
> Gate-users at lists.opengatecollaboration.org
> http://lists.opengatecollaboration.org/mailman/listinfo/gate-users


More information about the Gate-users mailing list