Information Systems Services and the local GridPP team at Lancaster University provide access for NGS users to the new High End Computing Cluster. Each execution host in the cluster is powered by two quad core Intel Nehalem 2.26 GHz CPUs and has 24GB of Memory. Jobs are scheduled using the LSF job scheduler, and access is provided through a glite CREAM resource manager.
All users registered with the NGS have access to this resource through the ngs Workload Management System and the ngs.ac.uk Virtual Organisation (which all ngs users are automatically a member of).
Instructions on how ngs users can submit to this resource through the ngs User Interface (UI) can be found on the ngs wiki here. To specify the Lancaster HEC as the destination for your jobs add the following lines to your JDL (Job Description Language) files:
Requirements = other.GlueSubClusterUniqueID == "abaddon.hec.lancs.ac.uk"; Requirements = other.GlueCEUniqueID == "abaddon.hec.lancs.ac.uk:8443/cream-lsf-normal";
"abaddon.hec.lancs.ac.uk" is the name of the resource manager.
The compute nodes are installed with a x86_64 CentOS5 operating system, and have many different 32 and 64-bit libraries for C, C++, fortran and python installed. Reasonable requests for package installs will always be considered, but licenced software cannot currently be run by ngs users submitting to the HEC. Due to job submission limitations the maximum amount of RAM that any single job submitted to abaddon is currently permitted to consume is 3GB.
User support requests for Lancaster should in the first instance be directed to the NGS Support Centre: support@grid-support.ac.uk
