The IT Centre of the Aristotle University of Thessaloniki hosts and operates a medium sized institutional HPC infrastructure. The "Aristotelis" cluster is comprised of more than 30 physical nodes that cumulatively amount to more than 1000 CPU cores. It is logically split upon partitions (queues) of similar (homogenous) resources. Jobs are scheduled upon these partitions via the Slurm Workload Manager, an open source cluster management and job scheduling system.
Upon the largest partition (named batch), which is comprised of 20 physical nodes and 400 CPU Cores the backend interconnect is Infiband (FDR) allowing for the efficient execution of large parallel (i.e. MPI) jobs and intensive I/O workloads. The amount of memory per node within this partition is 128GB, hence a total amount of 2.5TB of RAM would be available if one where to fully utilise this partition.
The "gpu" partition is comprised of nodes that include nVidia Tesla (P100) GPU cards (each of which accounts to 3584 accelerator cores). It is also worth mentioning that the Centre operates a special partition named "vnc" that allows HPC users to obtain a remote desktop environment directly upon the HPC cluster resources. This is extremely helpful especially for newcomers and users that depend heavily on applications that utilise graphical user interfaces (which is for example relatively common in engineering applications).
The operating system used is CentOS 7. In terms of compilers the following options are available:
- GCC Compiler Suite,
- PGI (Professional Edition) Compiler Suite and
- Intel Parallel Studio (Composer Edition).
A wide variety of scientific software is available for the HPC users. An indicative (yet not extensive) list of applications is available here.
The outcomes of the HPC activity and focus of the IT Centre are recorded and published here (page in Greek). These include research publications, doctoral and master students' thesis and more among others.
Beyond operating the institutional HPC infrastructure the IT Centre further provides support services for accessing and using efficiently the underlying resources addressed towards students, researchers and the teaching faculty of the University.