Forschungsinfrastruktur

University of Salzburg

Salzburg | Website

Short Description

The research infrastructure consists of a network of 15 high-end server systems from Supermicro with multi core CPUs based on the AMD64 architecture. It can be used for calculations and test setups in the fields of software development and algorithm research. The nodes are connected with 10Gbit/s ethernet and 2x 56Gbit/s (per computing node) Infiniband FDR and can be used for dedicated low level programming and development of RDMA capable software.

Additionally applications and operating systems can be run in an isolated and abstracted way using virtualisation and container technologies. The research infrastructure allows to satisfy diverse research needs. Memory- and CPU-intensive research projects as well as distributed computing projects utilizing multiple nodes at the same time can be run on the available infrastructure. Available configurations range from systems with 96GB up to 1TB RAM and 12 to 64 cores per node. Fast storage systems using RAID with double parity and backup systems in different locations allow safe storage of research data.

A Continuous Integration (CI) system for rapid development and distribution of research software to all nodes is available.

Contact Person

Prof. DI Dr. Nikolaus Augsten

Research Services

Systems and algorithms for processing data
Storing and querying large data
Approximate query processing
Spatio-temporal database systems
GIS enabled databases
Interconnected nodes with 2x56Gbit/s Infiniband (per computing node) and 10Gbit/s ethernet technology
Direct Infiniband hardware access for RDMA-enabled software and services
Abstraction and Isolation of running systems using virtualisation and container technologies

Methods & Expertise for Research Infrastructure

The research infrastructure is designed for research in the field of data engineering with the aim of solving efficiency problems in data processing. In this branch of research, new algorithms are developed, implemented and empirically evaluated. The empirical evaluation requires precise runtime measurements, measurements of memory consumption and network traffic. This usually requires exclusive and physical access (bare metal) to individual servers or a cluster of servers. The use is characterized by frequently changing configurations in order to be able to carry out tests under different conditions.

The research infrastructure is technically professionally administered and offers support services for the execution of experiments, e.g. versioned storage of experimental setups and large experimental data, E2EE for sensitive research data, fully automatic provisioning of the cluster nodes, as well as databases for measurement results. Researchers are supported and advised by technical staff when setting up their experiments. From a scientific point of view, there is a strong wealth of experience in design and the empirical evaluation of single-core, multi-core and parallel as well as distributed shared-nothing algorithms.