IPE 10-19 Internship or Master Thesis: Optimizing the bandwidth of high performance detectors with online compression
Nowadays high-speed streaming cameras are able to produce data with rates of up to 10 GB/s. To process such large amounts of data, the computing infrastructure relies on clusters with GPU-accelerated servers connected by an Infiniband fabric. However, storage operations are still a major performance bottleneck and the duration of experiments is limited by the available system memory. To reduce the load on the storage subsystem and prolong the duration of experiments, the available computing resources can be used for online data compression.
The internship will be performed within an international project that aims to develop a novel instrumentation for ultrafast imaging at synchrotron light sources. The student will work with the selected image data from the local ANKA synchrotron light source. She/he is expected to analyze the available datasets and propose a data reduction scheme for video streams and arbitrary floating-point data. The selected approach should bring essential reduction of the data flow for the selected data sets and be fast enough for execution on the available computing resources in real-time.
Very good knowledge of the C programming language and parallel programming models. Familiarity with existing image and video compression algorithms is a plus.
Institute for Data Processing and Electronics (IPE)
as soon as possible
limited, according to the study regulations
Suren Chilingaryan, IPE, Phone: +49 721 / 608 26579 (email@example.com)
Andreas Kopmann, IPE, Phone: +49 721 / 608 24910 (firstname.lastname@example.org)
Please apply online using the button below for this vacancy number IPE 10-19.
Personnel Support is provided by
Telefon: +49 721 608-25184,
Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany
We prefer to balance the number of female and male employees. Therefore we kindly ask female applicants to apply for this job.
If qualified, handicapped applicants will be preferred.