Endless amounts of data are available nowadays. But how can you process, analyse and (re)use them safely and securely? The projects in this Labs theme explore these aspects.
The LOFAR radio telescope makes it possible to measure changes in the universe quickly, but so far this is not possible in real time. As a result, we are missing a lot of information. In the PADRE project, we are changing this by speeding up the processing of signals with GPU cards.
Why are we doing this project?
Observing changes in the Universe with radio telescopes
The Universe is full of objects such as stars, black holes and gas clouds that are constantly changing. However, these changes happen so slowly that the Universe appears static to us. With radio telescopes such as LOFAR, we see light at frequencies that we cannot see with our eyes. At these frequencies, we can see black holes and neutron stars, for example, which do show rapid changes due to outbursts. By measuring these changes and linking them to models, we can gain a better understanding of how these objects are formed and evolve.
PADRE: detecting changes in real-time
LOFAR is the world's largest radio telescope with hundreds of thousands of sensors spread across Europe. The sensors in combination with the AARTFAAC instrument make it possible to quickly measure changes in the universe. So far, we have been able to detect these changes quickly, but not in real time. As a result, we cannot follow these objects quickly with other telescopes and we miss important information. The PADRE project (PetaFlop AARTFAAC Data Reduction Engine) will change this by accelerating the entire AARTFAAC signal processing chain with graphics cards (GPUs).
What are the main activities?
- We are building knowledge on how to create a real-time processing workflow/pipeline for LOFAR's AARTFAAC instrument using GPU-based data processing. CPUs have a few dozen complex cores that can perform complex tasks very well. GPUs, on the other hand, have thousands of simple cores and can therefore perform many simple tasks simultaneously. By making smart use of CPUs and GPUs, we can speed up data processing enormously.
- SURF, together with ASTRON and the Netherlands eScience Center, is building the AARTFAAC real-time pipeline and we are going to implement it.
- We are working on converting (porting) existing CPU-only code to GPU-specific code. While porting the code, we also make sure that the code is as reusable as possible in other data-intensive projects that want to accelerate their workflows.
Who are we collaborating with?
In this project we work together with ASTRON and the Netherlands eScience Center.
You can find more information about this project at servicedesk.surfsara.nl, or mail to Raymond Oonk at firstname.lastname@example.org.