Two big issues in the scientific community at the moment are transparency and reproducibility of research results. Workflow systems are a part of the solution of these problems when it comes to orchestrated computing. At the eScience Center we use containers and workflow languages such as CWL to make tools from different research groups work together. To document what is happening in these workflows and to show which tools can be combined, we use semantics to describe the data and processing steps. We apply workflows in a wide spectrum of research fields, ranging from bioinformatics to linguistics.
Modeling the complex behavior of dynamic systems like fluids or plasmas is the core of many fields of research, ranging from large eddy simulation in oceanography to electron transport in batteries, and from modeling protein-ligand docking to fluid dynamics in cooling water. At the eScience Center we work on speeding up specific algorithms as well as improving general purpose software such as computational fluid dynamics solvers.
The applicability of eScience techniques is often limited by their performance. A common example is weather forecasting, where results should not only be accurate, but also quick to keep ahead of the actual weather. At the eScience Center we commonly design and improve scientific code for efficiency, depending on what is necessary to achieve scientific results. Sometimes this means aiming for low latency, and sometimes it means improving throughput or lowering power usage.
At the eScience Center we apply many kinds of distributed computing, from clouds to supercomputers, depending on what kind of compute resources are necessary and how the problem can be split up into parts. We work together in close cooperation with SURFsara, and various other cloud providers to provide dependable and scalable solutions. We apply distributed computing to many research topics, ranging from climate simulations to machine reading for news summaries to the simulation of star clusters.
For some problems there are dedicated hardware solutions such as Graphics Processing Units (GPU) and Field Programmable Gate Arrays (FPGA). Writing the code that makes most of the hardware is an art. At the eScience Center we optimize code for various goals (e.g. speed or low power use), and we make tools that can help you tune your CUDA or OpenCL code semi-automatically.