SCCS Colloquium - Aug 19, 2020

From Sccswiki
Revision as of 17:57, 3 August 2020 by Makis (talk | contribs) (Created page with "{| class="wikitable" |- | '''Date:''' || Aug 19, 2020 |- | '''Room:''' || [ Online] (password: SCCS)...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
Date: Aug 19, 2020
Room: Online (password: SCCS)
Time: 15:00 - 16:00

Michael Zintl: Design and Implementation of a Modular Primary Flight Display for a Flight Simulator

Master's thesis submission talk. Michael is advised by Pranav Nagarajan and Julian Rhein and examined by Prof. Hans-Joachim Bungartz.

Cockpit instrumentation is often limited to a specific aircraft. The goal of this master's thesis was the development of a Primary Flight Display (PFD) that is as modular as possible. This allows any researcher to dynamically adapt the PFD to the respective research project and to display only the information that is relevant for the experiment and the aircraft type. The requirements for the PFD were defined through common standards and interviews. The applied implementation approach ensures maintainability and reduces the amount of necessary modifications for the integration into other projects. In the end, the display was evaluated through simulator flight testing and subsequent interviews.

Keywords: 2D rendering, OpenGL, C++, Qt, HMI, Cockpit, Flight Simulator

Alexander Hölzl: Integrating TeaMPI with ULFM for Hard Failure Tolerance in Simulation Software

Bachelor's thesis submission talk. Alexander is advised by Philipp Samfass.

With ever growing system complexity the mean time between failure of HPC systems is getting shorter. The conventional solution to failures in those kind of systems is checkpointing, but checkpointing does not scale well with regards to an increase in system scale. One of the solutions to this problem is replication, fault tolerance through redundant computation.

Transparent group replication with MPI has been implemented with teaMPI. TeaMPI alleviates some of the performance issues caused by replication by dividing the workload into different tasks and sharing the outcomes of the tasks between the groups/teams. Although teaMPI achieves replication it does not yet offer actual fault tolerance and will crash if a failure of a process occurs. Implementing fault tolerance in teaMPI was the goal of my thesis. As standard MPI does not offer needed fault tolerance functionality a modified version of OpenMPI, ULFM is used. ULFM offers a set of features that allows the programmer to respond to failures of MPI processes. The implementation has been tested using a simple solver for shallow water equations, SWE, and compared against an implementation using traditional checkpoint/restart.

Keywords: teaMPI, ULFM, Fault Tolerance, HPC