Type of Work:
Master or Diploma Thesis
In this work a novel checkpointing mechanism for discrete event simulations should be implemented. The idea is to implement a new syscall in the operating system (OS) to enable a fast interaction between the OS and the simulation kernel.
- Simulation background (e.g. SystemC)
- Operating systems knowledge (e.g. Linux, FreeBSD, etc)
- Programming skills in C/C++
Checkpointing is used to accelerate discrete event simulations using the temporal decoupling technique, where the simulation processes are allowed to run ahead of the global simulation time. This optimistic simulation can lead to simulation errors, that could be corrected by returning to a previous point in time and repeat the simulation again with a tighter timing synchronization. In a previous work, we used the syscall fork() in order to create these simulation checkpoints (http://dx.doi.org/10.23919/DATE.2019.8714823). However, as shown in this work, fork() implies a large timing overhead, such that such checkpoints cannot be created with a high frequency.