The German funded project ARAMiS included work on several demonstrators one of which was a multicore approach on large scale software integration (LSSI) for the automotive domain. Here BMW and Audi intentionally implemented two different integration platforms to gain both experience and real life data on a Hypervisor based concept on one side as well as using only native AUTOSAR-based methods on the other side for later comparison.
The idea was to obtain figures on the added overhead both for multicore as well as safety, based on practical work and close-to-production implementations. During implementation and evaluation on one hand there were a lot of valuable lessons learned about multicore in conjunction with safety.
On the other hand valuable information was gathered to make it finally possible to set up a cost model for estimation of potential overhead generated by different integration approaches for safety related software functions. This model is intended as an aid for developers to decide for the most appropriate and efficient solution out of the two shown approaches at an early point of their development.
It shall provide more specific figures taking into account the particular aspects of the final aim of a planned integration project. Here incorporation of the results for the overhead of switching between partitions, isolation of resources and system monitoring helps us to attain those figures as well as the application of statistical models for occurrence based on Audi's practical measurements done on a close-to-production implementation.
This paper will describe the findings and experiences during implementation of the demonstrator platforms as well as the cost model, its usage and how it could help on development decisions for software projects aiming on integration of safety related automotive functions.