Resource-Aware Scheduling for L4Re

Mixed-criticality systems require strong safety and security guarantees such as
isolation and freedom from interference between subsystems. Further
consolidation of these diverse subsystems on shared multi-core hardware
introduces contention for shared microarchitectural resources such as caches,
interconnects and energy budgets.
In this talk, I address these challenges and explore mechanisms for temporal
isolation and resource-aware scheduling within the L4Re framework. I present a
novel approach to scheduling in L4Re that incorporates generic scheduling
constraints (SCs) as first-class microkernel objects. This abstraction allows to
specify execution environments and reservations for system resources with great
flexibility. I show that SCs incur minimal performance overhead and have the
ability to effectively bound or mitigate shared resource contention for
realistic application workloads.
The goal of this talk is to discuss the SC concept and how to further extend the
strong isolation guarantees of the L4Re design.

(Diploma Thesis defense)