An Essential Relationship between Real-time and Resource Partitioning
summary
Presents a kernel-level resource-partitioning approach to run real-time and general-purpose processes on the same CPU without virtualization by assigning concrete resources: core pinning, memory-bandwidth partitioning, execution-time slices, and device access controls. Describes a prototype implementation with scheduler and resource-controller changes, measurements showing bounded worst-case response times, effective isolation across CPU-bound, I/O-bound and mixed workloads, and an implementation overhead around 0.5%.
AI-written summary · about
chapters (6)
- 0:00Real-time and partitioning
- 5:38Experimental system design
- 14:40Real-time system challenges
- 23:55Resource management performance
- 34:02Summary and outlook
- 39:48Q&A
related talks
20mGet More Computing Power by Helping the OS Scheduler
30mIntroduction and Consideration of Temporal Partitioning in Avionics With Open Source Eco-System
52mEffectively Measure and Reduce Kernel Latencies for Real-time Constraints
40mCase Study for Resource Control in Embedded Linux Container Integration
29mSecure and Dynamic Hardware Partitioning Management on Heterogeneous SoC
44mReal-Time Linux on Embedded Multicore Processors
44mReducing cpu jitter for deterministic performance
53mCPU Isolation by Memory Bandwidth