Project

Multicore Resource Allocation

Dynamic cache and memory-bandwidth allocation that pairs with schedulers to keep multicore real-time systems predictable—even as workloads, phases, and modes change.

Overview

Why this matters

Multicore platforms multiplex latency-critical, throughput-oriented, and best-effort workloads on shared caches and memory controllers. Static isolation leaves performance on the table and cannot handle shifting phases, mixed modes, or new tasks.

This project builds dynamic allocation policies that measure phase behavior, reassign cache capacity and bandwidth in real time, and co-design those decisions with the scheduler. The result: lower tail latency, higher schedulability, and predictable mode changes without sacrificing analyzability.

Documentation

Frameworks and artifacts from the dissertation.

DNA / DADNA

Phase-aware profiling and runtime reallocation of cache and bandwidth to speed up soft real-time tasks and cut deadline misses.

Omni

Mode-aware allocations and transition protocol that keep multicore systems schedulable while modes change.

DECNTR

Control and resource co-design: choose controller variants and allocations jointly to maintain safety and robustness.

RASCO

Co-design of DAG scheduling and per-phase budgets to meet end-to-end deadlines on shared hardware.

Runtime & Platforms

PREEMPT-RT and LITMUSRT prototypes, allocation hooks, and measurement tooling.

Tutorials

Step-by-step guides for setup, profiling phases, and running experiments.

Open tutorials

Next steps

Fill in benchmarks, traces, algorithms, and configuration guides as documentation grows.