# Modernizing RedoxOS Scheduling: EEVDF & Performance Analysis


## Personal Information

-   **Name:** Akshit Gaur
-   **Matrix/GitLab:** akshitgaur2005
-   **Website:** [himwant.org](https://himwant.org)


# 1. Project Goals

The goal of this project is to modernize the Redox OS process scheduling subsystem and establish a rigorous performance testing baseline. This involves three key objectives:

1.  **Benchmarking:** Establishing a “Source of Truth” for system performance to measure future improvements (specifically for Direct Context Switching).
2.  **Simulation & Correctness:** Developing a **Scheduler Test Harness** to verify scheduling logic in isolation under heavy load (1-1024 cores).
3.  **Kernel Scheduling:** Implementing a **Priority-Based Scheduler** followed by the **EEVDF (Earliest Eligible Virtual Deadline First)** algorithm to replace the legacy Round-Robin one.


# 2. The Problem

-   **Performance Visibility:** Optimizing Redox is difficult because we lack a standardized suite to isolate/measure syscall latencies and context-switch overheads.
-   **Testing Complexity:** Testing scheduler changes currently requires full system recompilation and booting, making iteration slow and regression testing difficult.
-   **Scheduling Latency:** The current Round-Robin scheduler does not handle interactive workloads or “lag” as effectively as modern algorithms like EEVDF.


# 3. Implementation Roadmap & Milestones


## Phase 1: Foundation & Prototype (Total: $1,800) (Target: Feb 15 - March 31)


### Milestone 1.1: Benchmarking & Profiling ($500) (Target: Feb 28)

I will build/port a benchmarking suite tailored for Redox.

-   **Deliverable:** Porting tools (inspired by `lmbench` / `UnixBench`) and writing test harness tools required by `4lDO2` for testing Direct Context Switching.
-   **Status:** Port of `sysbench` is already complete for `x86_64` and `aarch64`.


### Milestone 1.2: Priority-Ranked Scheduler ($800) (Target: March 20)

I will implement a simple priority-based scheduler integrated into Redox.

-   **Mechanism:** Nice value determines priority; context switch selects highest priority task.
-   **Goal:** Demonstrate basic operation and correctness of the scheduling API changes.


### Milestone 1.3: Isolated Test Harness ($500) (Target: March 31)

I will implement a non-Redox program to test the scheduler algorithms in isolation.

-   **Scope:** Simulation of tasks (continuous and blocked) under heavy load.
-   **Scale:** Verifying correctness on 1 to 1024 simulated cores.


## Phase 2: EEVDF Implementation & Tuning (Total: $3,700) (Target: April 1 - August 31)


### Milestone 2.1: EEVDF Implementation ($2,000) (Target: May 30)

I will implement the EEVDF algorithm within the kernel.

-   **Logic:** Replacing legacy Round Robin with the modern EEVDF.
-   **Deliverable:** A functioning EEVDF scheduler demonstrating basic operation.


### Milestone 2.2: Optimization ($700) (Target: June 30)

Improve the scheduler performance to match or exceed the current scheduler’s raw throughput using the Test Harness.


### Milestone 2.3: Analysis & Tuning ($500) (Target: July 31)

Analyze and tune the performance of the system with EEVDF, applying lessons from Linux and using the Benchmarking suite (Phase 1) to validate gains.


### Milestone 2.4: Documentation ($500) (Target: August 31)

Produce a comprehensive report/news post covering an overview of EEVDF scheduling theory (referencing the original paper), the Redox implementation, parameter tuning results, and how-to instructions for users to configure scheduling priorities themselves.

**Note**: The end date for final claims for payment will be October 31, 2026.


# 4. About Me

I am a Systems Engineering undergraduate (JNU) focused on Operating Systems, Rust, and ML Infrastructure.

-   **Redox Experience:** I have already contributed multiple PRs to the Redox kernel, base, relibc and recipes (including `sysbench` port).
-   **Relevant Skills:** Rust, C, OS Internals (xv6 analysis), RISC-V Assembly.

**Note:** All code submitted will be written by hand without the use of AI code generation tools.