Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 32 additions & 6 deletions README.md
Original file line number Diff line number Diff line change
@@ -1,20 +1,46 @@
# ⚖️ arbiter

[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
![Status: Experimental](https://img.shields.io/badge/Status-Experimental-orange)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg "License: MIT - Open source software license")](https://opensource.org/licenses/MIT)
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

Screen readers primarily use the alt attribute (the text in the first set of brackets) rather than the title attribute (the quoted text after the URL) to describe images. To improve accessibility as intended in the PR description, consider moving the descriptive text to the alt attribute.

Suggested change
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg "License: MIT - Open source software license")](https://opensource.org/licenses/MIT)
[![License: MIT - Open source software license](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)


> [!CAUTION]
> This project is currently **Experimental**. It is not recommended for production use and is subject to significant changes as the architecture evolves.

Dual nature—combining bare-metal virtualized hardware management (aSHARD VRAM pinning) with quantum-accelerated Kubernetes scheduling.
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

There is a terminology conflict between this tagline and the 'Key Features' section. Line 40 emphasizes 'bypassing virtualization overhead' with 'direct hardware management,' which contradicts the 'virtualized hardware management' mentioned here. Clarifying whether the system manages virtualized resources or provides direct bare-metal access will improve documentation clarity.

Suggested change
Dual nature—combining bare-metal virtualized hardware management (aSHARD VRAM pinning) with quantum-accelerated Kubernetes scheduling.
Dual nature—combining direct bare-metal hardware management (aSHARD VRAM pinning) with quantum-accelerated Kubernetes scheduling.


## 📖 Overview

`arbiter` is a specialized orchestration layer designed for high-performance computing environments. It bridges the gap between low-level hardware management and cloud-native scheduling, providing a unified interface for managing virtualized resources with precision.

## 🏗️ Architecture

```mermaid
graph TD
subgraph CloudNative [Cloud Native Layer]
K8s[Kubernetes Cluster]
end

subgraph Orchestration [Orchestration Layer]
Arbiter((Arbiter Core))
end

subgraph Infrastructure [Infrastructure Layer]
BareMetal[Bare Metal Servers]
GPU[GPU Resources / VRAM]
end

K8s <--> Arbiter
Arbiter <--> BareMetal
Arbiter <--> GPU
Comment on lines +31 to +33
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

In Mermaid flowcharts, the standard syntax for a bidirectional arrow is <->. While some renderers may be lenient with <-->, using the standard syntax ensures better compatibility across different Markdown viewers and IDE plugins.

Suggested change
K8s <--> Arbiter
Arbiter <--> BareMetal
Arbiter <--> GPU
K8s <-> Arbiter
Arbiter <-> BareMetal
Arbiter <-> GPU


style Arbiter fill:#f96,stroke:#333,stroke-width:4px
```

## 🚀 Key Features

- 🏗️ **Infrastructure Awareness**: Directly manages bare-metal resources for maximum performance.
- 📍 **VRAM Optimization**: Uses aSHARD pinning to eliminate GPU memory fragmentation.
- ⚛️ **Next-Gen Scheduling**: Leverages quantum-accelerated algorithms for complex Kubernetes workloads.
- ⚖️ **Unified Orchestration**: A single control plane for both hardware and cluster-level operations.
- 🏗️ **Bare-Metal Precision**: Bypass virtualization overhead with direct hardware management for latency-sensitive AI workloads.
- 📍 **Intelligent VRAM Pinning**: Maximize GPU utilization and eliminate fragmentation using aSHARD-driven memory allocation.
- ⚛️ **Quantum-Accelerated Scheduling**: Resolve complex multi-constraint resource allocations faster than traditional heuristics.
- ⚖️ **Unified Control Plane**: Seamlessly bridge the gap between low-level hardware states and high-level Kubernetes orchestration.

## ⚖️ License

Expand Down