dynamic-resource-allocation

Tensor Fusion is a state-of-the-art GPU virtualization and pooling solution designed to optimize GPU cluster utilization to its fullest potential.

Kubernetes DRA driver for exclusive, topology-aware CPU allocation for high-performance workloads

Linux + Kubernetes

🦄 Distributed Inference on Kubernetes with DRA and MIG

Kubernetes DRA driver for deterministic CPU time-sharing using Linux SCHED_DEADLINE. Partitions cores into guaranteed, non-overlapping time slots exposed as DRA devices.

Developed as a semester project for CS-315 (Cloud Computing), this project implements a robust, end-to-end data processing pipeline designed for cloud environments. It focuses on the core cloud principle of elasticity by implementing a system that dynamically adjusts computing resources (Spark Executors) based on real-time traffic volume.

Publishes paired Bluetooth controllers as DRA devices on liken clusters

⚡ Cloud-Native AI/ML & Jupyter Lab Platform powered by Kubernetes Operator & Dynamic Resource Allocation (DRA). Automates GPU scheduling, stateful workspaces, and multi-framework ML lab environments.

Provider-neutral Kubernetes DRA observability, simulator, and diagnostics for GPU and accelerator workloads—model virtual device pools, inspect ResourceClaim/ResourceSlice state, and explain allocation outcomes without physical hardware.

Kubernetes Network (DRA) Driver orchestrating host-backed networks across physical and virtual network interfaces (Macvlan, IPVLAN, SR-IOV VF, RDMA...)

Crop-disease inference on Kubernetes: a three-tier ladder, GPU scheduling, KEDA autoscaling on queue depth rather than CPU, and warm headroom for a cold start you cannot scale away. M1 and M2 built and measured; the on-device tier runs in the browser. The measurement says the ladder fails on field photos, and the README says so.

Kubernetes DRA driver for deterministic L3 cache partitioning via resctrl (Intel RDT, AMD QoS, ARM MPAM). Partitions cache ways into non-overlapping slices exposed as DRA devices.

Anchor is a Kubernetes DRA driver that gives pods static AWS IP addresses which follow them when they are rescheduled. most common use case is SCTP/SIGTRAN in EKS and on-premises

Kubernetes x NVIDIA DRA Workshop

A library and a cli to view K8s cluster resources

🔬 6G Network Slicing with AI-Driven Dynamic Resource Management | Multi-Agent DRL (SAC) | Explainable AI (SHAP) | Energy-Aware Optimization | Dynamic Soft Slicing | MLOps Pipeline | FastAPI | SimPy | PyTorch

high-performance, resilient Python framework for building Kubernetes Dynamic Resource Allocation (DRA) hardware drivers.

Routing and control of media playback on liken clusters

Publishes monitor outputs as DRA devices on liken clusters

Publishes audio outputs as DRA devices on liken clusters