0 views collected around this technical thread.
This article details 360 Cloud Platform's practical implementation of GPU passthrough and Docker‑MIG solutions, covering the underlying principles, host and OpenStack configuration steps, verification methods, and future directions for full GPU virtualization.
Baidu Intelligent Cloud’s GPU Container Virtualization 2.0 combines user‑mode and kernel‑mode isolation in a dual‑engine design that unifies scheduling of AI compute, rendering and encoding, supports mixed deployment and multi‑scheduler integration, and boosts GPU utilization across inference, offline tasks, autonomous‑driving simulation, and cloud‑gaming workloads.
This article introduces Alibaba Cloud's PAI‑DLC, a cloud‑native deep learning platform that integrates machine‑learning capabilities, containerized services, AI‑aware scheduling, GPU virtualization, elastic training with EasyScale, data access, and observability, and discusses its architecture, key features, and future directions.
Baidu Intelligent Cloud’s new Cloud‑native AI 2.0 platform tackles AI engineering bottlenecks by offering hybrid‑parallel large‑model training, flexible GPU virtualization, and an AI Accelerate Kit that boosts training efficiency over 50 % and cuts inference latency up to 63 %, raising GPU utilization from ~13 % to about 50 %.
The article surveys GPU container virtualization for AI heterogeneous computing, detailing utilization challenges, historical architectures, various virtualization methods, Baidu's dual-engine user- and kernel-space design with isolation and scheduling features, performance benefits, best‑practice scenarios, and deployment guidance, concluding with a technical Q&A.
This article presents a comprehensive overview of Beike's machine learning platform and its inference service, detailing the platform's architecture, GPU virtualization, cloud‑native migration, MLOps implementation, and future roadmap to achieve cost‑effective, automated AI model deployment at scale.
The article presents a GPU virtual‑sharing solution for AI inference workloads that isolates memory and compute resources via CUDA API interception, integrates with Kubernetes using the open‑source aliyun‑gpushare scheduler, and demonstrates doubled GPU utilization and minimal performance loss across multiple tests.
iQIYI created a custom GPU‑virtual‑sharing system that intercepts CUDA calls to enforce per‑container memory limits, rewrites kernel launches for compute isolation, and integrates with a Kubernetes scheduler extender, allowing multiple AI inference containers to share a single V100 with minimal overhead and more than doubling overall GPU utilization.
This article details how 58.com’s end‑to‑end deep‑learning platform was optimized for higher CPU and GPU inference performance using Intel MKL, OpenVINO, mixed TensorFlow deployment, GPU virtualization, and a Prometheus‑Grafana monitoring system, achieving a 37% reduction in GPU usage and a 146% increase in average GPU utilization.
This article provides an in‑depth technical overview of server virtualization, covering its historical evolution, CPU, memory, I/O and GPU virtualization techniques, hardware‑assisted extensions such as VT‑x/VT‑d/VT‑c, and the classification of virtualization architectures for modern cloud environments.
This article explains NVIDIA GRID vGPU virtualization, detailing how GPUs are partitioned by memory size, the supported hypervisors, the operation of virtual GPU resources, differences between full‑allocation vGPU and GPU pass‑through, licensing requirements, and performance considerations for cloud and data‑center environments.
This article provides a detailed technical overview of server virtualization, covering its historical roots, CPU, memory, I/O and GPU virtualization techniques, hardware-assisted extensions, and various hypervisor architectures, highlighting why virtualization remains essential in modern cloud computing environments.