XDP/BPF in RDMA - RXE Implementation
Explore XDP/BPF integration in RDMA(RXE) for enhanced network performance and programmability. Learn about cutting-edge techniques in Linux networking and data processing.
Scale and Accelerate Distributed Model Training in Kubernetes Clusters
Orchestrate and accelerate distributed deep learning workloads across multiple GPUs and nodes using Kubernetes, Kubeflow PytorchJob, RDMA, RoCE, and SRIOV for near-linear performance scaling.
RDMA Programming Using Rust Async
使用Rust异步框架实现RDMA编程,构建Rust友好的RDMA绑定,提供自动内存管理和异步I/O,大大简化RDMA应用开发,促进Rust社区利用RDMA构建高性能应用。
ML Training Acceleration with Heterogeneous Resources in ByteDance
探讨如何利用字节跳动的异构计算资源加速机器学习模型训练,包括GPU共享机制、NUMA亲和性资源分配和RDMA网络通信等技术。
QuarkContainers: High Performance Secure Container for Serverless
高性能安全容器QuarkContainers解决无服务器计算挑战:强隔离、低开销、快速启动、高性能。通过应用内核、IO优化、休眠加速和Kuasar集成实现突破性改进。
ACCL+: FPGA-Based Collective Engine for Distributed Applications
Explore ACCL+, an open-source FPGA-based collective communication library for distributed applications, offering portability, multiple protocol support, and user extensibility for enhanced performance.
Deep Dive into GPU Support in Apache Spark 3.x - Accelerator-Aware Scheduling and RAPIDS Plugin
Explore GPU acceleration in Apache Spark 3.x, covering task scheduling, columnar processing, and the RAPIDS plugin for enhanced ETL, ML, and DL performance without code changes.
Direct Drive - Azure's Next-Generation Block Storage Architecture and Ultra Disk Implementation
Dive into Azure's next-generation block storage architecture, exploring Direct Drive's innovative design, Ultra Disk capabilities, and solutions for cloud-scale storage challenges.
Building a High Performance Network in the Public Cloud Using RDMA - First Principles
Explore Oracle Cloud Infrastructure's high-performance networking using RDMA, covering challenges, optimizations, and unique features for low-latency workloads in public cloud environments.
Enhancing Network Performance with RDMA on Microsoft Azure Network Adapter
Discover how to optimize network performance using RDMA technology on Azure Network Adapter, covering deployment strategies, security protocols, and real-world applications for improved throughput and reduced latency.
Persistent Write Log Cache with RDMA Replication in Ceph
Dive into implementing Persistent Write Log Cache with replication in Ceph, exploring client-side PMEM storage, RDMA technology, and efficient write-back caching for distributed systems.
Satisfying AI's Demanding Data Needs Using Infrastructure as Code
Discover how to optimize AI infrastructure deployment using OpenStack, Terraform, and Ansible for efficient IPU-POD management, focusing on maximizing data processing and enabling flexible resource allocation.
Accessing Files Remotely with Linux - Recent Progress with SMB3.1.1 Client and Servers
Explore recent advancements in Linux's SMB3.1.1 protocol implementation, including improved performance, security features, and enhanced remote file access capabilities across various server platforms.
Ethernet Evolution: Powering AI's Future Through Ultra Ethernet Consortium
Discover how the Ultra Ethernet Consortium is revolutionizing network capabilities for AI and HPC, exploring technical innovations, enhanced performance metrics, and advanced RDMA solutions for next-gen computing.
Database Consolidation With Persistent Memory
Discover how Exadata leverages Intel PMEM and RoCE 100Gbs infrastructure to accelerate OLTP and OLAP workloads, optimizing database performance and consolidation.
LoLKV - Logless, Linearizable, RDMA-based Key-Value Storage System
Innovative key-value storage system using lock-free design for concurrent updates, achieving higher throughput and lower latency than state-of-the-art alternatives.
Reverie: Low Pass Filter-Based Switch Buffer Sharing for Datacenters with RDMA and TCP Traffic
Innovative buffer-sharing scheme for datacenters, improving RDMA and TCP performance through unified admission control and low-pass filtering, enhancing burst absorption and flow completion times.
DINT - Fast In-Kernel Distributed Transactions with eBPF
Explore DINT, a high-performance distributed transaction system using eBPF for kernel-level optimization, offering improved throughput and latency while maintaining security and debuggability.
Harmonic - Hardware-assisted RDMA Performance Isolation for Public Clouds
Innovative hardware-assisted solution for RDMA performance isolation in multi-tenant clouds, enhancing resource sharing and preventing interference between tenants.
TeRM - Extending RDMA-Attached Memory with SSD
Innovative approach to extend RDMA-attached memory with SSD, enhancing storage capacity while maintaining high performance. Introduces techniques to optimize CPU and network resource usage in RDMA-based systems.
The World of 100G Networking
Explore 100G networking technologies, their capabilities, and challenges in high-speed communication. Learn about competing solutions and potential improvements for efficient data transfer.