Tagged articles
GPU clusters
16 articles · Page 1 of 1
Tsinghua University's BaGuaLu-2 system dynamically selects per-operator recomputation or offloading after MoE gating to keep activation memory within device limits, achieving 1.14× throughput over baselines on 12,288 GPUs with 32.9% average MFU, zero config changes, and open-source code.
BaGuaLuGPU clustersMoE training
0 likes · 11 min read

At CNCC 2026, Sugon unveiled Fabricview, a platform that centrally manages millions of network devices and uses AI-driven analytics to detect and resolve link-level faults before they stall large-scale model training.
AI infrastructureFabricviewGPU clusters
0 likes · 6 min read

This article details Alibaba Cloud ACK's fault self-healing system for heterogeneous compute clusters, covering its Kubernetes-native architecture, multi-layer detection and event conversion, orchestrated repair workflows with stepwise authorization, rate limiting and circuit breaking, full observability, and practical deployment strategies to maximize effective compute throughput.
ACKGPU clustersKubernetes
0 likes · 64 min read


SuanNiMay 8, 2026 · Artificial Intelligence
OpenAI, together with AMD, Broadcom, Intel, Microsoft and Nvidia, introduced the Multipath Reliable Connection (MRC) protocol, which splits a single 800 Gb/s link into eight 100 Gb/s planes, enabling full‑mesh connectivity for over 100 k GPUs with fewer switches, lower cost, higher resilience, and dynamic load‑balancing that eliminates congestion and hardware‑failure impacts during large‑scale AI training.
AI networkingGPU clustersMRC
0 likes · 12 min read

The article analyzes how AI supernodes, massive GPU clusters, knowledge‑base activation, embodied intelligence, optical interconnect and open‑source agents like OpenClaw together form a complete AI industry ecosystem in 2025, highlighting performance breakthroughs, domestic competition, market share shifts, and emerging security concerns.
AI supernodesGPU clustersembodied intelligence
0 likes · 16 min read

This article details Baidu Intelligent Cloud's journey in designing, constructing, and operating a 32,000‑card hybrid AI compute cluster, covering challenges in power, cooling, networking, multi‑cluster scheduling, and security, and explains how innovative hardware, software, and operational strategies achieved over 50% MFU improvement and industry‑first performance records.
AI infrastructureGPU clustershybrid cloud
0 likes · 15 min read

The article explains how AI large‑model training demands massive GPU resources and how carefully designed network architectures—such as Clos/Fat‑Tree, Spine‑Leaf, multi‑rail versus single‑rail connections, Dragonfly, and Torus—impact performance, scalability, cost, and reliability, guiding the selection of optimal data‑center networks.
AIGPU clustersHigh-Performance Computing
0 likes · 9 min read

This article summarizes the challenges of distributed training for massive language models and presents a suite of solutions—including DP/TP/PP overlap, context parallelism, efficient recomputation, and a performance‑aware cost model—that together boost training throughput by over 30% on large GPU clusters.
GPU clustersPerformance Modelingactivation rematerialization
0 likes · 27 min read

This article analyzes the technical roadmap for upgrading AI super‑large GPU clusters to support trillion‑parameter multimodal models, covering single‑chip performance, super‑node scaling, DPU‑based compute fusion, energy‑efficient designs, converged storage, high‑throughput networking, and fault‑tolerant checkpoint strategies.
AI computeDPUGPU clusters
0 likes · 18 min read

The article surveys the network architectures and congestion‑control techniques used in massive GPU clusters—such as Byte’s megascale, Baidu HPN, Alibaba HPN7, and Tencent Xingmai 2.0—highlighting how high‑bandwidth, low‑latency designs and advanced RDMA technologies enable training of trillion‑parameter multimodal AI models.
GPU clustersHPNInfiniBand
0 likes · 11 min read

The article analyzes three major high‑performance‑computing network topologies—Fat‑Tree, Dragonfly, and Torus—detailing their design principles, scalability formulas, routing strategies, advantages, and limitations to help architects choose the most suitable architecture for large‑scale GPU clusters.
DragonflyFat-TreeGPU clusters
0 likes · 13 min read

This article analyzes traditional cloud data center network limitations for AI workloads and compares various high‑bandwidth, low‑latency architectures—including two‑layer and three‑layer fat‑tree designs, InfiniBand, and RoCE—providing best‑practice recommendations for building scalable, non‑blocking AI‑Pool networks.
AI computingFat-TreeGPU clusters
0 likes · 12 min read

The article analyzes ByteDance and Peking University's MegaScale system that enables efficient, stable training of large language models on clusters exceeding ten thousand GPUs, detailing algorithmic tweaks, 3D parallel communication overlap, operator optimizations, data‑pipeline improvements, network tuning, and fault‑tolerance mechanisms that together achieve a 55.2% MFU on a 175B model.
GPU clustersLLM trainingMegaScale
0 likes · 15 min read

This article explains how Baidu built a massive, high‑performance GPU/IB cluster, optimized its architecture and software stack, and integrated AI frameworks and resource management to overcome compute, memory, and communication bottlenecks, enabling efficient training of trillion‑parameter large models.
AI infrastructureGPU clusterscloud computing
0 likes · 19 min read

Tencent’s Star Network delivers a 1.6 Tbps Ethernet‑RDMA fabric, fat‑tree topology supporting up to 4 K GPUs, multi‑track traffic aggregation and adaptive heterogeneous links plus a custom TCCL library, cutting AllReduce overhead from 35 % to 3.7 %, speeding AI training iterations by 32 % while automating deployment and providing sub‑second self‑healing.
AI TrainingGPU clustersRDMA
0 likes · 19 min read

The article explains how the massive compute and storage demands of today’s large language models create a “compute wall” and “storage wall,” and describes Baidu Intelligent Cloud’s four‑layer full‑stack infrastructure—combining advanced parallelism techniques, optimized GPU networking, static‑graph compilation, and cost‑model‑driven placement—to train trillion‑parameter models efficiently.
AI infrastructureGPU clusterscost-model
0 likes · 27 min read
