AI Data Center Systems

Study wiki

AI infrastructure notes for the places where models meet machines.

Networking, inference, training, storage, and performance engineering notes organized as connected study tracks.

Start with AI fabric Explore inference

AI Data Center Network RDMA, InfiniBand, RoCE, Clos fabrics, telemetry, and congestion control. Efficient LLM Inference Systems KV cache, batching, quantization, GPU profiling, and serving trade-offs. AI Systems Performance Engineering GPU hardware, OS and container tuning, CUDA, PyTorch, and distributed communication. Deep Learning for Network Engineers Training fundamentals, parallelism, collectives, RDMA, RoCE, and NCCL. CME295 Lecture Notes Transformer and LLM lecture notes with practical systems annotations. Training MLPerf Training, distributed training workloads, LLMs, MoE, and LoRA. Storage AI workload storage, ZFS, MLPerf Storage, and checkpoint data paths.

Labs and Talks

Clos Fabric Lab Series InfiniBand Packet Analysis RDMA Read/Write Examples Making DGX B200 RDMA-ready