%% You should probably cite draft-calabria-bmwg-ai-fabric-inference-bench-04 instead of this revision.
@techreport{calabria-bmwg-ai-fabric-inference-bench-03,
    number =    {draft-calabria-bmwg-ai-fabric-inference-bench-03},
    type =      {Internet-Draft},
    institution =   {Internet Engineering Task Force},
    publisher = {Internet Engineering Task Force},
    note =      {Work in Progress},
    url =       {https://datatracker.ietf.org/doc/draft-calabria-bmwg-ai-fabric-inference-bench/03/},
    author =    {Fernando Calabria and Carlos Pignataro and Qin Wu and Giuseppe Fioccola and Sowjanya Reddy},
    title =     {{Benchmarking Methodology for AI Inference Serving Network Fabrics}},
    pagetotal = 46,
    year =      ,
    month =     ,
    day =       ,
    abstract =  {This document defines benchmarking terminology, methodologies, and Key Performance Indicators (KPIs) for evaluating Ethernet-based AI inference serving network fabrics. As Large Language Model (LLM) inference deployments scale to disaggregated prefill/decode architectures spanning hundreds or thousands of accelerators (GPUs/ XPUs), the interconnect fabric determines Time to First Token (TTFT), Inter-Token Latency (ITL), and aggregate throughput in tokens per second (TPS). This document establishes vendor-independent, reproducible test procedures for benchmarking fabric-level performance under realistic AI inference workloads. Coverage includes RDMA-based KV cache transfer between disaggregated prefill and decode workers, Mixture-of-Experts (MoE) expert parallelism AllToAll communication, request routing and load balancing for inference serving, congestion management under bursty inference traffic patterns, and scale/soak testing. The methodology enables direct comparison across NIC transport stacks (RoCEv2 and UET) and fabric architectures. This document is a companion to the AI training fabric benchmarking methodology, which addresses training workloads.},
}