CCIE #69931 · AI infrastructure networking
Networks that keep GPUs busy.
Practical writing, labs and a book on the networks behind AI training and inference: RoCEv2 and InfiniBand back-ends, EVPN front-ends and storage, explained with the rigor of latency-critical infrastructure.
What you'll find here
AI fabric design
Topologies, oversubscription and sizing for training and inference, with every decision traced to a requirement.
Lossless RDMA
RoCEv2, PFC, ECN and congestion control, InfiniBand concepts, and the counters that show whether it all works.
Calculators
ECMP collision odds, incast queue growth and end-to-end oversubscription: the numbers that decide designs, in your browser.
Latest articles
- Two exits, one idle link: a hands-on look at BGP ADD-PATH
A ten-minute FRR lab: behind a route reflector, ADD-PATH and multipath-relax do nothing alone and fix an idle link together. Packet captures and a failover myth, tested.
Oct 22, 2026 - One link fails, every job slows: what ECMP hashing really does
Plain modulo hashing reshuffles almost every flow when one link fails. Why that hurts AI fabrics, and what resilient, symmetric and weighted hashing change.
Oct 17, 2026 - Decision: one spine ASN or many?
A short design decision for eBGP fabrics. Why a shared spine ASN turns BGP loop prevention into protection against valley paths after a failure.
Oct 17, 2026
Building the AI Factory
Eighteen chapters on designing, configuring and operating the networks behind AI training and inference, from GPU servers and collectives to lossless fabrics, load balancing, Ultra Ethernet, scale-up networks and three complete builds.