Oluwatosin Oladayo Aramide

At a few thousand GPUs, the backend network becomes a probability problem: something is always broken and the hash is always unlucky. The fix is two identical fabrics with zero cross-links. A deep dive into dual-plane design, sized for B300 nodes with 800G per GPU.

Two elephant flows, six trials, six collisions: one uplink carried everything while its twin sat at zero, and every dashboard stayed green. The measured polarization, the four-witness diagnostic chase, and the birthday math production fabrics live with.

Your AI cluster is two networks wearing one badge. One is judged by whether users notice it. The other is judged by whether GPUs wait on it. Frontend vs backend: the traffic, the rules, and how to design the GPU fabric so compute never waits.

5,412 connections on my Kubernetes cluster versus one RDMA flow moving 59.5 GB a minute. Then three identical transfers, one degraded path, and 32 percent of the workers' time gone. AI traffic measured, not described.

Credit-based flow control, a daemon that runs the whole fabric, adapters that hand out pointers into remote RAM, and switches that do math. Every InfiniBand idea an Ethernet engineer should know, demonstrated live on a real InfiniBand fabric.

The finale: one request traced live through all eight hops the series built, the five traps ranked by hours lost, and what I would do differently next time. Plus the honest single point of failure the design still carries.

Longhorn on the workers (the storage leaves finally earn their name), 60Gi of language models surviving node death, a dozen apps behind one login, and the five-object checklist that publishes anything.

One Gateway claims the BGP-advertised VIP, a cert-manager chain makes every app green-padlock HTTPS, sixteen HTTPRoutes share one door, and split-view DNS gives the same names two honest answers.

Every chip in the modern data center exists because one bottleneck got expensive enough to deserve its own silicon. The whole zoo in one sitting: real photos, one table, and a memory trick that makes the differences stick.