A mixture-of-experts model does an all-to-all twice per layer. What does that demand of the fabric, and what changes on a rack-scale system?

Two all-to-alls per layer across sixty layers is over a hundred collectives on the critical path of every decode step, each one small and latency-bound. The per-GPU volume worked out, the time on each link type, and why a single large NVLink domain changes the design rather than just improving it.

Updated Sep 2026 · Grounded in real AI infrastructure interview loops and written to a senior-engineer editorial bar, with every number worked and every diagram hand-built.

20 answers per topic instead of 10, plus saved progress and bookmarks · no cardor unlock all 283 remaining answers · ₹2,000 / $25

The concepts behind this question

Ranked by how closely each one overlaps this question's topic, so the first card is the thing to read if the answer above moved too fast.

Advanced

🕸️ Distributed Training🔒 Premium

Expert Parallelism for MoEA mixture-of-experts layer runs only a few of its experts per token, so the experts can be spread across GPUs and each token shipped to the ranks that hold its chosen experts. That shipping is an all-to-all in each direction, twice per layer per pass, and its cost plus the load imbalance between experts is what expert parallelism is really about.

Foundational

🔌 Networking & Storage

NCCL and Collective AlgorithmsNCCL is the library every PyTorch collective lands in, and its choice of ring or tree, channel count and protocol decides whether an all-reduce runs at fabric speed or at a third of it. Knowing what NCCL_DEBUG=INFO prints, and which environment variable changes which decision, is the difference between tuning a cluster and guessing at it.

Advanced

🧩 GPU & Accelerator Architecture🔒 Premium

NVLink, NVSwitch and PCIeInside a node, GPUs talk over NVLink at 900 GB/s per H100 through an NVSwitch fabric that gives all eight cards full bandwidth to each other; to the host and to anything outside the node they talk over PCIe at 64 GB/s or a 400 Gb/s NIC at 50 GB/s. That fifteen-fold gap is why tensor parallelism stays inside the eight-GPU domain, why NVL72 changes the serving math for MoE, and why the question "how many GPUs share an NVLink domain?" is the first thing to ask about any cluster.

Foundational

🧮 Open Weights & Serving Engines

Expert Parallel and All-to-All BackendsA mixture-of-experts model can be split two ways and the choice changes everything. Tensor parallelism shards each expert across GPUs, which keeps every GPU busy and reads every expert's shard on every token. Expert parallelism gives whole experts to whole GPUs, which reads only the selected experts but requires an all-to-all to route tokens to them and back. The all-to-all is the cost, its backend is a configuration choice matched to the interconnect, and expert load imbalance is what actually limits the result.

Explore the full curriculum →

UP NEXT ON YOUR JOURNEY

Next in this trackCheckpoints and datasets both live on storage. Why does one system sized for both usually serve neither well?Next in this trackSeveral unrelated jobs slow down at once and nothing is dropping packets. How do you confirm a pause cascade and find where it started?Next in this track8,192 GPUs exceeds what a two-tier Clos of 64-port switches supports. Compare adding a tier against using higher-radix switches.

FEDITOR'S NOTE

Scored on computing the per-GPU all-to-all volume rather than the aggregate, on recognising the collectives as latency-bound rather than bandwidth-bound, and on why a wide NVLink domain changes what is buildable.

DISCUSSION · 0

No comments yet — be the first to share your approach.