One fabric link goes down and comes back once an hour. What does that do to a 1,024-GPU training job, and how would you find it?

A collective stops when one participant stops, and a job of a thousand ranks has no way to distinguish a link that will come back in two seconds from a rank that has died. What the job actually does during those seconds, the cost per flap, and why the fault passes every test you would run.

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.

Foundational

🔌 Networking & Storage

Debugging a Slow All-ReduceA training job reports its all-reduce at a third of what the fabric should deliver, every node passed its health check, and nothing is logged. This page is the isolation order that finds the cause in an hour instead of a day: measure the collective in isolation, split the job until the slow pair or rank appears, then check the specific things that make a link, a node or a placement slow. Most cases end at one NIC, one topology mismatch, or GPUDirect silently off.

Advanced

💻 Coding for Infra🔒 Premium

Batching Queues and BackpressureWrite a request batcher is the coding round's version of the serving engine's scheduler: requests arrive one at a time, the GPU wants them in groups, and the batcher decides when a group is full enough to send without holding anyone too long or accepting more than it can hold. The two knobs are the maximum batch size and the maximum wait, the invariant is a bounded queue, and the follow-ups (priorities, cost-aware batching, cancellation, bounded in-flight batches) are the ideas the real engines carry. This page implements the batcher in asyncio, derives what each knob buys, and walks the follow-ups.

Foundational

🗂️ Scheduling & Orchestration

Node Lifecycle: Drain, Upgrade and ReturnA node moves through a fixed cycle between provisioning and decommissioning, and most fleet operations are one lap around it: cordon so nothing new lands, drain so running work finishes or moves, act, validate, then return to the pool. The wall-clock cost of a fleet-wide change is dominated by draining rather than by the change itself, which makes the plan a scheduling document rather than a technical one.

Advanced

🩺 Fleet Reliability & Observability🔒 Premium

Stragglers and HangsSynchronous training runs at the speed of its slowest rank, so one GPU that is 30% slow makes a thousand GPUs 30% slow, and one rank that never arrives at a collective makes the other 1,023 wait in silence until a watchdog fires ten minutes later. Finding the slow rank and the stuck rank is the most common on-call task on a training fleet, and the tooling for it (per-rank timing, the NCCL flight recorder, stack dumps across ranks) is specific and learnable. This page derives the straggler tax from first principles, lists the causes in the order they actually occur, and gives the procedure for a hang.

Explore the full curriculum →

UP NEXT ON YOUR JOURNEY

Next in this trackWhy does object storage behave badly for random reads, and what does a training pipeline do about it?Next in this trackDeduplicate and quality-filter a multi-petabyte web corpus. What does that pipeline cost and where does it bottleneck?Next in this trackA 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?

FEDITOR'S NOTE

Scored on knowing a collective blocks all ranks rather than degrading, on computing the cost per flap against the timeout policy, and on naming continuous counters as the only instrument that finds an intermittent fault.

DISCUSSION · 0

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