03InfiniBand or RoCE version 2 for a new GPU training cluster. Make the call and say what would reverse it.▼medium★ EssentialNVIDIACrusoexAI4 repliesunlockedChoose between supported InfiniBand and RoCE designs using workload tests, congestion behavior and a complete cost model.Open full answer →
08What is incast, why does it break an RDMA fabric, and what do ECN and priority flow control each do about it?▼hardxAINVIDIAMeta4 repliesunlockedCalculate queue growth during incast, then distinguish endpoint rate feedback from link-level priority pause and shared-resource contention.Open full answer →
18What is Ultra Ethernet trying to fix, and does it change how you would choose a fabric today?▼hardBroadcomMetaAMD4 replies○ sign inExplain packet spraying through a small path-collision model, then separate transport features from measured performance and a supported fabric purchase.Open full answer →
26Several unrelated jobs slow down at once and nothing is dropping packets. How do you confirm a pause cascade and find where it started?▼hardMetaxAI4 replies◆ premiumFollow pause requests against the direction of data, estimate queue growth and distinguish cascade evidence from its possible causes.Open full answer →
