Skip to content
RoCE course / Module 1

RoCE and RDMA for AI fabrics

0%

0 of 6 cases closed

Foundations

Explain the verbs object model, pick the right transport, and read a RoCEv2 packet and a GID table well enough to say which version and which priority a connection is actually using.

  1. 1.1 The verbs object model: device, PD, QP, CQ, MRThe card that says InfiniBand on an Ethernet cable
  2. 1.2 Transports and opcodes: what RC, UC and UD can actually doThe one requirement that changed the transport
  3. 1.3 RoCEv1 vs RoCEv2: the encapsulationWorks in the rack, dies between racks
  4. 1.4 GIDs: how a connection picks its RoCE versionThe benchmark crosses the subnet and the application does not
  5. 1.5 QP state machine, retries, and connection setupForty minutes, every time
  6. 1.6 Multipath RC: a new transport in the annex, not in the APIThe annex he cannot read
  7. Checkpoint