Part 4 of a series on RDMA in Kubernetes pods for GPU training. DRANET gives a pod an IPvlan child of each rail it claims. A claim can also put GPUs and rails on the same PCIe root.

Part 3 of a series on RDMA in Kubernetes pods for GPU training. The pod stays on the pod network and gets the RDMA devices of the rails. It also gets an IPvlan interface on each rail.

Part 2 of a series on RDMA in Kubernetes pods for GPU training. Each pod gets its own SR-IOV virtual function on each rail, with 1 or 2 VFs per physical function.

Part 1 of a series on RDMA in Kubernetes pods for GPU training. What a hostNetwork pod needs for RDMA, the NCCL baseline on two A100 nodes, and what the extra settings cost.

An interactive walkthrough of how Dranet finds a network device, gives it to a pod, and takes it back, with links to the code of each release.