Protocol
RDMA Verbs
InfiniBand verbs · RDMA API
Kernel-bypass networking model: queue pairs, memory regions, and work requests move data between registered buffers with minimal CPU. Completions signal success; transport may be InfiniBand, RoCE, or iWARP.
Layers
Tags
verbsqpmrcompletionzero-copy
Sources
- InfiniBand Architecture Spec (verbs)
- rdma-core / libibverbs documentation
Related protocols
RoCEv2
Lab readyRDMA over UDP/IPv4/IPv6 on Ethernet: InfiniBand BTH follows the UDP header (port 4791). Routable across L3 with ECN/PFC/DCQCN for congest…
RoCE v1
Lab readyRDMA over Converged Ethernet version 1 encapsulates InfiniBand verbs in Ethernet frames without IP/UDP headers. It is an L2-only design l…
InfiniBand
Lab readyLossless switched fabric with its own L2/L3 (LID/GID), subnet manager, and verbs transport (RC/UD/…). Credits, VLs, and partition keys pr…
iWARP
Lab readyRDMA over TCP/IP using MPA framing with DDP and RDMAP layers. It traverses lossy networks better than RoCE but needs different NIC offloa…
NVMe over Fabrics
Lab readyNVMe over Fabrics extends NVMe queues across RDMA, TCP, or Fibre Channel. Capsules map submission/completion queue entries over the fabri…
GPUDirect RDMA
Lab readyNIC DMA path that reads and writes GPU memory directly, bypassing host bounce buffers for RDMA verbs. Requires peer-to-peer BAR mapping, …