RDMA and HPC Interconnect
Move data between machines without waiting for the CPU.
5 courses20 labs
HPC, storage, and AI infrastructure engineers.
What you will be able to do
- Write verbs-based RDMA programs with queue pairs, memory regions, and completion queues.
- Design and tune a lossless RoCEv2 fabric with PFC and ECN.
- Explain and measure how MPI, UCX, and OFI use RDMA.
- Plan GPU-direct data paths and virtualized NIC access.
Courses
RDM-01≈ 30 h
RDMA Programming with Verbs
Write verbs programs for two-sided, one-sided, and atomic operations, and tune them.
9 units5 labs
RDM-02≈ 30 h
RoCEv2 and Converged Fabric Deployment
Design, configure, and troubleshoot a lossless Ethernet fabric that carries RDMA alongside TCP/IP.
7 units5 labs
RDM-03≈ 30 h
RDMA and MPI in High-Performance Computing
Explain how MPI reaches the wire through OFI and UCX, and tune a real MPI job.
9 units4 labs
RDM-04≈ 30 h
RDMA for Distributed Storage (Ceph)
Find where a storage cluster loses time to TCP, enable RDMA in Ceph, and prove the gain with measurements.
7 units4 labs
RDM-05≈ 30 h
GPUDirect and SR-IOV for HPC
Move GPU data across the cluster without a CPU copy, and keep the path fast in a multi-tenant system.
9 units4 labs