--- title: "#rdma articles | Hashnode" description: "Explore 11 articles tagged #rdma on Hashnode. Tutorials, guides, and how-tos from the community." url: "https://hashnode.com/tag/rdma" --- Tag # #rdma 11 posts · 1 followers [Nathanael FETUE](https://hashnode.com/@nathanael-fetue) · [1d ago](https://hashnode.com/posts/pairing-a-roce-request-with-its-ack-and-why-the-queue-pair-number-is-not-enough/6af44489a4d949f08425e1af) ### [Pairing a RoCE request with its ACK, and why the queue-pair number is not enough](https://pegasus-nathanael-tech.hashnode.dev/pairing-a-roce-request-with-its-ack-and-why-the-queue-pair-number-is-not-enough?utm_source=hashnode&utm_medium=feed) An RC packet carries only its destination QP, so a write and its ACK name different QPs. The key that pairs them in XDP, and the ACKs not to trust. 4 min read · [pegasus-nathanael-tech.hashnode.dev](https://pegasus-nathanael-tech.hashnode.dev/pairing-a-roce-request-with-its-ack-and-why-the-queue-pair-number-is-not-enough?utm_source=hashnode&utm_medium=feed) [Nathanael FETUE](https://hashnode.com/@nathanael-fetue) · [3d ago](https://hashnode.com/posts/busy-polling-or-events-timing-one-rdma-write-and-why-perftest-says-half/6af159d1709e0bfcfc8f45f1) ### [Busy polling or events ? Timing one RDMA write, and why perftest says half](https://pegasus-nathanael-tech.hashnode.dev/busy-polling-or-events-timing-one-rdma-write-and-why-perftest-says-half?utm_source=hashnode&utm_medium=feed) Building AgentXDP, part 13. Part 12 gave the testbed four links. This part gives it a stopwatch. In part 6, 20 ms of ACK delay left every RDMA error counter at zero. A delay shows up in one place: how 5 min read · [pegasus-nathanael-tech.hashnode.dev](https://pegasus-nathanael-tech.hashnode.dev/busy-polling-or-events-timing-one-rdma-write-and-why-perftest-says-half?utm_source=hashnode&utm_medium=feed) [3](https://hashnode.com/posts/busy-polling-or-events-timing-one-rdma-write-and-why-perftest-says-half/6af159d1709e0bfcfc8f45f1) [Nathanael FETUE](https://hashnode.com/@nathanael-fetue) · [Nov 5](https://hashnode.com/posts/four-rails-on-one-vm-a-multi-link-rdma-testbed-without-losing-ssh/6aebe0b832bf0a32a58bea5b) ### [Four rails on one VM: a multi-link RDMA testbed without losing SSH](https://pegasus-nathanael-tech.hashnode.dev/four-rails-on-one-vm-a-multi-link-rdma-testbed-without-losing-ssh?utm_source=hashnode&utm_medium=feed) One NIC per rail, four rails on one VM: computed names, a priority collision a test caught, and one slow link that sped the others up. 5 min read · [pegasus-nathanael-tech.hashnode.dev](https://pegasus-nathanael-tech.hashnode.dev/four-rails-on-one-vm-a-multi-link-rdma-testbed-without-losing-ssh?utm_source=hashnode&utm_medium=feed) [Weiwei Zhang](https://hashnode.com/@aicplight) · [Nov 4](https://hashnode.com/posts/1-6t-ethernet-vs-infiniband-a-technical-guide-to-next-generation-ai-network-architecture/6aeafe0dc22f616c288f1fe5) ### [1.6T Ethernet vs. InfiniBand: A Technical Guide to Next-Generation AI Network Architecture](https://aicplight.hashnode.dev/1-6t-ethernet-vs-infiniband-a-technical-guide-to-next-generation-ai-network-architecture?utm_source=hashnode&utm_medium=feed) Explore 1.6T Ethernet vs. InfiniBand for AI networking. Learn about RoCEv2, RDMA, PFC, ECN, DCQCN, GPU communication, and optical connectivity. 10 min read · [aicplight.hashnode.dev](https://aicplight.hashnode.dev/1-6t-ethernet-vs-infiniband-a-technical-guide-to-next-generation-ai-network-architecture?utm_source=hashnode&utm_medium=feed) [Nathanael FETUE](https://hashnode.com/@nathanael-fetue) · [Nov 1](https://hashnode.com/posts/week-2-from-a-sensor-to-an-agent/6ae7cce438877bcd79edc645) ### [Week 2: from a sensor to an agent](https://pegasus-nathanael-tech.hashnode.dev/week-2-from-a-sensor-to-an-agent?utm_source=hashnode&utm_medium=feed) A week turning one XDP sensor into an agent: what a host can really measure on RDMA, the parts it took, and what is still not proven. 4 min read · [pegasus-nathanael-tech.hashnode.dev](https://pegasus-nathanael-tech.hashnode.dev/week-2-from-a-sensor-to-an-agent?utm_source=hashnode&utm_medium=feed) [2](https://hashnode.com/posts/week-2-from-a-sensor-to-an-agent/6ae7cce438877bcd79edc645) [Nathanael FETUE](https://hashnode.com/@nathanael-fetue) · [Oct 30](https://hashnode.com/posts/attaching-xdp-without-leaving-a-trace-bpf-link-kill-9-and-a-signal-mask-nobody-tests/6ae44cb13538dc023f94f887) ### [Attaching XDP without leaving a trace: bpf\_link, kill -9 and a signal mask nobody tests](https://pegasus-nathanael-tech.hashnode.dev/attaching-xdp-without-leaving-a-trace-bpf-link-kill-9-and-a-signal-mask-nobody-tests?utm_source=hashnode&utm_medium=feed) Building AgentXDP, part 9. Previous: part 8, a lock-free queue broken on purpose. Next: part 10, a metrics endpoint a slow client cannot stall. An observability agent must do no harm. If it crashes, o 4 min read · [pegasus-nathanael-tech.hashnode.dev](https://pegasus-nathanael-tech.hashnode.dev/attaching-xdp-without-leaving-a-trace-bpf-link-kill-9-and-a-signal-mask-nobody-tests?utm_source=hashnode&utm_medium=feed) [Weiwei Zhang](https://hashnode.com/@aicplight) · [Oct 13](https://hashnode.com/posts/how-infiniband-xdr-and-800g-roce-handle-congestion-in-ai-clusters/6acdf7daecdd61e07179fde1) ### [How InfiniBand XDR and 800G RoCE Handle Congestion in AI Clusters](https://aicplight.hashnode.dev/how-infiniband-xdr-and-800g-roce-handle-congestion-in-ai-clusters?utm_source=hashnode&utm_medium=feed) Explore how InfiniBand XDR and 800G RoCE manage network congestion in AI clusters. Learn how RDMA, PFC, ECN, and tail latency affect distributed training performance, scalability, and network design. 6 min read · [aicplight.hashnode.dev](https://aicplight.hashnode.dev/how-infiniband-xdr-and-800g-roce-handle-congestion-in-ai-clusters?utm_source=hashnode&utm_medium=feed) [Nathanael FETUE](https://hashnode.com/@nathanael-fetue) · [Oct 7](https://hashnode.com/posts/my-roce-testbed-passed-it-was-lying/6ac5a323a97d85f79be4f6e2) ### [My RoCE testbed passed. It was lying](https://pegasus-nathanael-tech.hashnode.dev/my-roce-testbed-passed-it-was-lying?utm_source=hashnode&utm_medium=feed) A RoCE testbed passed its test while no RDMA write completed. How bpftrace on kfree\_skb found the firewall dropping every packet. 6 min read · [pegasus-nathanael-tech.hashnode.dev](https://pegasus-nathanael-tech.hashnode.dev/my-roce-testbed-passed-it-was-lying?utm_source=hashnode&utm_medium=feed) [1](https://hashnode.com/posts/my-roce-testbed-passed-it-was-lying/6ac5a323a97d85f79be4f6e2) [Weiwei Zhang](https://hashnode.com/@aicplight) · [Aug 18](https://hashnode.com/posts/roce-explained-how-rdma-over-ethernet-enables-scalable-ai-inference/6a841b702514a98192246b72) ### [RoCE Explained: How RDMA over Ethernet Enables Scalable AI Inference](https://aicplight.hashnode.dev/roce-explained-how-rdma-over-ethernet-enables-scalable-ai-inference?utm_source=hashnode&utm_medium=feed) Explore RoCE v2 networks with RDMA, lossless Ethernet, RoCE adapters, switches, and InfiniBand comparisons for AI inference, cloud, and HPC scenarios. 6 min read · [aicplight.hashnode.dev](https://aicplight.hashnode.dev/roce-explained-how-rdma-over-ethernet-enables-scalable-ai-inference?utm_source=hashnode&utm_medium=feed) [ania0-art](https://hashnode.com/@6371393) · [Jul 26](https://hashnode.com/posts/0-let-s-enjoy-reading-the-paper/6a65c54b7ff946909ee33285) ### [0: Let's enjoy reading the paper!](https://once2twice.hashnode.dev/0-let-s-enjoy-reading-the-paper?utm_source=hashnode&utm_medium=feed) This blog post is for sharing the research paper only and does not involve specific technical details. 9 min read · [once2twice.hashnode.dev](https://once2twice.hashnode.dev/0-let-s-enjoy-reading-the-paper?utm_source=hashnode&utm_medium=feed)