I’ve been working with RDMA/EFA while benchmarking distributed training workloads, and found that there isn’t really a good TUI tool (like htop/iftop) to monitor RDMA traffic. While rdma statistic exposes counters, it’s not very intuitive for real-time debugging.
So I built rdmatop, a small netlink-based TUI to visualize live RDMA RX/TX activity. It’s been useful for troubleshooting performance issues in RDMA-heavy workloads (e.g., multi-node training).
[link] [comments]
SOCIAL SHARE CARD GENERATOR