Is there any recommended setup for Valkey+RDMA to fully utilized interconnect bandwidth?

[taekyounghan]

Hello, I'm new to Valkey and am interested in its utilization of network devices.

As shown below, I have two connected nodes with identical setup for the Valkey server and client, respectively

             CPU (intel i7 @ 2.90GHz)  -----------PCIe gen 3 x16 (128Gbps) --------NIC (100Gbps,NVIDIA ConnectX6-Dx, RDMA enabled)

I've tried some benchmark tests and tried to monitor the bottleneck in PCIe.

However, its utilization is up to 30Gbps.

In my current benchmarking setup, I'm not observing full PCIe or network saturation.

Could this be due to CPU bottlenecks, or is my workload insufficiently demanding?

Are there any known references or profiling data that quantify the CPU resources needed to drive, for example, 100 Gbps RDMA or PCIe traffic?

Any comments will be greatly helpful.

Best regards,
Taekyoung