What is the approximate latency between two BM.HPC2.36 nodes using RDMA on the same rack?

The approximate latency between two BM.HPC2.36 nodes using RDMA (Remote Direct Memory Access) on the same rack is around 1.7 microseconds. RDMA technology allows direct memory access from the memory of one computer to another without involving the operating system, which significantly reduces latency and increases throughput for data-intensive applications commonly found in high-performance computing (HPC) environments.

The choice of 1.7 microseconds aligns well with expected performance metrics from RDMA in high-bandwidth environments, especially when nodes are located in close proximity, such as within the same rack. This low latency is critical for applications that require rapid data exchange and tight synchronization between compute nodes, which is a common scenario in HPC workloads.

In contrast, the other latency values provided do not accurately reflect the typical performance characteristics achievable with RDMA in a well-optimized environment with high-quality networking hardware. The low latency observed in real RDMA implementations contributes to the overall efficiency and performance of distributed computing tasks, particularly in settings where every microsecond counts.