Reference Guide

Technical support and resources
5 Dell EMC XC Series Appliance and XC Core System Remote Direct Memory Access Deployment Guide
1 Introduction
This document describes the deployment of the Remote Direct Memory Access (RDMA) solution.
1.1 Remote Direct Memory Access
RDMA increases bandwidth while lowering latency and CPU utilization. It sidesteps the system software stack
components that process network traffic.
1.1.1 RDMA benefits
This section describes RDMA benefits.
1.1.1.1 Zero-Copy
Applications can perform data transfers without the involvement of the network software stack. Data is sent
and received directly to the network card buffers without being copied between the network layers, bypassing
TCP, which means that there is zero work to be done by the CPU; no caching, no context switching, resulting
in lower overall load on the system as this frees the CPU(s) for other workloads.
1.1.1.2 Kernel bypass
Applications can perform data transfers directly from user-space without kernel involvement. Kernel packets
are offloaded for the transport layer to the NIC; packet drops are addressed by using a lossless network using
Priority-based Flow Control (PFC).
Applications access remote memory without consuming any CPU time in the remote server. The remote
memory server (process) is read without any intervention from the remote processor. Likewise, the remote
CPU cache(s) are not filled with the accessed memory content.
Supported platforms
The following are supported platforms:
• XC740-24 with NVMe
o Intel rNDC and 2xCX-4 PCIe Adapters.
• XC940-24 with NVMe
o Intel rNDC and 2xCX-4 PCIe Adapters
NOTE:
Prior to Foundation 4.4.1, the following applies:
• Two CX4 PCIe adaptors are required.
• CX4 rNDC cannot be mixed with a CX4 adapter.
• Use the Intel rNDC slot to populate the rNDC slot.
rNDC cannot be used for RDMA.
1.1.2 Supported hypervisors
The following is the supported hypervisor:
• AHV version 20170830.184
• ESXi 6.7 U2 Build #13006603