Vxrail is based on vsan so even without a official document, we have some best practices for it. Because of the specific integration with vsan, we suggest:
- virtual proxies to read data locally from each host via hotadd
- ideally one proxy per host if the size of a single node is big enough to justify this
-Drs rules to stick each proxy to its node
With this design in place, Veeam reads placement data of each vmdk (it doesn't matter where the vm is running), and selects the proxy that has local access to the majority of disk data. In this way, data is read locally from the node without useless network traffic consumption.
EMEA Cloud Architect @ Veeam Software
Veeam VMCE #1