This is all about making sure there's an understanding of what is the lowest latency path to the storage system. Meaning that let;s say you have 2 datacenters, DC-A and DC-B, and you have a stretched compute across DC-A and DC-B and you have stretched ...