Can someone please explain to me how this scenario works in sheepdog? Suppose I have 12 nodes (either 12 servers with one dedicated sheep disk each, or 4 servers with 3 sheep disks in each) and I want 3 copies of my data to be maintained. Is it possible that none of my data will reside on the host which creates/accesses it or does sheepdog ensure the data has an affinity for the host that uses it? If it has a complete local copy does it still retrieve remote copies? Hope this makes sense. I'm just interested in the efficiency of VM storage access in sheepdog. If I can understand it then I will add this as a FAQ item for proposed user documentation. Thanks, Matt. |