maoravni

So that means I can't use the cache NVME and the entire data for the share must occupy a single drive of my array? So I get slow write performance AND I have to manage my data, which I thought I could skip on Unraid?

I've suspected this for quite a while, so I conducted a small test. I took a folder with ~95K files, ~24GB and copied to and from my Unraid server and to and from an old NAS I have. The Unraid server has AMD Ryzen 5 5600G, 16GB RAM. I also have an NVME cache drive. The old NAS has Atom D525, 4GB RAM, and 5 3TB drives, running latest OpenMediaVault. All machines are connected to the same switch, 1GB ethernet. The source machine is an Intel I7 6700T with a SATA SSD. I copied the files to the NAS, to the Unraid server using the standard shares, where the files will be stored on the cache drive, and to the Unraid server to a direct share to cache drive. The copy operation was performed with Robocopy using 4 threads. Times: Local -> Unraid user share: 23m Local -> Old NAS: 19m Local -> Unraid direct share: 6m Unraid user share -> Local: 10m Old NAS -> Local: 8m Unraid direct share -> Local: 6m This explains why I get so low performance running applications that store data directly on the network shares. I know that Unraid adds latency to network operations because of shfs, and I don't expect write operations to be as fast as RAID-5, and that's why I added the NVME drive, but I didn't expect this. Is there any way to reduce the latency (that does not include direct mounting, as that defeats the purpose) ?